maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/lib/Headers/avxintrin.h

5068 lines
194 KiB
C
Raw Normal View History

Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/*===---- avxintrin.h - AVX intrinsics -------------------------------------===
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*
*===-----------------------------------------------------------------------===
*/
Disallow direct inclusion of avxintrin.h. Users should include immintrin.h instead. This matches GCC's behavior. llvm-svn: 111692
2010-08-21 07:00:03 +08:00
#ifndef __IMMINTRIN_H
#error "Never use <avxintrin.h> directly; include <immintrin.h> instead."
#endif
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Add missing include guards into headers in lib/Headers. While it may appear that these headers should not be included more than once, they are in fact included twice when building our builtins module (in order for it to generate submodules for them), and without this, any modular build enabling AVX and including any builtin header fails. Testing this is tricky because including any of these headers in a modular build is liable to fail, due to unrelated builtin headers in the same module including headers which might not be available on the system running the tests. Suggestion on that front are welcome (but we're getting close to being able to run a buildbot that has modules enabled for all tests, which would nicely solve the testing problem). llvm-svn: 186275
2013-07-14 13:41:45 +08:00
#ifndef __AVXINTRIN_H
#define __AVXINTRIN_H
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef double __v4df __attribute__ ((__vector_size__ (32)));
typedef float __v8sf __attribute__ ((__vector_size__ (32)));
typedef long long __v4di __attribute__ ((__vector_size__ (32)));
typedef int __v8si __attribute__ ((__vector_size__ (32)));
typedef short __v16hi __attribute__ ((__vector_size__ (32)));
typedef char __v32qi __attribute__ ((__vector_size__ (32)));
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
/* Unsigned types */
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef unsigned long long __v4du __attribute__ ((__vector_size__ (32)));
typedef unsigned int __v8su __attribute__ ((__vector_size__ (32)));
typedef unsigned short __v16hu __attribute__ ((__vector_size__ (32)));
typedef unsigned char __v32qu __attribute__ ((__vector_size__ (32)));
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
Fix the SSE4 byte sign extension in a cleaner way, and more thoroughly test that our intrinsics behave the same under -fsigned-char and -funsigned-char. This further testing uncovered that AVX-2 has a broken cmpgt for 8-bit elements, and has for a long time. This is fixed in the same way as SSE4 handles the case. The other ISA extensions currently work correctly because they use specific instruction intrinsics. As soon as they are rewritten in terms of generic IR, they will need to add these special casts. I've added the necessary testing to catch this however, so we shouldn't have to chase it down again. I considered changing the core typedef to be signed, but that seems like a bad idea. Notably, it would be an ABI break if anyone is reaching into the innards of the intrinsic headers and passing __v16qi on an API boundary. I can't be completely confident that this wouldn't happen due to a macro expanding in a lambda, etc., so it seems much better to leave it alone. It also matches GCC's behavior exactly. A fun side note is that for both GCC and Clang, -funsigned-char really does change the semantics of __v16qi. To observe this, consider: % cat x.cc #include <smmintrin.h> #include <iostream> int main() { __v16qi a = { 1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; __v16qi b = _mm_set1_epi8(-1); std::cout << (int)(a / b)[0] << ", " << (int)(a / b)[1] << '\n'; } % clang++ -o x x.cc && ./x -1, 1 % clang++ -funsigned-char -o x x.cc && ./x 0, 1 However, while this may be surprising, both Clang and GCC agree. Differential Revision: http://reviews.llvm.org/D13324 llvm-svn: 249097
2015-10-02 07:40:12 +08:00
/* We need an explicitly signed variant for char. Note that this shouldn't
* appear in the interface though. */
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef signed char __v32qs __attribute__((__vector_size__(32)));
Fix the SSE4 byte sign extension in a cleaner way, and more thoroughly test that our intrinsics behave the same under -fsigned-char and -funsigned-char. This further testing uncovered that AVX-2 has a broken cmpgt for 8-bit elements, and has for a long time. This is fixed in the same way as SSE4 handles the case. The other ISA extensions currently work correctly because they use specific instruction intrinsics. As soon as they are rewritten in terms of generic IR, they will need to add these special casts. I've added the necessary testing to catch this however, so we shouldn't have to chase it down again. I considered changing the core typedef to be signed, but that seems like a bad idea. Notably, it would be an ABI break if anyone is reaching into the innards of the intrinsic headers and passing __v16qi on an API boundary. I can't be completely confident that this wouldn't happen due to a macro expanding in a lambda, etc., so it seems much better to leave it alone. It also matches GCC's behavior exactly. A fun side note is that for both GCC and Clang, -funsigned-char really does change the semantics of __v16qi. To observe this, consider: % cat x.cc #include <smmintrin.h> #include <iostream> int main() { __v16qi a = { 1, -1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; __v16qi b = _mm_set1_epi8(-1); std::cout << (int)(a / b)[0] << ", " << (int)(a / b)[1] << '\n'; } % clang++ -o x x.cc && ./x -1, 1 % clang++ -funsigned-char -o x x.cc && ./x 0, 1 However, while this may be surprising, both Clang and GCC agree. Differential Revision: http://reviews.llvm.org/D13324 llvm-svn: 249097
2015-10-02 07:40:12 +08:00
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
typedef float __m256 __attribute__ ((__vector_size__ (32), __aligned__(32)));
typedef double __m256d __attribute__((__vector_size__(32), __aligned__(32)));
typedef long long __m256i __attribute__((__vector_size__(32), __aligned__(32)));
typedef float __m256_u __attribute__ ((__vector_size__ (32), __aligned__(1)));
typedef double __m256d_u __attribute__((__vector_size__(32), __aligned__(1)));
typedef long long __m256i_u __attribute__((__vector_size__(32), __aligned__(1)));
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Use a define for per-file function attributes for the Intel intrinsic headers. This is a precursor to changing them to use the new target attribute code. llvm-svn: 239882
2015-06-17 15:09:20 +08:00
/* Define the default attributes for the functions in this file. */
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("avx"), __min_vector_width__(256)))
#define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("avx"), __min_vector_width__(128)))
Use a define for per-file function attributes for the Intel intrinsic headers. This is a precursor to changing them to use the new target attribute code. llvm-svn: 239882
2015-06-17 15:09:20 +08:00
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/* Arithmetic */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Adds two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VADDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the sums of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_add_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256d)((__v4df)__a+(__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Adds two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VADDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the sums of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_add_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256)((__v8sf)__a+(__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Subtracts two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSUBPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing the minuend.
/// \param __b
/// A 256-bit vector of [4 x double] containing the subtrahend.
/// \returns A 256-bit vector of [4 x double] containing the differences between
/// both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_sub_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256d)((__v4df)__a-(__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Subtracts two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSUBPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing the minuend.
/// \param __b
/// A 256-bit vector of [8 x float] containing the subtrahend.
/// \returns A 256-bit vector of [8 x float] containing the differences between
/// both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_sub_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256)((__v8sf)__a-(__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Adds the even-indexed values and subtracts the odd-indexed values of
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// two 256-bit vectors of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VADDSUBPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing the left source operand.
/// \param __b
/// A 256-bit vector of [4 x double] containing the right source operand.
/// \returns A 256-bit vector of [4 x double] containing the alternating sums
/// and differences between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_addsub_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_addsubpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Adds the even-indexed values and subtracts the odd-indexed values of
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// two 256-bit vectors of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VADDSUBPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing the left source operand.
/// \param __b
/// A 256-bit vector of [8 x float] containing the right source operand.
/// \returns A 256-bit vector of [8 x float] containing the alternating sums and
/// differences between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_addsub_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_addsubps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Divides two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VDIVPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing the dividend.
/// \param __b
/// A 256-bit vector of [4 x double] containing the divisor.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [4 x double] containing the quotients of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_div_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256d)((__v4df)__a/(__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Divides two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VDIVPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing the dividend.
/// \param __b
/// A 256-bit vector of [8 x float] containing the divisor.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [8 x float] containing the quotients of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_div_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256)((__v8sf)__a/(__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares two 256-bit vectors of [4 x double] and returns the greater
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// of each pair of values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMAXPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the operands.
/// \returns A 256-bit vector of [4 x double] containing the maximum values
/// between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_max_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_maxpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares two 256-bit vectors of [8 x float] and returns the greater
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// of each pair of values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMAXPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the operands.
/// \returns A 256-bit vector of [8 x float] containing the maximum values
/// between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_max_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_maxps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares two 256-bit vectors of [4 x double] and returns the lesser
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// of each pair of values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMINPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the operands.
/// \returns A 256-bit vector of [4 x double] containing the minimum values
/// between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_min_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_minpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares two 256-bit vectors of [8 x float] and returns the lesser
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// of each pair of values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMINPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the operands.
/// \returns A 256-bit vector of [8 x float] containing the minimum values
/// between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_min_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_minps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Multiplies two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMULPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the operands.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [4 x double] containing the products of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_mul_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256d)((__v4df)__a * (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Multiplies two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMULPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the operands.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [8 x float] containing the products of both
/// operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_mul_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256)((__v8sf)__a * (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Calculates the square roots of the values in a 256-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSQRTPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the square roots of the
/// values in the operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_sqrt_pd(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_sqrtpd256((__v4df)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Calculates the square roots of the values in a 256-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSQRTPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the square roots of the
/// values in the operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_sqrt_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_sqrtps256((__v8sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Calculates the reciprocal square roots of the values in a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// vector of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VRSQRTPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the reciprocal square
/// roots of the values in the operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_rsqrt_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_rsqrtps256((__v8sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Calculates the reciprocals of the values in a 256-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VRCPPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the reciprocals of the
/// values in the operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_rcp_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_rcpps256((__v8sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds the values in a 256-bit vector of [4 x double] as specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// by the byte operand. The source values are rounded to integer values and
/// returned as 64-bit double-precision floating-point values.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_round_pd(__m256d V, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [4 x double].
/// \param M
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// An integer value that specifies the rounding operation. \n
/// Bits [7:4] are reserved. \n
/// Bit [3] is a precision exception value: \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 0: A normal PE exception is used. \n
/// 1: The PE field is not updated. \n
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// Bit [2] is the rounding control source: \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 0: Use bits [1:0] of \a M. \n
/// 1: Use the current MXCSR setting. \n
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// Bits [1:0] contain the rounding control definition: \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 00: Nearest. \n
/// 01: Downward (toward negative infinity). \n
/// 10: Upward (toward positive infinity). \n
/// 11: Truncated.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [4 x double] containing the rounded values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_round_pd(V, M) \
(__m256d)__builtin_ia32_roundpd256((__v4df)(__m256d)(V), (M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds the values stored in a 256-bit vector of [8 x float] as
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// specified by the byte operand. The source values are rounded to integer
/// values and returned as floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_round_ps(__m256 V, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [8 x float].
/// \param M
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// An integer value that specifies the rounding operation. \n
/// Bits [7:4] are reserved. \n
/// Bit [3] is a precision exception value: \n
/// 0: A normal PE exception is used. \n
/// 1: The PE field is not updated. \n
/// Bit [2] is the rounding control source: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 0: Use bits [1:0] of \a M. \n
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// 1: Use the current MXCSR setting. \n
/// Bits [1:0] contain the rounding control definition: \n
/// 00: Nearest. \n
/// 01: Downward (toward negative infinity). \n
/// 10: Upward (toward positive infinity). \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 11: Truncated.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [8 x float] containing the rounded values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_round_ps(V, M) \
(__m256)__builtin_ia32_roundps256((__v8sf)(__m256)(V), (M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds up the values stored in a 256-bit vector of [4 x double]. The
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// source values are rounded up to integer values and returned as 64-bit
/// double-precision floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_ceil_pd(__m256d V);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the rounded up values.
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _mm256_ceil_pd(V) _mm256_round_pd((V), _MM_FROUND_CEIL)
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds down the values stored in a 256-bit vector of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// The source values are rounded down to integer values and returned as
/// 64-bit double-precision floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_floor_pd(__m256d V);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the rounded down
/// values.
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _mm256_floor_pd(V) _mm256_round_pd((V), _MM_FROUND_FLOOR)
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds up the values stored in a 256-bit vector of [8 x float]. The
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// source values are rounded up to integer values and returned as
/// floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_ceil_ps(__m256 V);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the rounded up values.
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _mm256_ceil_ps(V) _mm256_round_ps((V), _MM_FROUND_CEIL)
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Rounds down the values stored in a 256-bit vector of [8 x float]. The
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// source values are rounded down to integer values and returned as
/// floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_floor_ps(__m256 V);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VROUNDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param V
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the rounded down values.
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _mm256_floor_ps(V) _mm256_round_ps((V), _MM_FROUND_FLOOR)
/* Logical */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise AND of two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VANDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise AND of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_and_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256d)((__v4du)__a & (__v4du)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise AND of two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VANDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise AND of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_and_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256)((__v8su)__a & (__v8su)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise AND of two 256-bit vectors of [4 x double], using
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// the one's complement of the values contained in the first source operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VANDNPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing the left source operand. The
/// one's complement of this value is used in the bitwise AND.
/// \param __b
/// A 256-bit vector of [4 x double] containing the right source operand.
/// \returns A 256-bit vector of [4 x double] containing the bitwise AND of the
/// values of the second operand and the one's complement of the first
/// operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_andnot_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256d)(~(__v4du)__a & (__v4du)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise AND of two 256-bit vectors of [8 x float], using
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// the one's complement of the values contained in the first source operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VANDNPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing the left source operand. The
/// one's complement of this value is used in the bitwise AND.
/// \param __b
/// A 256-bit vector of [8 x float] containing the right source operand.
/// \returns A 256-bit vector of [8 x float] containing the bitwise AND of the
/// values of the second operand and the one's complement of the first
/// operand.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_andnot_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256)(~(__v8su)__a & (__v8su)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise OR of two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VORPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise OR of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_or_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256d)((__v4du)__a | (__v4du)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise OR of two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VORPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise OR of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_or_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256)((__v8su)__a | (__v8su)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise XOR of two 256-bit vectors of [4 x double].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VXORPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// \returns A 256-bit vector of [4 x double] containing the bitwise XOR of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_xor_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256d)((__v4du)__a ^ (__v4du)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Performs a bitwise XOR of two 256-bit vectors of [8 x float].
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VXORPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// \returns A 256-bit vector of [8 x float] containing the bitwise XOR of the
/// values between both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_xor_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use unsigned types for vector arithmetic in intrinsics to avoid undefined behavior for signed integer overflow. This is really only needed for addition, subtraction, and multiplication, but I did the bitwise ops too for overall consistency. Clang currently doesn't set NSW for signed vector operations so the undefined behavior shouldn't happen today. llvm-svn: 271778
2016-06-04 13:43:41 +08:00
return (__m256)((__v8su)__a ^ (__v8su)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Horizontal arithmetic */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Horizontally adds the adjacent pairs of values contained in two
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// 256-bit vectors of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VHADDPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// The horizontal sums of the values are returned in the even-indexed
/// elements of a vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// The horizontal sums of the values are returned in the odd-indexed
/// elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal sums of
/// both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_hadd_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_haddpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Horizontally adds the adjacent pairs of values contained in two
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// 256-bit vectors of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VHADDPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// The horizontal sums of the values are returned in the elements with
/// index 0, 1, 4, 5 of a vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// The horizontal sums of the values are returned in the elements with
/// index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal sums of
/// both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_hadd_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_haddps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Horizontally subtracts the adjacent pairs of values contained in two
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// 256-bit vectors of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VHSUBPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// The horizontal differences between the values are returned in the
/// even-indexed elements of a vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double] containing one of the source operands.
/// The horizontal differences between the values are returned in the
/// odd-indexed elements of a vector of [4 x double].
/// \returns A 256-bit vector of [4 x double] containing the horizontal
/// differences of both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_hsub_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_hsubpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Horizontally subtracts the adjacent pairs of values contained in two
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// 256-bit vectors of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VHSUBPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// The horizontal differences between the values are returned in the
/// elements with index 0, 1, 4, 5 of a vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float] containing one of the source operands.
/// The horizontal differences between the values are returned in the
/// elements with index 2, 3, 6, 7 of a vector of [8 x float].
/// \returns A 256-bit vector of [8 x float] containing the horizontal
/// differences of both operands.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_hsub_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256)__builtin_ia32_hsubps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Vector permutations */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 128-bit vector of [2 x double] as specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// by the 128-bit integer vector operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \param __c
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 128-bit integer vector operand specifying how the values are to be
/// copied. \n
/// Bit [1]: \n
/// 0: Bits [63:0] of the source are copied to bits [63:0] of the returned
/// vector. \n
/// 1: Bits [127:64] of the source are copied to bits [63:0] of the
/// returned vector. \n
/// Bit [65]: \n
/// 0: Bits [63:0] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// 1: Bits [127:64] of the source are copied to bits [127:64] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 128-bit vector of [2 x double] containing the copied values.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline __m128d __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_permutevar_pd(__m128d __a, __m128i __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m128d)__builtin_ia32_vpermilvarpd((__v2df)__a, (__v2di)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 256-bit vector of [4 x double] as specified
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// by the 256-bit integer vector operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \param __c
/// A 256-bit integer vector operand specifying how the values are to be
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// copied. \n
/// Bit [1]: \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 0: Bits [63:0] of the source are copied to bits [63:0] of the returned
/// vector. \n
/// 1: Bits [127:64] of the source are copied to bits [63:0] of the
/// returned vector. \n
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// Bit [65]: \n
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// 0: Bits [63:0] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// 1: Bits [127:64] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// Bit [129]: \n
/// 0: Bits [191:128] of the source are copied to bits [191:128] of the
/// returned vector. \n
/// 1: Bits [255:192] of the source are copied to bits [191:128] of the
/// returned vector. \n
/// Bit [193]: \n
/// 0: Bits [191:128] of the source are copied to bits [255:192] of the
/// returned vector. \n
/// 1: Bits [255:192] of the source are copied to bits [255:192] of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [4 x double] containing the copied values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_permutevar_pd(__m256d __a, __m256i __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_vpermilvarpd256((__v4df)__a, (__v4di)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values stored in a 128-bit vector of [4 x float] as
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// specified by the 128-bit integer vector operand.
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \param __c
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 128-bit integer vector operand specifying how the values are to be
/// copied. \n
/// Bits [1:0]: \n
/// 00: Bits [31:0] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// Bits [33:32]: \n
/// 00: Bits [31:0] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// Bits [65:64]: \n
/// 00: Bits [31:0] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// Bits [97:96]: \n
/// 00: Bits [31:0] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [127:96] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 128-bit vector of [4 x float] containing the copied values.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline __m128 __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_permutevar_ps(__m128 __a, __m128i __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m128)__builtin_ia32_vpermilvarps((__v4sf)__a, (__v4si)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values stored in a 256-bit vector of [8 x float] as
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// specified by the 256-bit integer vector operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \param __c
/// A 256-bit integer vector operand specifying how the values are to be
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// copied. \n
/// Bits [1:0]: \n
/// 00: Bits [31:0] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// Bits [33:32]: \n
/// 00: Bits [31:0] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// Bits [65:64]: \n
/// 00: Bits [31:0] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// Bits [97:96]: \n
/// 00: Bits [31:0] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// Bits [129:128]: \n
/// 00: Bits [159:128] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// Bits [161:160]: \n
/// 00: Bits [159:128] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// Bits [193:192]: \n
/// 00: Bits [159:128] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// Bits [225:224]: \n
/// 00: Bits [159:128] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [255:224] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [8 x float] containing the copied values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_permutevar_ps(__m256 __a, __m256i __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
return (__m256)__builtin_ia32_vpermilvarps256((__v8sf)__a, (__v8si)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 128-bit vector of [2 x double] as specified
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// by the immediate integer operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_permute_pd(__m128d A, const int C);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param A
/// A 128-bit vector of [2 x double].
/// \param C
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// An immediate integer operand specifying how the values are to be
/// copied. \n
/// Bit [0]: \n
/// 0: Bits [63:0] of the source are copied to bits [63:0] of the returned
/// vector. \n
/// 1: Bits [127:64] of the source are copied to bits [63:0] of the
/// returned vector. \n
/// Bit [1]: \n
/// 0: Bits [63:0] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// 1: Bits [127:64] of the source are copied to bits [127:64] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 128-bit vector of [2 x double] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_permute_pd(A, C) \
[X86] Add builtins for vpermilps/pd instructions to enable target feature checking. llvm-svn: 334256
2018-06-08 08:59:27 +08:00
(__m128d)__builtin_ia32_vpermilpd((__v2df)(__m128d)(A), (int)(C))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 256-bit vector of [4 x double] as specified by
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// the immediate integer operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_permute_pd(__m256d A, const int C);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPD </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param A
/// A 256-bit vector of [4 x double].
/// \param C
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// An immediate integer operand specifying how the values are to be
/// copied. \n
/// Bit [0]: \n
/// 0: Bits [63:0] of the source are copied to bits [63:0] of the returned
/// vector. \n
/// 1: Bits [127:64] of the source are copied to bits [63:0] of the
/// returned vector. \n
/// Bit [1]: \n
/// 0: Bits [63:0] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// 1: Bits [127:64] of the source are copied to bits [127:64] of the
/// returned vector. \n
/// Bit [2]: \n
/// 0: Bits [191:128] of the source are copied to bits [191:128] of the
/// returned vector. \n
/// 1: Bits [255:192] of the source are copied to bits [191:128] of the
/// returned vector. \n
/// Bit [3]: \n
/// 0: Bits [191:128] of the source are copied to bits [255:192] of the
/// returned vector. \n
/// 1: Bits [255:192] of the source are copied to bits [255:192] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [4 x double] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_permute_pd(A, C) \
[X86] Add builtins for vpermilps/pd instructions to enable target feature checking. llvm-svn: 334256
2018-06-08 08:59:27 +08:00
(__m256d)__builtin_ia32_vpermilpd256((__v4df)(__m256d)(A), (int)(C))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 128-bit vector of [4 x float] as specified by
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// the immediate integer operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_permute_ps(__m128 A, const int C);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param A
/// A 128-bit vector of [4 x float].
/// \param C
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// An immediate integer operand specifying how the values are to be
/// copied. \n
/// Bits [1:0]: \n
/// 00: Bits [31:0] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// Bits [3:2]: \n
/// 00: Bits [31:0] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// Bits [5:4]: \n
/// 00: Bits [31:0] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// Bits [7:6]: \n
/// 00: Bits [31:0] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [127:96] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 128-bit vector of [4 x float] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_permute_ps(A, C) \
[X86] Add builtins for vpermilps/pd instructions to enable target feature checking. llvm-svn: 334256
2018-06-08 08:59:27 +08:00
(__m128)__builtin_ia32_vpermilps((__v4sf)(__m128)(A), (int)(C))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Copies the values in a 256-bit vector of [8 x float] as specified by
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// the immediate integer operand.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_permute_ps(__m256 A, const int C);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS </c> instruction.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
///
/// \param A
/// A 256-bit vector of [8 x float].
/// \param C
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// An immediate integer operand specifying how the values are to be
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// copied. \n
/// Bits [1:0]: \n
/// 00: Bits [31:0] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [31:0] of the
/// returned vector. \n
/// Bits [3:2]: \n
/// 00: Bits [31:0] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [63:32] of the
/// returned vector. \n
/// Bits [5:4]: \n
/// 00: Bits [31:0] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [95:64] of the
/// returned vector. \n
/// Bits [7:6]: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 00: Bits [31:0] of the source are copied to bits [127:96] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// returned vector. \n
/// 01: Bits [63:32] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 10: Bits [95:64] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// 11: Bits [127:96] of the source are copied to bits [127:96] of the
/// returned vector. \n
/// Bits [1:0]: \n
/// 00: Bits [159:128] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [159:128] of the
/// returned vector. \n
/// Bits [3:2]: \n
/// 00: Bits [159:128] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [191:160] of the
/// returned vector. \n
/// Bits [5:4]: \n
/// 00: Bits [159:128] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [223:192] of the
/// returned vector. \n
/// Bits [7:6]: \n
/// 00: Bits [159:128] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 01: Bits [191:160] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 10: Bits [223:192] of the source are copied to bits [255:224] of the
/// returned vector. \n
/// 11: Bits [255:224] of the source are copied to bits [255:224] of the
/// returned vector.
Add doxygen comments to avxintrin.h's intrinsics. Only around 25% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 263175
2016-03-11 08:05:54 +08:00
/// \returns A 256-bit vector of [8 x float] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_permute_ps(A, C) \
[X86] Add builtins for vpermilps/pd instructions to enable target feature checking. llvm-svn: 334256
2018-06-08 08:59:27 +08:00
(__m256)__builtin_ia32_vpermilps256((__v8sf)(__m256)(A), (int)(C))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Permutes 128-bit data values stored in two 256-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [4 x double], as specified by the immediate integer operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_permute2f128_pd(__m256d V1, __m256d V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERM2F128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A 256-bit vector of [4 x double].
/// \param V2
/// A 256-bit vector of [4 x double.
/// \param M
/// An immediate integer operand specifying how the values are to be
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// permuted. \n
/// Bits [1:0]: \n
/// 00: Bits [127:0] of operand \a V1 are copied to bits [127:0] of the
/// destination. \n
/// 01: Bits [255:128] of operand \a V1 are copied to bits [127:0] of the
/// destination. \n
/// 10: Bits [127:0] of operand \a V2 are copied to bits [127:0] of the
/// destination. \n
/// 11: Bits [255:128] of operand \a V2 are copied to bits [127:0] of the
/// destination. \n
/// Bits [5:4]: \n
/// 00: Bits [127:0] of operand \a V1 are copied to bits [255:128] of the
/// destination. \n
/// 01: Bits [255:128] of operand \a V1 are copied to bits [255:128] of the
/// destination. \n
/// 10: Bits [127:0] of operand \a V2 are copied to bits [255:128] of the
/// destination. \n
/// 11: Bits [255:128] of operand \a V2 are copied to bits [255:128] of the
/// destination.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [4 x double] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_permute2f128_pd(V1, V2, M) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256d)__builtin_ia32_vperm2f128_pd256((__v4df)(__m256d)(V1), \
[X86] Change immediate type for some builtins from char to int. These builtins are all handled by CGBuiltin.cpp so it doesn't much matter what the immediate type is, but int matches the intrinsic spec. llvm-svn: 334310
2018-06-09 02:00:22 +08:00
(__v4df)(__m256d)(V2), (int)(M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Permutes 128-bit data values stored in two 256-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [8 x float], as specified by the immediate integer operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_permute2f128_ps(__m256 V1, __m256 V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERM2F128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A 256-bit vector of [8 x float].
/// \param V2
/// A 256-bit vector of [8 x float].
/// \param M
/// An immediate integer operand specifying how the values are to be
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// permuted. \n
/// Bits [1:0]: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 00: Bits [127:0] of operand \a V1 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 01: Bits [255:128] of operand \a V1 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 10: Bits [127:0] of operand \a V2 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 11: Bits [255:128] of operand \a V2 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
/// Bits [5:4]: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 00: Bits [127:0] of operand \a V1 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 01: Bits [255:128] of operand \a V1 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 10: Bits [127:0] of operand \a V2 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 11: Bits [255:128] of operand \a V2 are copied to bits [255:128] of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// destination.
/// \returns A 256-bit vector of [8 x float] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_permute2f128_ps(V1, V2, M) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256)__builtin_ia32_vperm2f128_ps256((__v8sf)(__m256)(V1), \
[X86] Change immediate type for some builtins from char to int. These builtins are all handled by CGBuiltin.cpp so it doesn't much matter what the immediate type is, but int matches the intrinsic spec. llvm-svn: 334310
2018-06-09 02:00:22 +08:00
(__v8sf)(__m256)(V2), (int)(M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Permutes 128-bit data values stored in two 256-bit integer vectors,
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// as specified by the immediate integer operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256i _mm256_permute2f128_si256(__m256i V1, __m256i V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPERM2F128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A 256-bit integer vector.
/// \param V2
/// A 256-bit integer vector.
/// \param M
/// An immediate integer operand specifying how the values are to be copied.
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// Bits [1:0]: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 00: Bits [127:0] of operand \a V1 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 01: Bits [255:128] of operand \a V1 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 10: Bits [127:0] of operand \a V2 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 11: Bits [255:128] of operand \a V2 are copied to bits [127:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
/// Bits [5:4]: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 00: Bits [127:0] of operand \a V1 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 01: Bits [255:128] of operand \a V1 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 10: Bits [127:0] of operand \a V2 are copied to bits [255:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// 11: Bits [255:128] of operand \a V2 are copied to bits [255:128] of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// destination.
/// \returns A 256-bit integer vector containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_permute2f128_si256(V1, V2, M) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256i)__builtin_ia32_vperm2f128_si256((__v8si)(__m256i)(V1), \
[X86] Change immediate type for some builtins from char to int. These builtins are all handled by CGBuiltin.cpp so it doesn't much matter what the immediate type is, but int matches the intrinsic spec. llvm-svn: 334310
2018-06-09 02:00:22 +08:00
(__v8si)(__m256i)(V2), (int)(M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/* Vector Blend */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Merges 64-bit double-precision data values stored in either of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// two 256-bit vectors of [4 x double], as specified by the immediate
/// integer operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_blend_pd(__m256d V1, __m256d V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBLENDPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A 256-bit vector of [4 x double].
/// \param V2
/// A 256-bit vector of [4 x double].
/// \param M
/// An immediate integer operand, with mask bits [3:0] specifying how the
/// values are to be copied. The position of the mask bit corresponds to the
/// index of a copied value. When a mask bit is 0, the corresponding 64-bit
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// element in operand \a V1 is copied to the same position in the
/// destination. When a mask bit is 1, the corresponding 64-bit element in
/// operand \a V2 is copied to the same position in the destination.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [4 x double] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_blend_pd(V1, V2, M) \
[X86] Add builtins for blend with immediate control to enforce target feature requirements and check immediate range. llvm-svn: 334249
2018-06-08 08:00:21 +08:00
(__m256d)__builtin_ia32_blendpd256((__v4df)(__m256d)(V1), \
(__v4df)(__m256d)(V2), (int)(M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Merges 32-bit single-precision data values stored in either of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// two 256-bit vectors of [8 x float], as specified by the immediate
/// integer operand.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_blend_ps(__m256 V1, __m256 V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBLENDPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A 256-bit vector of [8 x float].
/// \param V2
/// A 256-bit vector of [8 x float].
/// \param M
/// An immediate integer operand, with mask bits [7:0] specifying how the
/// values are to be copied. The position of the mask bit corresponds to the
/// index of a copied value. When a mask bit is 0, the corresponding 32-bit
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// element in operand \a V1 is copied to the same position in the
/// destination. When a mask bit is 1, the corresponding 32-bit element in
/// operand \a V2 is copied to the same position in the destination.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [8 x float] containing the copied values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_blend_ps(V1, V2, M) \
[X86] Add builtins for blend with immediate control to enforce target feature requirements and check immediate range. llvm-svn: 334249
2018-06-08 08:00:21 +08:00
(__m256)__builtin_ia32_blendps256((__v8sf)(__m256)(V1), \
(__v8sf)(__m256)(V2), (int)(M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Merges 64-bit double-precision data values stored in either of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// two 256-bit vectors of [4 x double], as specified by the 256-bit vector
/// operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBLENDVPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double].
/// \param __c
/// A 256-bit vector operand, with mask bits 255, 191, 127, and 63 specifying
/// how the values are to be copied. The position of the mask bit corresponds
/// to the most significant bit of a copied value. When a mask bit is 0, the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// corresponding 64-bit element in operand \a __a is copied to the same
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// position in the destination. When a mask bit is 1, the corresponding
[DOXYGEN] Updated instruction names corresponding to avxintrin.h intrinsics. Documentation for some of the avxintrin.h's intrinsics errorneously said that non VEX-prefixed instructions could be generated. This was fixed. I tried several different solutions to achieve pretty printing of unordered lists (nested and non-nested) in param sections in doxygen. llvm-svn: 287990
2016-11-27 03:38:19 +08:00
/// 64-bit element in operand \a __b is copied to the same position in the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// destination.
/// \returns A 256-bit vector of [4 x double] containing the copied values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_blendv_pd(__m256d __a, __m256d __b, __m256d __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_blendvpd256(
(__v4df)__a, (__v4df)__b, (__v4df)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Merges 32-bit single-precision data values stored in either of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// two 256-bit vectors of [8 x float], as specified by the 256-bit vector
/// operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBLENDVPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float].
/// \param __c
/// A 256-bit vector operand, with mask bits 255, 223, 191, 159, 127, 95, 63,
/// and 31 specifying how the values are to be copied. The position of the
/// mask bit corresponds to the most significant bit of a copied value. When
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// a mask bit is 0, the corresponding 32-bit element in operand \a __a is
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// copied to the same position in the destination. When a mask bit is 1, the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// corresponding 32-bit element in operand \a __b is copied to the same
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// position in the destination.
/// \returns A 256-bit vector of [8 x float] containing the copied values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Readd an open paren that was lost while reformatting code. llvm-svn: 172669
2013-01-17 07:13:42 +08:00
return (__m256)__builtin_ia32_blendvps256(
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
(__v8sf)__a, (__v8sf)__b, (__v8sf)__c);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Vector Dot Product */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Computes two dot products in parallel, using the lower and upper
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// halves of two [8 x float] vectors as input to the two computations, and
/// returning the two dot products in the lower and upper halves of the
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// [8 x float] result.
///
/// The immediate integer operand controls which input elements will
/// contribute to the dot product, and where the final results are returned.
/// In general, for each dot product, the four corresponding elements of the
/// input vectors are multiplied; the first two and second two products are
/// summed, then the two sums are added to form the final result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_dp_ps(__m256 V1, __m256 V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VDPPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param V1
/// A vector of [8 x float] values, treated as two [4 x float] vectors.
/// \param V2
/// A vector of [8 x float] values, treated as two [4 x float] vectors.
/// \param M
/// An immediate integer argument. Bits [7:4] determine which elements of
/// the input vectors are used, with bit [4] corresponding to the lowest
/// element and bit [7] corresponding to the highest element of each [4 x
/// float] subvector. If a bit is set, the corresponding elements from the
/// two input vectors are used as an input for dot product; otherwise that
/// input is treated as zero. Bits [3:0] determine which elements of the
/// result will receive a copy of the final dot product, with bit [0]
/// corresponding to the lowest element and bit [3] corresponding to the
/// highest element of each [4 x float] subvector. If a bit is set, the dot
/// product is returned in the corresponding element; otherwise that element
/// is set to zero. The bitmask is applied in the same way to each of the
/// two parallel dot product computations.
/// \returns A 256-bit vector of [8 x float] containing the two dot products.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_dp_ps(V1, V2, M) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256)__builtin_ia32_dpps256((__v8sf)(__m256)(V1), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v8sf)(__m256)(V2), (M))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/* Vector shuffle */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Selects 8 float values from the 256-bit operands of [8 x float], as
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// specified by the immediate value operand.
///
/// The four selected elements in each operand are copied to the destination
/// according to the bits specified in the immediate operand. The selected
/// elements from the first 256-bit operand are copied to bits [63:0] and
/// bits [191:128] of the destination, and the selected elements from the
/// second 256-bit operand are copied to bits [127:64] and bits [255:192] of
/// the destination. For example, if bits [7:0] of the immediate operand
/// contain a value of 0xFF, the 256-bit destination vector would contain the
/// following values: b[7], b[7], a[7], a[7], b[3], b[3], a[3], a[3].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_shuffle_ps(__m256 a, __m256 b, const int mask);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSHUFPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 256-bit vector of [8 x float]. The four selected elements in this
/// operand are copied to bits [63:0] and bits [191:128] in the destination,
/// according to the bits specified in the immediate operand.
/// \param b
/// A 256-bit vector of [8 x float]. The four selected elements in this
/// operand are copied to bits [127:64] and bits [255:192] in the
/// destination, according to the bits specified in the immediate operand.
/// \param mask
/// An immediate value containing an 8-bit value specifying which elements to
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// copy from \a a and \a b \n.
/// Bits [3:0] specify the values copied from operand \a a. \n
/// Bits [7:4] specify the values copied from operand \a b. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// The destinations within the 256-bit destination are assigned values as
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// follows, according to the bit value assignments described below: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// Bits [1:0] are used to assign values to bits [31:0] and [159:128] in the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// Bits [3:2] are used to assign values to bits [63:32] and [191:160] in the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// Bits [5:4] are used to assign values to bits [95:64] and [223:192] in the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// Bits [7:6] are used to assign values to bits [127:96] and [255:224] in
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// the destination. \n
/// Bit value assignments: \n
/// 00: Bits [31:0] and [159:128] are copied from the selected operand. \n
/// 01: Bits [63:32] and [191:160] are copied from the selected operand. \n
/// 10: Bits [95:64] and [223:192] are copied from the selected operand. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// 11: Bits [127:96] and [255:224] are copied from the selected operand.
/// \returns A 256-bit vector of [8 x float] containing the shuffled values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_shuffle_ps(a, b, mask) \
[X86] Add builtins for shufps and shufpd to enable target feature and immediate range checking. llvm-svn: 334266
2018-06-08 15:18:33 +08:00
(__m256)__builtin_ia32_shufps256((__v8sf)(__m256)(a), \
(__v8sf)(__m256)(b), (int)(mask))
Remove 256-bit shuffle built-ins and make the AVX intrinsic call llvm __builtin_shufflevector with the appropriate arguments llvm-svn: 110766
2010-08-11 09:17:34 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Selects four double-precision values from the 256-bit operands of
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// [4 x double], as specified by the immediate value operand.
///
/// The selected elements from the first 256-bit operand are copied to bits
/// [63:0] and bits [191:128] in the destination, and the selected elements
/// from the second 256-bit operand are copied to bits [127:64] and bits
/// [255:192] in the destination. For example, if bits [3:0] of the immediate
/// operand contain a value of 0xF, the 256-bit destination vector would
/// contain the following values: b[3], a[3], b[1], a[1].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_shuffle_pd(__m256d a, __m256d b, const int mask);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VSHUFPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 256-bit vector of [4 x double].
/// \param b
/// A 256-bit vector of [4 x double].
/// \param mask
/// An immediate value containing 8-bit values specifying which elements to
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// copy from \a a and \a b: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [0]=0: Bits [63:0] are copied from \a a to bits [63:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [0]=1: Bits [127:64] are copied from \a a to bits [63:0] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [1]=0: Bits [63:0] are copied from \a b to bits [127:64] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [1]=1: Bits [127:64] are copied from \a b to bits [127:64] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [2]=0: Bits [191:128] are copied from \a a to bits [191:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [2]=1: Bits [255:192] are copied from \a a to bits [191:128] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [3]=0: Bits [191:128] are copied from \a b to bits [255:192] of the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// destination. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// Bit [3]=1: Bits [255:192] are copied from \a b to bits [255:192] of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// destination.
/// \returns A 256-bit vector of [4 x double] containing the shuffled values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_shuffle_pd(a, b, mask) \
[X86] Add builtins for shufps and shufpd to enable target feature and immediate range checking. llvm-svn: 334266
2018-06-08 15:18:33 +08:00
(__m256d)__builtin_ia32_shufpd256((__v4df)(__m256d)(a), \
(__v4df)(__m256d)(b), (int)(mask))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/* Compare */
#define _CMP_EQ_OQ 0x00 /* Equal (ordered, non-signaling) */
#define _CMP_LT_OS 0x01 /* Less-than (ordered, signaling) */
#define _CMP_LE_OS 0x02 /* Less-than-or-equal (ordered, signaling) */
#define _CMP_UNORD_Q 0x03 /* Unordered (non-signaling) */
#define _CMP_NEQ_UQ 0x04 /* Not-equal (unordered, non-signaling) */
#define _CMP_NLT_US 0x05 /* Not-less-than (unordered, signaling) */
#define _CMP_NLE_US 0x06 /* Not-less-than-or-equal (unordered, signaling) */
[x86] fix AVX FP cmp intrinsic documentation (PR28110) This copies the text used in the #define statements to the code comments. The conflicting text comes from AMD manuals, but those are wrong. Sadly, that FP cmp text has not been updated even after some docs were updated for Zen: http://support.amd.com/en-us/search/tech-docs ( AMD64 Architecture Programmer's Manual Volume 4 ) See PR28110 for more discussion: https://bugs.llvm.org/show_bug.cgi?id=28110 Differential Revision: https://reviews.llvm.org/D31428 llvm-svn: 300068
2017-04-12 23:19:08 +08:00
#define _CMP_ORD_Q 0x07 /* Ordered (non-signaling) */
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _CMP_EQ_UQ 0x08 /* Equal (unordered, non-signaling) */
[x86] fix AVX FP cmp intrinsic documentation (PR28110) This copies the text used in the #define statements to the code comments. The conflicting text comes from AMD manuals, but those are wrong. Sadly, that FP cmp text has not been updated even after some docs were updated for Zen: http://support.amd.com/en-us/search/tech-docs ( AMD64 Architecture Programmer's Manual Volume 4 ) See PR28110 for more discussion: https://bugs.llvm.org/show_bug.cgi?id=28110 Differential Revision: https://reviews.llvm.org/D31428 llvm-svn: 300068
2017-04-12 23:19:08 +08:00
#define _CMP_NGE_US 0x09 /* Not-greater-than-or-equal (unordered, signaling) */
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _CMP_NGT_US 0x0a /* Not-greater-than (unordered, signaling) */
#define _CMP_FALSE_OQ 0x0b /* False (ordered, non-signaling) */
#define _CMP_NEQ_OQ 0x0c /* Not-equal (ordered, non-signaling) */
#define _CMP_GE_OS 0x0d /* Greater-than-or-equal (ordered, signaling) */
#define _CMP_GT_OS 0x0e /* Greater-than (ordered, signaling) */
#define _CMP_TRUE_UQ 0x0f /* True (unordered, non-signaling) */
#define _CMP_EQ_OS 0x10 /* Equal (ordered, signaling) */
#define _CMP_LT_OQ 0x11 /* Less-than (ordered, non-signaling) */
#define _CMP_LE_OQ 0x12 /* Less-than-or-equal (ordered, non-signaling) */
#define _CMP_UNORD_S 0x13 /* Unordered (signaling) */
#define _CMP_NEQ_US 0x14 /* Not-equal (unordered, signaling) */
#define _CMP_NLT_UQ 0x15 /* Not-less-than (unordered, non-signaling) */
[x86] fix AVX FP cmp intrinsic documentation (PR28110) This copies the text used in the #define statements to the code comments. The conflicting text comes from AMD manuals, but those are wrong. Sadly, that FP cmp text has not been updated even after some docs were updated for Zen: http://support.amd.com/en-us/search/tech-docs ( AMD64 Architecture Programmer's Manual Volume 4 ) See PR28110 for more discussion: https://bugs.llvm.org/show_bug.cgi?id=28110 Differential Revision: https://reviews.llvm.org/D31428 llvm-svn: 300068
2017-04-12 23:19:08 +08:00
#define _CMP_NLE_UQ 0x16 /* Not-less-than-or-equal (unordered, non-signaling) */
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _CMP_ORD_S 0x17 /* Ordered (signaling) */
#define _CMP_EQ_US 0x18 /* Equal (unordered, signaling) */
[x86] fix AVX FP cmp intrinsic documentation (PR28110) This copies the text used in the #define statements to the code comments. The conflicting text comes from AMD manuals, but those are wrong. Sadly, that FP cmp text has not been updated even after some docs were updated for Zen: http://support.amd.com/en-us/search/tech-docs ( AMD64 Architecture Programmer's Manual Volume 4 ) See PR28110 for more discussion: https://bugs.llvm.org/show_bug.cgi?id=28110 Differential Revision: https://reviews.llvm.org/D31428 llvm-svn: 300068
2017-04-12 23:19:08 +08:00
#define _CMP_NGE_UQ 0x19 /* Not-greater-than-or-equal (unordered, non-signaling) */
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#define _CMP_NGT_UQ 0x1a /* Not-greater-than (unordered, non-signaling) */
#define _CMP_FALSE_OS 0x1b /* False (ordered, signaling) */
#define _CMP_NEQ_OS 0x1c /* Not-equal (ordered, signaling) */
#define _CMP_GE_OQ 0x1d /* Greater-than-or-equal (ordered, non-signaling) */
#define _CMP_GT_OQ 0x1e /* Greater-than (ordered, non-signaling) */
#define _CMP_TRUE_US 0x1f /* True (unordered, signaling) */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding double-precision values of two
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// 128-bit vectors of [2 x double], using the operation specified by the
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// immediate integer operand.
///
/// Returns a [2 x double] vector consisting of two doubles corresponding to
/// the two comparison results: zero if the comparison is false, and all 1's
/// if the comparison is true.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_cmp_pd(__m128d a, __m128d b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 128-bit vector of [2 x double].
/// \param b
/// A 128-bit vector of [2 x double].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_cmp_pd(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m128d)__builtin_ia32_cmppd((__v2df)(__m128d)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v2df)(__m128d)(b), (c))
Fix vector macros to correctly check argument types. <rdar://problem/10261670> llvm-svn: 143792
2011-11-05 14:08:06 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding values of two 128-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [4 x float], using the operation specified by the immediate integer
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// operand.
///
/// Returns a [4 x float] vector consisting of four floats corresponding to
/// the four comparison results: zero if the comparison is false, and all 1's
/// if the comparison is true.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_cmp_ps(__m128 a, __m128 b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 128-bit vector of [4 x float].
/// \param b
/// A 128-bit vector of [4 x float].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_cmp_ps(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m128)__builtin_ia32_cmpps((__v4sf)(__m128)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v4sf)(__m128)(b), (c))
Fix vector macros to correctly check argument types. <rdar://problem/10261670> llvm-svn: 143792
2011-11-05 14:08:06 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding double-precision values of two
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// 256-bit vectors of [4 x double], using the operation specified by the
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// immediate integer operand.
///
/// Returns a [4 x double] vector consisting of four doubles corresponding to
/// the four comparison results: zero if the comparison is false, and all 1's
/// if the comparison is true.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_cmp_pd(__m256d a, __m256d b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 256-bit vector of [4 x double].
/// \param b
/// A 256-bit vector of [4 x double].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [4 x double] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_cmp_pd(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256d)__builtin_ia32_cmppd256((__v4df)(__m256d)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v4df)(__m256d)(b), (c))
Fix vector macros to correctly check argument types. <rdar://problem/10261670> llvm-svn: 143792
2011-11-05 14:08:06 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding values of two 256-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [8 x float], using the operation specified by the immediate integer
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// operand.
///
/// Returns a [8 x float] vector consisting of eight floats corresponding to
/// the eight comparison results: zero if the comparison is false, and all
/// 1's if the comparison is true.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_cmp_ps(__m256 a, __m256 b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 256-bit vector of [8 x float].
/// \param b
/// A 256-bit vector of [8 x float].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 256-bit vector of [8 x float] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_cmp_ps(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m256)__builtin_ia32_cmpps256((__v8sf)(__m256)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v8sf)(__m256)(b), (c))
Fix vector macros to correctly check argument types. <rdar://problem/10261670> llvm-svn: 143792
2011-11-05 14:08:06 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding scalar double-precision values of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// two 128-bit vectors of [2 x double], using the operation specified by the
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// immediate integer operand.
///
/// If the result is true, all 64 bits of the destination vector are set;
/// otherwise they are cleared.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_cmp_sd(__m128d a, __m128d b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPSD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 128-bit vector of [2 x double].
/// \param b
/// A 128-bit vector of [2 x double].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_cmp_sd(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m128d)__builtin_ia32_cmpsd((__v2df)(__m128d)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v2df)(__m128d)(b), (c))
Fix vector macros to correctly check argument types. <rdar://problem/10261670> llvm-svn: 143792
2011-11-05 14:08:06 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Compares each of the corresponding scalar values of two 128-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// vectors of [4 x float], using the operation specified by the immediate
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// integer operand.
///
/// If the result is true, all 32 bits of the destination vector are set;
/// otherwise they are cleared.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_cmp_ss(__m128 a, __m128 b, const int c);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCMPSS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param a
/// A 128-bit vector of [4 x float].
/// \param b
/// A 128-bit vector of [4 x float].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// operation to use: \n
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// 0x08: Equal (unordered, non-signaling) \n
/// 0x09: Not-greater-than-or-equal (unordered, signaling) \n
/// 0x0A: Not-greater-than (unordered, signaling) \n
/// 0x0B: False (ordered, non-signaling) \n
/// 0x0C: Not-equal (ordered, non-signaling) \n
/// 0x0D: Greater-than-or-equal (ordered, signaling) \n
/// 0x0E: Greater-than (ordered, signaling) \n
/// 0x0F: True (unordered, non-signaling) \n
/// 0x10: Equal (ordered, signaling) \n
/// 0x11: Less-than (ordered, non-signaling) \n
/// 0x12: Less-than-or-equal (ordered, non-signaling) \n
/// 0x13: Unordered (signaling) \n
/// 0x14: Not-equal (unordered, signaling) \n
/// 0x15: Not-less-than (unordered, non-signaling) \n
/// 0x16: Not-less-than-or-equal (unordered, non-signaling) \n
/// 0x17: Ordered (signaling) \n
/// 0x18: Equal (unordered, signaling) \n
/// 0x19: Not-greater-than-or-equal (unordered, non-signaling) \n
/// 0x1A: Not-greater-than (unordered, non-signaling) \n
/// 0x1B: False (ordered, signaling) \n
/// 0x1C: Not-equal (ordered, signaling) \n
/// 0x1D: Greater-than-or-equal (ordered, non-signaling) \n
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm_cmp_ss(a, b, c) \
[X86] Remove temporary variables from macros in x86 intrinsic headers. Prevents duplicate names appearing from multiple macro expansions. NFC llvm-svn: 252586
2015-11-10 13:08:05 +08:00
(__m128)__builtin_ia32_cmpss((__v4sf)(__m128)(a), \
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
(__v4sf)(__m128)(b), (c))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [8 x i32] vector and returns the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x i32].
/// \param __imm
/// An immediate integer operand with bits [2:0] determining which vector
/// element is extracted and returned.
/// \returns A 32-bit integer containing the extracted 32 bits of extended
/// packed data.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_extract_epi32(X, N) \
(int)__builtin_ia32_vec_ext_v8si((__v8si)(__m256i)(X), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [16 x i16] vector and returns the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit integer vector of [16 x i16].
/// \param __imm
/// An immediate integer operand with bits [3:0] determining which vector
/// element is extracted and returned.
[X86][AVX] Ensure zero-extension of _mm256_extract_epi8 and _mm256_extract_epi16 Ensure _mm256_extract_epi8 and _mm256_extract_epi16 zero extend their i8/i16 result to i32. This matches _mm_extract_epi8 and _mm_extract_epi16. Fix for PR27594 Differential Revision: http://reviews.llvm.org/D20468 llvm-svn: 270330
2016-05-22 05:14:35 +08:00
/// \returns A 32-bit integer containing the extracted 16 bits of zero extended
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// packed data.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_extract_epi16(X, N) \
(int)(unsigned short)__builtin_ia32_vec_ext_v16hi((__v16hi)(__m256i)(X), \
(int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [32 x i8] vector and returns the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit integer vector of [32 x i8].
/// \param __imm
/// An immediate integer operand with bits [4:0] determining which vector
/// element is extracted and returned.
[X86][AVX] Ensure zero-extension of _mm256_extract_epi8 and _mm256_extract_epi16 Ensure _mm256_extract_epi8 and _mm256_extract_epi16 zero extend their i8/i16 result to i32. This matches _mm_extract_epi8 and _mm_extract_epi16. Fix for PR27594 Differential Revision: http://reviews.llvm.org/D20468 llvm-svn: 270330
2016-05-22 05:14:35 +08:00
/// \returns A 32-bit integer containing the extracted 8 bits of zero extended
/// packed data.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_extract_epi8(X, N) \
(int)(unsigned char)__builtin_ia32_vec_ext_v32qi((__v32qi)(__m256i)(X), \
(int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#ifdef __x86_64__
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [4 x i64] vector and returns the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit integer vector of [4 x i64].
/// \param __imm
/// An immediate integer operand with bits [1:0] determining which vector
/// element is extracted and returned.
/// \returns A 64-bit integer containing the extracted 64 bits of extended
/// packed data.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_extract_epi64(X, N) \
(long long)__builtin_ia32_vec_ext_v4di((__v4di)(__m256i)(X), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#endif
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [8 x i32] vector and replaces the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand by a new value. Returns the
/// modified vector.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A vector of [8 x i32] to be used by the insert operation.
/// \param __b
/// An integer value. The replacement value for the insert operation.
/// \param __imm
/// An immediate integer specifying the index of the vector element to be
/// replaced.
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// \returns A copy of vector \a __a, after replacing its element indexed by
/// \a __imm with \a __b.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_insert_epi32(X, I, N) \
(__m256i)__builtin_ia32_vec_set_v8si((__v8si)(__m256i)(X), \
(int)(I), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [16 x i16] vector and replaces the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand with a new value. Returns the
/// modified vector.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A vector of [16 x i16] to be used by the insert operation.
/// \param __b
/// An i16 integer value. The replacement value for the insert operation.
/// \param __imm
/// An immediate integer specifying the index of the vector element to be
/// replaced.
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// \returns A copy of vector \a __a, after replacing its element indexed by
/// \a __imm with \a __b.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_insert_epi16(X, I, N) \
(__m256i)__builtin_ia32_vec_set_v16hi((__v16hi)(__m256i)(X), \
(int)(I), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [32 x i8] vector and replaces the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand with a new value. Returns the
/// modified vector.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A vector of [32 x i8] to be used by the insert operation.
/// \param __b
/// An i8 integer value. The replacement value for the insert operation.
/// \param __imm
/// An immediate integer specifying the index of the vector element to be
/// replaced.
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// \returns A copy of vector \a __a, after replacing its element indexed by
/// \a __imm with \a __b.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_insert_epi8(X, I, N) \
(__m256i)__builtin_ia32_vec_set_v32qi((__v32qi)(__m256i)(X), \
(int)(I), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#ifdef __x86_64__
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Takes a [4 x i64] vector and replaces the vector element value
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// indexed by the immediate constant operand with a new value. Returns the
/// modified vector.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128+COMPOSITE </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A vector of [4 x i64] to be used by the insert operation.
/// \param __b
/// A 64-bit integer value. The replacement value for the insert operation.
/// \param __imm
/// An immediate integer specifying the index of the vector element to be
/// replaced.
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// \returns A copy of vector \a __a, after replacing its element indexed by
/// \a __imm with \a __b.
[X86] Add builtins for vector element insert and extract for different 128 and 256 bit vector types. Use them to implement the extract and insert intrinsics. Previously we were just using extended vector operations in the header file. This unfortunately allowed non-constant indices to be used with the intrinsics. This is incompatible with gcc, icc, and MSVC. It also introduces a different performance characteristic because non-constant index gets lowered to a vector store and an element sized load. By adding the builtins we can check for the index to be a constant and ensure its in range of the vector element count. User code still has the option to use extended vector operations themselves if they need non-constant indexing. llvm-svn: 334057
2018-06-06 08:24:55 +08:00
#define _mm256_insert_epi64(X, I, N) \
(__m256i)__builtin_ia32_vec_set_v4di((__v4di)(__m256i)(X), \
(long long)(I), (int)(N))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
#endif
/* Conversion */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a vector of [4 x i32] into a vector of [4 x double].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTDQ2PD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 128-bit integer vector of [4 x i32].
/// \returns A 256-bit vector of [4 x double] containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtepi32_pd(__m128i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86][SSE] Replace lossless i32/f32 to f64 conversion intrinsics with generic IR Both the (V)CVTDQ2PD(Y) (i32 to f64) and (V)CVTPS2PD(Y) (f32 to f64) conversion instructions are lossless and can be safely represented as generic __builtin_convertvector calls instead of x86 intrinsics without affecting final codegen. This patch removes the clang builtins and their use in the sse2/avx headers - a future patch will deal with removing the llvm intrinsics, but that will require a bit more work. Differential Revision: http://reviews.llvm.org/D20528 llvm-svn: 270499
2016-05-24 06:13:02 +08:00
return (__m256d)__builtin_convertvector((__v4si)__a, __v4df);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a vector of [8 x i32] into a vector of [8 x float].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTDQ2PS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit integer vector.
/// \returns A 256-bit vector of [8 x float] containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtepi32_ps(__m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Use __builtin_convertvector to implement some of the packed integer to packed float conversion intrinsics. I believe this is safe assuming default default FP environment. The conversion might be inexact, but it can never overflow the FP type so this shouldn't be undefined behavior for the uitofp/sitofp instructions. We already do something similar for scalar conversions. Differential Revision: https://reviews.llvm.org/D46863 llvm-svn: 332882
2018-05-22 04:19:17 +08:00
return (__m256)__builtin_convertvector((__v8si)__a, __v8sf);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a 256-bit vector of [4 x double] into a 128-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
/// [4 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTPD2PS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \returns A 128-bit vector of [4 x float] containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtpd_ps(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m128)__builtin_ia32_cvtpd2ps256((__v4df) __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a vector of [8 x float] into a vector of [8 x i32].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTPS2DQ </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, only around 50% of the intrinsics in this file are documented here. The patches for the other half will be sent out later. Updated bmiintrin.h to fix an incorrect section name. Updated f16cintrin.h to fix incorect parameter names. The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 269718
2016-05-17 06:54:45 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit integer vector containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtps_epi32(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256i)__builtin_ia32_cvtps2dq256((__v8sf) __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a 128-bit vector of [4 x float] into a 256-bit vector of [4
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTPS2PD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \returns A 256-bit vector of [4 x double] containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtps_pd(__m128 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86][SSE] Replace lossless i32/f32 to f64 conversion intrinsics with generic IR Both the (V)CVTDQ2PD(Y) (i32 to f64) and (V)CVTPS2PD(Y) (f32 to f64) conversion instructions are lossless and can be safely represented as generic __builtin_convertvector calls instead of x86 intrinsics without affecting final codegen. This patch removes the clang builtins and their use in the sse2/avx headers - a future patch will deal with removing the llvm intrinsics, but that will require a bit more work. Differential Revision: http://reviews.llvm.org/D20528 llvm-svn: 270499
2016-05-24 06:13:02 +08:00
return (__m256d)__builtin_convertvector((__v4sf)__a, __v4df);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a 256-bit vector of [4 x double] into a 128-bit vector of [4
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// x i32], truncating the result by rounding towards zero when it is
/// inexact.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTTPD2DQ </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \returns A 128-bit integer vector containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvttpd_epi32(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86][SSE] Reimplement SSE fp2si conversion intrinsics instead of using generic IR D20859 and D20860 attempted to replace the SSE (V)CVTTPS2DQ and VCVTTPD2DQ truncating conversions with generic IR instead. It turns out that the behaviour of these intrinsics is different enough from generic IR that this will cause problems, INF/NAN/out of range values are guaranteed to result in a 0x80000000 value - which plays havoc with constant folding which converts them to either zero or UNDEF. This is also an issue with the scalar implementations (which were already generic IR and what I was trying to match). This patch changes both scalar and packed versions back to using x86-specific builtins. It also deals with the other scalar conversion cases that are runtime rounding mode dependent and can have similar issues with constant folding. Differential Revision: https://reviews.llvm.org/D22105 llvm-svn: 276102
2016-07-20 18:18:01 +08:00
return (__m128i)__builtin_ia32_cvttpd2dq256((__v4df) __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a 256-bit vector of [4 x double] into a 128-bit vector of [4
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// x i32]. When a conversion is inexact, the value returned is rounded
/// according to the rounding control bits in the MXCSR register.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTPD2DQ </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \returns A 128-bit integer vector containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvtpd_epi32(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m128i)__builtin_ia32_cvtpd2dq256((__v4df) __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Converts a vector of [8 x float] into a vector of [8 x i32],
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// truncating the result by rounding towards zero when it is inexact.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VCVTTPS2DQ </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 256-bit integer vector containing the converted values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_cvttps_epi32(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86][SSE] Reimplement SSE fp2si conversion intrinsics instead of using generic IR D20859 and D20860 attempted to replace the SSE (V)CVTTPS2DQ and VCVTTPD2DQ truncating conversions with generic IR instead. It turns out that the behaviour of these intrinsics is different enough from generic IR that this will cause problems, INF/NAN/out of range values are guaranteed to result in a 0x80000000 value - which plays havoc with constant folding which converts them to either zero or UNDEF. This is also an issue with the scalar implementations (which were already generic IR and what I was trying to match). This patch changes both scalar and packed versions back to using x86-specific builtins. It also deals with the other scalar conversion cases that are runtime rounding mode dependent and can have similar issues with constant folding. Differential Revision: https://reviews.llvm.org/D22105 llvm-svn: 276102
2016-07-20 18:18:01 +08:00
return (__m256i)__builtin_ia32_cvttps2dq256((__v8sf) __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Returns the first element of the input vector of [4 x double].
[DOXYGEN] Documentation for the newly added x86 intrinsics. Added doxygen comments for the newly added intrinsics in avxintrin.h, namely _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32 Added doxygen comments for the new intrinsics in emmintrin.h, namely _mm_loadu_si64 and _mm_load_sd. Explicit parameter names were added for _mm_clflush and _mm_setcsr The rest of the changes are editorial, removing trailing spaces at the end of the lines. Differential Revision: https://reviews.llvm.org/D28503 llvm-svn: 291876
2017-01-13 09:14:08 +08:00
///
/// \headerfile <avxintrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
/// instruction.
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \returns A 64 bit double containing the first element of the input vector.
Adding front-end support to several intrinsics (bit scanning, conversion and state reading intrinsics) Adding LLVM front-end support to two intrinsics dealing with bit scan: _bit_scan_forward and _bit_scan_reverse. Their functionality is as described in Intel intrinsics guide: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_forward&expand=371,370 https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_reverse&expand=371,370 Furthermore, adding clang front-end support to these conversion intrinsics: _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32. Finally, adding tests to all of the above, as well as to the state reading intrinsics _rdpmc and _rdtsc. Their functionality is also specified in the Intel intrinsics guide. Commit on behalf of Omer Paparo Bivas llvm-svn: 271387
2016-06-01 20:21:00 +08:00
static __inline double __DEFAULT_FN_ATTRS
_mm256_cvtsd_f64(__m256d __a)
{
return __a[0];
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Returns the first element of the input vector of [8 x i32].
[DOXYGEN] Documentation for the newly added x86 intrinsics. Added doxygen comments for the newly added intrinsics in avxintrin.h, namely _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32 Added doxygen comments for the new intrinsics in emmintrin.h, namely _mm_loadu_si64 and _mm_load_sd. Explicit parameter names were added for _mm_clflush and _mm_setcsr The rest of the changes are editorial, removing trailing spaces at the end of the lines. Differential Revision: https://reviews.llvm.org/D28503 llvm-svn: 291876
2017-01-13 09:14:08 +08:00
///
/// \headerfile <avxintrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
/// instruction.
///
/// \param __a
/// A 256-bit vector of [8 x i32].
/// \returns A 32 bit integer containing the first element of the input vector.
Adding front-end support to several intrinsics (bit scanning, conversion and state reading intrinsics) Adding LLVM front-end support to two intrinsics dealing with bit scan: _bit_scan_forward and _bit_scan_reverse. Their functionality is as described in Intel intrinsics guide: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_forward&expand=371,370 https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_reverse&expand=371,370 Furthermore, adding clang front-end support to these conversion intrinsics: _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32. Finally, adding tests to all of the above, as well as to the state reading intrinsics _rdpmc and _rdtsc. Their functionality is also specified in the Intel intrinsics guide. Commit on behalf of Omer Paparo Bivas llvm-svn: 271387
2016-06-01 20:21:00 +08:00
static __inline int __DEFAULT_FN_ATTRS
_mm256_cvtsi256_si32(__m256i __a)
{
__v8si __b = (__v8si)__a;
return __b[0];
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Returns the first element of the input vector of [8 x float].
[DOXYGEN] Documentation for the newly added x86 intrinsics. Added doxygen comments for the newly added intrinsics in avxintrin.h, namely _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32 Added doxygen comments for the new intrinsics in emmintrin.h, namely _mm_loadu_si64 and _mm_load_sd. Explicit parameter names were added for _mm_clflush and _mm_setcsr The rest of the changes are editorial, removing trailing spaces at the end of the lines. Differential Revision: https://reviews.llvm.org/D28503 llvm-svn: 291876
2017-01-13 09:14:08 +08:00
///
/// \headerfile <avxintrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
/// instruction.
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \returns A 32 bit float containing the first element of the input vector.
Adding front-end support to several intrinsics (bit scanning, conversion and state reading intrinsics) Adding LLVM front-end support to two intrinsics dealing with bit scan: _bit_scan_forward and _bit_scan_reverse. Their functionality is as described in Intel intrinsics guide: https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_forward&expand=371,370 https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_bit_scan_reverse&expand=371,370 Furthermore, adding clang front-end support to these conversion intrinsics: _mm256_cvtsd_f64, _mm256_cvtsi256_si32 and _mm256_cvtss_f32. Finally, adding tests to all of the above, as well as to the state reading intrinsics _rdpmc and _rdtsc. Their functionality is also specified in the Intel intrinsics guide. Commit on behalf of Omer Paparo Bivas llvm-svn: 271387
2016-06-01 20:21:00 +08:00
static __inline float __DEFAULT_FN_ATTRS
_mm256_cvtss_f32(__m256 __a)
{
return __a[0];
}
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
/* Vector replicate */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves and duplicates odd-indexed values from a 256-bit vector of
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// [8 x float] to float values in a 256-bit vector of [8 x float].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVSHDUP </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [255:224] of \a __a are written to bits [255:224] and [223:192] of
/// the return value. \n
/// Bits [191:160] of \a __a are written to bits [191:160] and [159:128] of
/// the return value. \n
/// Bits [127:96] of \a __a are written to bits [127:96] and [95:64] of the
/// return value. \n
/// Bits [63:32] of \a __a are written to bits [63:32] and [31:0] of the
/// return value.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \returns A 256-bit vector of [8 x float] containing the moved and duplicated
/// values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_movehdup_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v8sf)__a, (__v8sf)__a, 1, 1, 3, 3, 5, 5, 7, 7);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves and duplicates even-indexed values from a 256-bit vector of
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// [8 x float] to float values in a 256-bit vector of [8 x float].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVSLDUP </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [223:192] of \a __a are written to bits [255:224] and [223:192] of
/// the return value. \n
/// Bits [159:128] of \a __a are written to bits [191:160] and [159:128] of
/// the return value. \n
/// Bits [95:64] of \a __a are written to bits [127:96] and [95:64] of the
/// return value. \n
/// Bits [31:0] of \a __a are written to bits [63:32] and [31:0] of the
/// return value.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \returns A 256-bit vector of [8 x float] containing the moved and duplicated
/// values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_moveldup_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v8sf)__a, (__v8sf)__a, 0, 0, 2, 2, 4, 4, 6, 6);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves and duplicates double-precision floating point values from a
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// 256-bit vector of [4 x double] to double-precision values in a 256-bit
/// vector of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDDUP </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [4 x double]. \n
/// Bits [63:0] of \a __a are written to bits [127:64] and [63:0] of the
/// return value. \n
/// Bits [191:128] of \a __a are written to bits [255:192] and [191:128] of
/// the return value.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \returns A 256-bit vector of [4 x double] containing the moved and
/// duplicated values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_movedup_pd(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4df)__a, (__v4df)__a, 0, 0, 2, 2);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Unpack and Interleave */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Unpacks the odd-indexed vector elements from two 256-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// [4 x double] and interleaves them into a 256-bit vector of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VUNPCKHPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit floating-point vector of [4 x double]. \n
/// Bits [127:64] are written to bits [63:0] of the return value. \n
/// Bits [255:192] are written to bits [191:128] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \param __b
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit floating-point vector of [4 x double]. \n
/// Bits [127:64] are written to bits [127:64] of the return value. \n
/// Bits [255:192] are written to bits [255:192] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \returns A 256-bit vector of [4 x double] containing the interleaved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_unpackhi_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4df)__a, (__v4df)__b, 1, 5, 1+2, 5+2);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Unpacks the even-indexed vector elements from two 256-bit vectors of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// [4 x double] and interleaves them into a 256-bit vector of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VUNPCKLPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit floating-point vector of [4 x double]. \n
/// Bits [63:0] are written to bits [63:0] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// Bits [191:128] are written to bits [191:128] of the return value.
/// \param __b
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit floating-point vector of [4 x double]. \n
/// Bits [63:0] are written to bits [127:64] of the return value. \n
/// Bits [191:128] are written to bits [255:192] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \returns A 256-bit vector of [4 x double] containing the interleaved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_unpacklo_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4df)__a, (__v4df)__b, 0, 4, 0+2, 4+2);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Unpacks the 32-bit vector elements 2, 3, 6 and 7 from each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// two 256-bit vectors of [8 x float] and interleaves them into a 256-bit
/// vector of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VUNPCKHPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [95:64] are written to bits [31:0] of the return value. \n
/// Bits [127:96] are written to bits [95:64] of the return value. \n
/// Bits [223:192] are written to bits [159:128] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// Bits [255:224] are written to bits [223:192] of the return value.
/// \param __b
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [95:64] are written to bits [63:32] of the return value. \n
/// Bits [127:96] are written to bits [127:96] of the return value. \n
/// Bits [223:192] are written to bits [191:160] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// Bits [255:224] are written to bits [255:224] of the return value.
/// \returns A 256-bit vector of [8 x float] containing the interleaved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_unpackhi_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v8sf)__a, (__v8sf)__b, 2, 10, 2+1, 10+1, 6, 14, 6+1, 14+1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Unpacks the 32-bit vector elements 0, 1, 4 and 5 from each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// two 256-bit vectors of [8 x float] and interleaves them into a 256-bit
/// vector of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VUNPCKLPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [31:0] are written to bits [31:0] of the return value. \n
/// Bits [63:32] are written to bits [95:64] of the return value. \n
/// Bits [159:128] are written to bits [159:128] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// Bits [191:160] are written to bits [223:192] of the return value.
/// \param __b
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// A 256-bit vector of [8 x float]. \n
/// Bits [31:0] are written to bits [63:32] of the return value. \n
/// Bits [63:32] are written to bits [127:96] of the return value. \n
/// Bits [159:128] are written to bits [191:160] of the return value. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// Bits [191:160] are written to bits [255:224] of the return value.
/// \returns A 256-bit vector of [8 x float] containing the interleaved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_unpacklo_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v8sf)__a, (__v8sf)__b, 0, 8, 0+1, 8+1, 4, 12, 4+1, 12+1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Bit Test */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [2 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the ZF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \param __b
/// A 128-bit vector of [2 x double].
/// \returns the ZF flag in the EFLAGS register.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testz_pd(__m128d __a, __m128d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestzpd((__v2df)__a, (__v2df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [2 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the CF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \param __b
/// A 128-bit vector of [2 x double].
/// \returns the CF flag in the EFLAGS register.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testc_pd(__m128d __a, __m128d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestcpd((__v2df)__a, (__v2df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [2 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns 1 if both the ZF and CF flags are set to 0,
/// otherwise it returns 0.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \param __b
/// A 128-bit vector of [2 x double].
/// \returns 1 if both the ZF and CF flags are set to 0, otherwise returns 0.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testnzc_pd(__m128d __a, __m128d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestnzcpd((__v2df)__a, (__v2df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [4 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the ZF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \param __b
/// A 128-bit vector of [4 x float].
/// \returns the ZF flag.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testz_ps(__m128 __a, __m128 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestzps((__v4sf)__a, (__v4sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [4 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the CF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \param __b
/// A 128-bit vector of [4 x float].
/// \returns the CF flag.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testc_ps(__m128 __a, __m128 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestcps((__v4sf)__a, (__v4sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 128-bit floating-point vectors of [4 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns 1 if both the ZF and CF flags are set to 0,
/// otherwise it returns 0.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \param __b
/// A 128-bit vector of [4 x float].
/// \returns 1 if both the ZF and CF flags are set to 0, otherwise returns 0.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline int __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_testnzc_ps(__m128 __a, __m128 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestnzcps((__v4sf)__a, (__v4sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [4 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision elements in the
/// first source vector and the corresponding elements in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the ZF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double].
/// \returns the ZF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testz_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestzpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [4 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision elements in the
/// first source vector and the corresponding elements in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the CF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double].
/// \returns the CF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testc_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestcpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [4 x double], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the double-precision elements in the
/// first source vector and the corresponding elements in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of double-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns 1 if both the ZF and CF flags are set to 0,
/// otherwise it returns 0.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double].
/// \param __b
/// A 256-bit vector of [4 x double].
/// \returns 1 if both the ZF and CF flags are set to 0, otherwise returns 0.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testnzc_pd(__m256d __a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestnzcpd256((__v4df)__a, (__v4df)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [8 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the ZF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float].
/// \returns the ZF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testz_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestzps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [8 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision element in the
/// first source vector and the corresponding element in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the CF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float].
/// \returns the CF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testc_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestcps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit floating-point vectors of [8 x float], perform an
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// element-by-element comparison of the single-precision elements in the
/// first source vector and the corresponding elements in the second source
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// vector.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bits of both elements are 1, the ZF flag is set to 0. Otherwise the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of single-precision elements where the
/// sign-bit of the first element is 0 and the sign-bit of the second element
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is 1, the CF flag is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns 1 if both the ZF and CF flags are set to 0,
/// otherwise it returns 0.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VTESTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [8 x float].
/// \param __b
/// A 256-bit vector of [8 x float].
/// \returns 1 if both the ZF and CF flags are set to 0, otherwise returns 0.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testnzc_ps(__m256 __a, __m256 __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_vtestnzcps256((__v8sf)__a, (__v8sf)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit integer vectors, perform a bit-by-bit comparison
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// of the two source vectors.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where both bits are 1, the ZF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where the bit from the first source
/// vector is 0 and the bit from the second source vector is 1, the CF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the ZF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPTEST </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit integer vector.
/// \param __b
/// A 256-bit integer vector.
/// \returns the ZF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testz_si256(__m256i __a, __m256i __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_ptestz256((__v4di)__a, (__v4di)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit integer vectors, perform a bit-by-bit comparison
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// of the two source vectors.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where both bits are 1, the ZF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where the bit from the first source
/// vector is 0 and the bit from the second source vector is 1, the CF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns the value of the CF flag.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPTEST </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit integer vector.
/// \param __b
/// A 256-bit integer vector.
/// \returns the CF flag.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testc_si256(__m256i __a, __m256i __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_ptestc256((__v4di)__a, (__v4di)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Given two 256-bit integer vectors, perform a bit-by-bit comparison
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// of the two source vectors.
///
/// The EFLAGS register is updated as follows: \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where both bits are 1, the ZF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the ZF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// If there is at least one pair of bits where the bit from the first source
/// vector is 0 and the bit from the second source vector is 1, the CF flag
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// is set to 0. Otherwise the CF flag is set to 1. \n
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// This intrinsic returns 1 if both the ZF and CF flags are set to 0,
/// otherwise it returns 0.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPTEST </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit integer vector.
/// \param __b
/// A 256-bit integer vector.
/// \returns 1 if both the ZF and CF flags are set to 0, otherwise returns 0.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_testnzc_si256(__m256i __a, __m256i __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_ptestnzc256((__v4di)__a, (__v4di)__b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Vector extract sign mask */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Extracts the sign bits of double-precision floating point elements
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// in a 256-bit vector of [4 x double] and writes them to the lower order
/// bits of the return value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVMSKPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A 256-bit vector of [4 x double] containing the double-precision
/// floating point values with sign bits to be extracted.
/// \returns The sign bits from the operand, written to bits [3:0].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_movemask_pd(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_movmskpd256((__v4df)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Extracts the sign bits of single-precision floating point elements
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// in a 256-bit vector of [8 x float] and writes them to the lower order
/// bits of the return value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVMSKPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
[DOXYGEN] Fix doxygen and content issues in avxintrin.h - Fix incorrect wording in various intrinsic descriptions. Previously the descriptions used "low-order" and "high-order" when the intended meaning was "even-indexed" and "odd-indexed". - Fix a few typos and errors found during review. - Restore new line endings. This patch was made by Craig Flores llvm-svn: 322027
2018-01-09 05:21:17 +08:00
/// A 256-bit vector of [8 x float] containing the single-precision floating
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// point values with sign bits to be extracted.
/// \returns The sign bits from the operand, written to bits [7:0].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline int __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_movemask_ps(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return __builtin_ia32_movmskps256((__v8sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
/* Vector __zero */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Zeroes the contents of all XMM or YMM registers.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VZEROALL </c> instruction.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline void __attribute__((__always_inline__, __nodebug__, __target__("avx")))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
_mm256_zeroall(void)
{
__builtin_ia32_vzeroall();
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Zeroes the upper 128 bits (bits 255:128) of all YMM registers.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VZEROUPPER </c> instruction.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline void __attribute__((__always_inline__, __nodebug__, __target__("avx")))
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
_mm256_zeroupper(void)
{
__builtin_ia32_vzeroupper();
}
/* Vector load with broadcast */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads a scalar single-precision floating point value from the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// specified address pointed to by \a __a and broadcasts it to the elements
/// of a [4 x float] vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBROADCASTSS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// The single-precision floating point value to be broadcast.
/// \returns A 128-bit vector of [4 x float] whose 32-bit elements are set
/// equal to the broadcast value.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline __m128 __DEFAULT_FN_ATTRS128
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm_broadcast_ss(float const *__a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Implement AVX1 vbroadcast intrinsics with vector initializers These intrinsics are special because they directly take a memory operand (AVX2 adds the register counterparts). Typically, other non-memop intrinsics take registers and then it's left to isel to fold memory operands. In order to LICM intrinsics directly reading memory, we require that no stores are in the loop (LICM) or that the folded load accesses constant memory (MachineLICM). When neither is the case we fail to hoist a loop-invariant broadcast. We can work around this limitation if we expose the load as a regular load and then just implement the broadcast using the vector initializer syntax. This exposes the load to LICM and other optimizations. At the IR level this is translated into a series of insertelements. The sequence is already recognized as a broadcast so there is no impact on the quality of codegen. _mm256_broadcast_pd and _mm256_broadcast_ps are not updated by this patch because right now we lack the DAG-combiner smartness to recover the broadcast instructions. This will be tackled in a follow-on. There will be completing changes on the LLVM side to remove the LLVM intrinsics and to auto-upgrade bitcode files. Fixes <rdar://problem/16494520> llvm-svn: 209846
2014-05-30 04:47:29 +08:00
float __f = *__a;
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m128)(__v4sf){ __f, __f, __f, __f };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads a scalar double-precision floating point value from the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// specified address pointed to by \a __a and broadcasts it to the elements
/// of a [4 x double] vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBROADCASTSD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// The double-precision floating point value to be broadcast.
/// \returns A 256-bit vector of [4 x double] whose 64-bit elements are set
/// equal to the broadcast value.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_broadcast_sd(double const *__a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Implement AVX1 vbroadcast intrinsics with vector initializers These intrinsics are special because they directly take a memory operand (AVX2 adds the register counterparts). Typically, other non-memop intrinsics take registers and then it's left to isel to fold memory operands. In order to LICM intrinsics directly reading memory, we require that no stores are in the loop (LICM) or that the folded load accesses constant memory (MachineLICM). When neither is the case we fail to hoist a loop-invariant broadcast. We can work around this limitation if we expose the load as a regular load and then just implement the broadcast using the vector initializer syntax. This exposes the load to LICM and other optimizations. At the IR level this is translated into a series of insertelements. The sequence is already recognized as a broadcast so there is no impact on the quality of codegen. _mm256_broadcast_pd and _mm256_broadcast_ps are not updated by this patch because right now we lack the DAG-combiner smartness to recover the broadcast instructions. This will be tackled in a follow-on. There will be completing changes on the LLVM side to remove the LLVM intrinsics and to auto-upgrade bitcode files. Fixes <rdar://problem/16494520> llvm-svn: 209846
2014-05-30 04:47:29 +08:00
double __d = *__a;
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256d)(__v4df){ __d, __d, __d, __d };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads a scalar single-precision floating point value from the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// specified address pointed to by \a __a and broadcasts it to the elements
/// of a [8 x float] vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBROADCASTSS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// The single-precision floating point value to be broadcast.
/// \returns A 256-bit vector of [8 x float] whose 32-bit elements are set
/// equal to the broadcast value.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_broadcast_ss(float const *__a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Implement AVX1 vbroadcast intrinsics with vector initializers These intrinsics are special because they directly take a memory operand (AVX2 adds the register counterparts). Typically, other non-memop intrinsics take registers and then it's left to isel to fold memory operands. In order to LICM intrinsics directly reading memory, we require that no stores are in the loop (LICM) or that the folded load accesses constant memory (MachineLICM). When neither is the case we fail to hoist a loop-invariant broadcast. We can work around this limitation if we expose the load as a regular load and then just implement the broadcast using the vector initializer syntax. This exposes the load to LICM and other optimizations. At the IR level this is translated into a series of insertelements. The sequence is already recognized as a broadcast so there is no impact on the quality of codegen. _mm256_broadcast_pd and _mm256_broadcast_ps are not updated by this patch because right now we lack the DAG-combiner smartness to recover the broadcast instructions. This will be tackled in a follow-on. There will be completing changes on the LLVM side to remove the LLVM intrinsics and to auto-upgrade bitcode files. Fixes <rdar://problem/16494520> llvm-svn: 209846
2014-05-30 04:47:29 +08:00
float __f = *__a;
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256)(__v8sf){ __f, __f, __f, __f, __f, __f, __f, __f };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads the data from a 128-bit vector of [2 x double] from the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// specified address pointed to by \a __a and broadcasts it to 128-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// elements in a 256-bit vector of [4 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBROADCASTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// The 128-bit vector of [2 x double] to be broadcast.
/// \returns A 256-bit vector of [4 x double] whose 128-bit elements are set
/// equal to the broadcast value.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_broadcast_pd(__m128d const *__a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Replace __builtin_ia32_vbroadcastf128_pd256 and __builtin_ia32_vbroadcastf128_ps256 with an unaligned load intrinsics and a __builtin_shufflevector call. llvm-svn: 333853
2018-06-04 03:42:59 +08:00
__m128d __b = _mm_loadu_pd((const double *)__a);
return (__m256d)__builtin_shufflevector((__v2df)__b, (__v2df)__b,
0, 1, 0, 1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads the data from a 128-bit vector of [4 x float] from the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// specified address pointed to by \a __a and broadcasts it to 128-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// elements in a 256-bit vector of [8 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VBROADCASTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// The 128-bit vector of [4 x float] to be broadcast.
/// \returns A 256-bit vector of [8 x float] whose 128-bit elements are set
/// equal to the broadcast value.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_broadcast_ps(__m128 const *__a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Replace __builtin_ia32_vbroadcastf128_pd256 and __builtin_ia32_vbroadcastf128_ps256 with an unaligned load intrinsics and a __builtin_shufflevector call. llvm-svn: 333853
2018-06-04 03:42:59 +08:00
__m128 __b = _mm_loadu_ps((const float *)__a);
return (__m256)__builtin_shufflevector((__v4sf)__b, (__v4sf)__b,
0, 1, 2, 3, 0, 1, 2, 3);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* SIMD load ops */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 4 double-precision floating point values from a 32-byte aligned
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p into a vector of [4 x double].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVAPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a memory location containing
/// double-precision floating point values.
/// \returns A 256-bit vector of [4 x double] containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_load_pd(double const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return *(__m256d *)__p;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 8 single-precision floating point values from a 32-byte aligned
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p into a vector of [8 x float].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVAPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a memory location containing float values.
/// \returns A 256-bit vector of [8 x float] containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_load_ps(float const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return *(__m256 *)__p;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 4 double-precision floating point values from an unaligned
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p into a vector of [4 x double].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVUPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location containing double-precision floating
/// point values.
/// \returns A 256-bit vector of [4 x double] containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu_pd(double const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Represent 256-bit unaligned loads natively and remove the builtins. Similar change was made for 128-bit versions a while back. llvm-svn: 148919
2012-01-25 12:26:17 +08:00
struct __loadu_pd {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256d_u __v;
Headers: Don't use attribute keywords which aren't reserved Instead of using 'unavailable', use '__unavailable__' llvm-svn: 228087
2015-02-04 08:26:10 +08:00
} __attribute__((__packed__, __may_alias__));
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return ((struct __loadu_pd*)__p)->__v;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 8 single-precision floating point values from an unaligned
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p into a vector of [8 x float].
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVUPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location containing single-precision floating
/// point values.
/// \returns A 256-bit vector of [8 x float] containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu_ps(float const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Represent 256-bit unaligned loads natively and remove the builtins. Similar change was made for 128-bit versions a while back. llvm-svn: 148919
2012-01-25 12:26:17 +08:00
struct __loadu_ps {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256_u __v;
Headers: Don't use attribute keywords which aren't reserved Instead of using 'unavailable', use '__unavailable__' llvm-svn: 228087
2015-02-04 08:26:10 +08:00
} __attribute__((__packed__, __may_alias__));
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return ((struct __loadu_ps*)__p)->__v;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 256 bits of integer data from a 32-byte aligned memory
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// location pointed to by \a __p into elements of a 256-bit integer vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDQA </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a 256-bit integer vector containing integer
/// values.
/// \returns A 256-bit integer vector containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_load_si256(__m256i const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return *__p;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 256 bits of integer data from an unaligned memory location
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// pointed to by \a __p into a 256-bit integer vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDQU </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a 256-bit integer vector containing integer values.
/// \returns A 256-bit integer vector containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu_si256(__m256i const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Represent 256-bit unaligned loads natively and remove the builtins. Similar change was made for 128-bit versions a while back. llvm-svn: 148919
2012-01-25 12:26:17 +08:00
struct __loadu_si256 {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256i_u __v;
Headers: Don't use attribute keywords which aren't reserved Instead of using 'unavailable', use '__unavailable__' llvm-svn: 228087
2015-02-04 08:26:10 +08:00
} __attribute__((__packed__, __may_alias__));
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return ((struct __loadu_si256*)__p)->__v;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads 256 bits of integer data from an unaligned memory location
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// pointed to by \a __p into a 256-bit integer vector. This intrinsic may
/// perform better than \c _mm256_loadu_si256 when the data crosses a cache
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// line boundary.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VLDDQU </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a 256-bit integer vector containing integer values.
/// \returns A 256-bit integer vector containing the moved values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_lddqu_si256(__m256i const *__p)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256i)__builtin_ia32_lddqu256((char const *)__p);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* SIMD store ops */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores double-precision floating point values from a 256-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [4 x double] to a 32-byte aligned memory location pointed to by
/// \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVAPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a memory location that will receive the
/// double-precision floaing point values.
/// \param __a
/// A 256-bit vector of [4 x double] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_store_pd(double *__p, __m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
*(__m256d *)__p = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores single-precision floating point values from a 256-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [8 x float] to a 32-byte aligned memory location pointed to by \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVAPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a memory location that will receive the
/// float values.
/// \param __a
/// A 256-bit vector of [8 x float] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_store_ps(float *__p, __m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
*(__m256 *)__p = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores double-precision floating point values from a 256-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [4 x double] to an unaligned memory location pointed to by \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVUPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the double-precision
/// floating point values.
/// \param __a
/// A 256-bit vector of [4 x double] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu_pd(double *__p, __m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
struct __storeu_pd {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256d_u __v;
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
} __attribute__((__packed__, __may_alias__));
((struct __storeu_pd*)__p)->__v = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores single-precision floating point values from a 256-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [8 x float] to an unaligned memory location pointed to by \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVUPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the float values.
/// \param __a
/// A 256-bit vector of [8 x float] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu_ps(float *__p, __m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
struct __storeu_ps {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256_u __v;
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
} __attribute__((__packed__, __may_alias__));
((struct __storeu_ps*)__p)->__v = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores integer values from a 256-bit integer vector to a 32-byte
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// aligned memory location pointed to by \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDQA </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A 32-byte aligned pointer to a memory location that will receive the
/// integer values.
/// \param __a
/// A 256-bit integer vector containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_store_si256(__m256i *__p, __m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
*__p = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores integer values from a 256-bit integer vector to an unaligned
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDQU </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the integer values.
/// \param __a
/// A 256-bit integer vector containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu_si256(__m256i *__p, __m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
struct __storeu_si256 {
[X86] Add explicit alignment to __m128/__m128i/__m128d/etc. to allow matching of MSVC behavior with #pragma pack. Summary: With MSVC, #pragma pack is ignored when there is explicit alignment. This differs from gcc. Clang emulates this difference when compiling for Windows. It appears that MSVC and its headers consider the __m128/__m128i/__m128d/etc. types to be explicitly aligned and ignores #pragma pack for them. Since we don't have explicit alignment on them in our headers, we don't match the MSVC behavior here. This patch adds explicit alignment to match this behavior. I'm hoping this won't cause any problems when we're not emulating MSVC. But if someone knows of something that would be different we can swith to conditionally adding the alignment based on _MSC_VER. I had to add explicitly unaligned types as well so we could use them in the loadu/storeu intrinsics which use __attribute__(__packed__). Using the now explicitly aligned types wouldn't produce align 1 accesses when targeting Windows. Reviewers: rnk, erichkeane, spatel, RKSimon Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57961 llvm-svn: 353555
2019-02-09 03:45:08 +08:00
__m256i_u __v;
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
} __attribute__((__packed__, __may_alias__));
((struct __storeu_si256*)__p)->__v = __a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Conditional load ops */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Conditionally loads double-precision floating point elements from a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// memory location pointed to by \a __p into a 128-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// [2 x double], depending on the mask bits associated with each data
/// element.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that contains the double-precision
/// floating point values.
/// \param __m
/// A 128-bit integer vector containing the mask. The most significant bit of
/// each data element represents the mask bits. If a mask bit is zero, the
/// corresponding value in the memory location is not loaded and the
/// corresponding field in the return value is set to zero.
/// \returns A 128-bit vector of [2 x double] containing the loaded values.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline __m128d __DEFAULT_FN_ATTRS128
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm_maskload_pd(double const *__p, __m128i __m)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
return (__m128d)__builtin_ia32_maskloadpd((const __v2df *)__p, (__v2di)__m);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Conditionally loads double-precision floating point elements from a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// memory location pointed to by \a __p into a 256-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// [4 x double], depending on the mask bits associated with each data
/// element.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that contains the double-precision
/// floating point values.
/// \param __m
/// A 256-bit integer vector of [4 x quadword] containing the mask. The most
/// significant bit of each quadword element represents the mask bits. If a
/// mask bit is zero, the corresponding value in the memory location is not
/// loaded and the corresponding field in the return value is set to zero.
/// \returns A 256-bit vector of [4 x double] containing the loaded values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm256_maskload_pd(double const *__p, __m256i __m)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
return (__m256d)__builtin_ia32_maskloadpd256((const __v4df *)__p,
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
(__v4di)__m);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Conditionally loads single-precision floating point elements from a
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// memory location pointed to by \a __p into a 128-bit vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// [4 x float], depending on the mask bits associated with each data
/// element.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that contains the single-precision
/// floating point values.
/// \param __m
/// A 128-bit integer vector containing the mask. The most significant bit of
/// each data element represents the mask bits. If a mask bit is zero, the
/// corresponding value in the memory location is not loaded and the
/// corresponding field in the return value is set to zero.
/// \returns A 128-bit vector of [4 x float] containing the loaded values.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline __m128 __DEFAULT_FN_ATTRS128
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm_maskload_ps(float const *__p, __m128i __m)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
return (__m128)__builtin_ia32_maskloadps((const __v4sf *)__p, (__v4si)__m);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Conditionally loads single-precision floating point elements from a
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// memory location pointed to by \a __p into a 256-bit vector of
/// [8 x float], depending on the mask bits associated with each data
/// element.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that contains the single-precision
/// floating point values.
/// \param __m
/// A 256-bit integer vector of [8 x dword] containing the mask. The most
/// significant bit of each dword element represents the mask bits. If a mask
/// bit is zero, the corresponding value in the memory location is not loaded
/// and the corresponding field in the return value is set to zero.
/// \returns A 256-bit vector of [8 x float] containing the loaded values.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm256_maskload_ps(float const *__p, __m256i __m)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
return (__m256)__builtin_ia32_maskloadps256((const __v8sf *)__p, (__v8si)__m);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Conditional store ops */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves single-precision floating point values from a 256-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [8 x float] to a memory location pointed to by \a __p, according to
/// the specified mask.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the float values.
/// \param __m
/// A 256-bit integer vector of [8 x dword] containing the mask. The most
/// significant bit of each dword element in the mask vector represents the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// mask bits. If a mask bit is zero, the corresponding value from vector
/// \a __a is not stored and the corresponding field in the memory location
/// pointed to by \a __p is not changed.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \param __a
/// A 256-bit vector of [8 x float] containing the values to be stored.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm256_maskstore_ps(float *__p, __m256i __m, __m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
__builtin_ia32_maskstoreps256((__v8sf *)__p, (__v8si)__m, (__v8sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves double-precision values from a 128-bit vector of [2 x double]
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// to a memory location pointed to by \a __p, according to the specified
/// mask.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the float values.
/// \param __m
/// A 128-bit integer vector containing the mask. The most significant bit of
/// each field in the mask vector represents the mask bits. If a mask bit is
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// zero, the corresponding value from vector \a __a is not stored and the
/// corresponding field in the memory location pointed to by \a __p is not
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// changed.
/// \param __a
/// A 128-bit vector of [2 x double] containing the values to be stored.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline void __DEFAULT_FN_ATTRS128
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm_maskstore_pd(double *__p, __m128i __m, __m128d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
__builtin_ia32_maskstorepd((__v2df *)__p, (__v2di)__m, (__v2df)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves double-precision values from a 256-bit vector of [4 x double]
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// to a memory location pointed to by \a __p, according to the specified
/// mask.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the float values.
/// \param __m
/// A 256-bit integer vector of [4 x quadword] containing the mask. The most
/// significant bit of each quadword element in the mask vector represents
/// the mask bits. If a mask bit is zero, the corresponding value from vector
/// __a is not stored and the corresponding field in the memory location
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// pointed to by \a __p is not changed.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \param __a
/// A 256-bit vector of [4 x double] containing the values to be stored.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm256_maskstore_pd(double *__p, __m256i __m, __m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
__builtin_ia32_maskstorepd256((__v4df *)__p, (__v4di)__m, (__v4df)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves single-precision floating point values from a 128-bit vector
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// of [4 x float] to a memory location pointed to by \a __p, according to
/// the specified mask.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMASKMOVPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a memory location that will receive the float values.
/// \param __m
/// A 128-bit integer vector containing the mask. The most significant bit of
/// each field in the mask vector represents the mask bits. If a mask bit is
/// zero, the corresponding value from vector __a is not stored and the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// corresponding field in the memory location pointed to by \a __p is not
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// changed.
/// \param __a
/// A 128-bit vector of [4 x float] containing the values to be stored.
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
static __inline void __DEFAULT_FN_ATTRS128
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
_mm_maskstore_ps(float *__p, __m128i __m, __m128 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[x86] Fix maskload/store intrinsic definitions in avxintrin.h According to the Intel documentation, the mask operand of a maskload and maskstore intrinsics is always a vector of packed integer/long integer values. This patch introduces the following two changes: 1. It fixes the avx maskload/store intrinsic definitions in avxintrin.h. 2. It changes BuiltinsX86.def to match the correct gcc definitions for avx maskload/store (see D13861 for more details). Differential Revision: http://reviews.llvm.org/D13861 llvm-svn: 250816
2015-10-20 19:19:54 +08:00
__builtin_ia32_maskstoreps((__v4sf *)__p, (__v4si)__m, (__v4sf)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Cacheability support ops */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves integer data from a 256-bit integer vector to a 32-byte
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// aligned memory location. To minimize caching, the data is flagged as
/// non-temporal (unlikely to be used again soon).
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVNTDQ </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A pointer to a 32-byte aligned memory location that will receive the
/// integer values.
/// \param __b
/// A 256-bit integer vector containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_stream_si256(__m256i *__a, __m256i __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef __v4di __v4di_aligned __attribute__((aligned(32)));
__builtin_nontemporal_store((__v4di_aligned)__b, (__v4di_aligned*)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves double-precision values from a 256-bit vector of [4 x double]
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// to a 32-byte aligned memory location. To minimize caching, the data is
/// flagged as non-temporal (unlikely to be used again soon).
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVNTPD </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __a
/// A pointer to a 32-byte aligned memory location that will receive the
[DOXYGEN] Corrected several typos and incorrect parameters description that Sony's techinical writer found during review. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 304840
2017-06-07 06:58:01 +08:00
/// double-precision floating-point values.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// \param __b
/// A 256-bit vector of [4 x double] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_stream_pd(double *__a, __m256d __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef __v4df __v4df_aligned __attribute__((aligned(32)));
__builtin_nontemporal_store((__v4df_aligned)__b, (__v4df_aligned*)__a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Moves single-precision floating point values from a 256-bit vector
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
/// of [8 x float] to a 32-byte aligned memory location. To minimize
/// caching, the data is flagged as non-temporal (unlikely to be used again
/// soon).
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVNTPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \param __p
/// A pointer to a 32-byte aligned memory location that will receive the
/// single-precision floating point values.
/// \param __a
/// A 256-bit vector of [8 x float] containing the values to be moved.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_stream_ps(float *__p, __m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Revert r333791 "Cap "voluntary" vector alignment at 16 for all Darwin platforms." Adding __attribute__((aligned(32))) to __m256 breaks the implementation of _mm256_loadu_ps on Windows. On Windows, alignment attributes have higher precedence than packing attributes. We also might want to carefully consider the consequences of changing our vector typedefs, since many users copy them and invent their own new, non-Intel specific vector type names. llvm-svn: 333958
2018-06-05 05:39:20 +08:00
typedef __v8sf __v8sf_aligned __attribute__((aligned(32)));
__builtin_nontemporal_store((__v8sf_aligned)__a, (__v8sf_aligned*)__p);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Create vectors */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Create a 256-bit vector of [4 x double] with undefined values.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \returns A 256-bit vector of [4 x double] containing undefined values.
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
static __inline__ __m256d __DEFAULT_FN_ATTRS
[X86] Add void to the argument list of intrinsics that don't take arguments since empty argument list mean something else in C. llvm-svn: 272244
2016-06-09 13:14:28 +08:00
_mm256_undefined_pd(void)
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
{
return (__m256d)__builtin_ia32_undef256();
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Create a 256-bit vector of [8 x float] with undefined values.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \returns A 256-bit vector of [8 x float] containing undefined values.
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
static __inline__ __m256 __DEFAULT_FN_ATTRS
[X86] Add void to the argument list of intrinsics that don't take arguments since empty argument list mean something else in C. llvm-svn: 272244
2016-06-09 13:14:28 +08:00
_mm256_undefined_ps(void)
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
{
return (__m256)__builtin_ia32_undef256();
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Create a 256-bit integer vector with undefined values.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, all the intrinsics in this file that were documented by Sony's intrinsics guide should have corresponding doxygen comments. Note: The doxygen comments are automatically generated based on Sony's intrinsic s document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. Reviewed by Wolfgang Pieb. llvm-svn: 287436
2016-11-19 12:59:08 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \returns A 256-bit integer vector containing undefined values.
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
static __inline__ __m256i __DEFAULT_FN_ATTRS
[X86] Add void to the argument list of intrinsics that don't take arguments since empty argument list mean something else in C. llvm-svn: 272244
2016-06-09 13:14:28 +08:00
_mm256_undefined_si256(void)
[X86][SSE] Add _mm_undefined_* intrinsics Added missing SSE/AVX 'undefined' intrinsics (PR24040): _mm_undefined_pd, _mm_undefined_ps + _mm_undefined_si128 _mm256_undefined_pd, _mm256_undefined_ps + _mm256_undefined_si256 _mm512_undefined, _mm512_undefined_ps, _mm512_undefined_pd + _mm512_undefined_epi32 Added builtin intrinsicss: __builtin_ia32_undef128, __builtin_ia32_undef256 + __builtin_ia32_undef512 Differential Revision: http://reviews.llvm.org/D12052 llvm-svn: 246083
2015-08-27 05:17:12 +08:00
{
return (__m256i)__builtin_ia32_undef256();
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double]
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// initialized with the specified double-precision floating-point values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VUNPCKLPD+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A double-precision floating-point value used to initialize bits [255:192]
/// of the result.
/// \param __b
/// A double-precision floating-point value used to initialize bits [191:128]
/// of the result.
/// \param __c
/// A double-precision floating-point value used to initialize bits [127:64]
/// of the result.
/// \param __d
/// A double-precision floating-point value used to initialize bits [63:0]
/// of the result.
/// \returns An initialized 256-bit floating-point vector of [4 x double].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_pd(double __a, double __b, double __c, double __d)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256d){ __d, __c, __b, __a };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] initialized
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// with the specified single-precision floating-point values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A single-precision floating-point value used to initialize bits [255:224]
/// of the result.
/// \param __b
/// A single-precision floating-point value used to initialize bits [223:192]
/// of the result.
/// \param __c
/// A single-precision floating-point value used to initialize bits [191:160]
/// of the result.
/// \param __d
/// A single-precision floating-point value used to initialize bits [159:128]
/// of the result.
/// \param __e
/// A single-precision floating-point value used to initialize bits [127:96]
/// of the result.
/// \param __f
/// A single-precision floating-point value used to initialize bits [95:64]
/// of the result.
/// \param __g
/// A single-precision floating-point value used to initialize bits [63:32]
/// of the result.
/// \param __h
/// A single-precision floating-point value used to initialize bits [31:0]
/// of the result.
/// \returns An initialized 256-bit floating-point vector of [8 x float].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_ps(float __a, float __b, float __c, float __d,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
float __e, float __f, float __g, float __h)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256){ __h, __g, __f, __e, __d, __c, __b, __a };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector initialized with the specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 32-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __i0
/// A 32-bit integral value used to initialize bits [255:224] of the result.
/// \param __i1
/// A 32-bit integral value used to initialize bits [223:192] of the result.
/// \param __i2
/// A 32-bit integral value used to initialize bits [191:160] of the result.
/// \param __i3
/// A 32-bit integral value used to initialize bits [159:128] of the result.
/// \param __i4
/// A 32-bit integral value used to initialize bits [127:96] of the result.
/// \param __i5
/// A 32-bit integral value used to initialize bits [95:64] of the result.
/// \param __i6
/// A 32-bit integral value used to initialize bits [63:32] of the result.
/// \param __i7
/// A 32-bit integral value used to initialize bits [31:0] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_epi32(int __i0, int __i1, int __i2, int __i3,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
int __i4, int __i5, int __i6, int __i7)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256i)(__v8si){ __i7, __i6, __i5, __i4, __i3, __i2, __i1, __i0 };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector initialized with the specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 16-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __w15
/// A 16-bit integral value used to initialize bits [255:240] of the result.
/// \param __w14
/// A 16-bit integral value used to initialize bits [239:224] of the result.
/// \param __w13
/// A 16-bit integral value used to initialize bits [223:208] of the result.
/// \param __w12
/// A 16-bit integral value used to initialize bits [207:192] of the result.
/// \param __w11
/// A 16-bit integral value used to initialize bits [191:176] of the result.
/// \param __w10
/// A 16-bit integral value used to initialize bits [175:160] of the result.
/// \param __w09
/// A 16-bit integral value used to initialize bits [159:144] of the result.
/// \param __w08
/// A 16-bit integral value used to initialize bits [143:128] of the result.
/// \param __w07
/// A 16-bit integral value used to initialize bits [127:112] of the result.
/// \param __w06
/// A 16-bit integral value used to initialize bits [111:96] of the result.
/// \param __w05
/// A 16-bit integral value used to initialize bits [95:80] of the result.
/// \param __w04
/// A 16-bit integral value used to initialize bits [79:64] of the result.
/// \param __w03
/// A 16-bit integral value used to initialize bits [63:48] of the result.
/// \param __w02
/// A 16-bit integral value used to initialize bits [47:32] of the result.
/// \param __w01
/// A 16-bit integral value used to initialize bits [31:16] of the result.
/// \param __w00
/// A 16-bit integral value used to initialize bits [15:0] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_epi16(short __w15, short __w14, short __w13, short __w12,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
short __w11, short __w10, short __w09, short __w08,
short __w07, short __w06, short __w05, short __w04,
short __w03, short __w02, short __w01, short __w00)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256i)(__v16hi){ __w00, __w01, __w02, __w03, __w04, __w05, __w06,
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__w07, __w08, __w09, __w10, __w11, __w12, __w13, __w14, __w15 };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector initialized with the specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 8-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __b31
/// An 8-bit integral value used to initialize bits [255:248] of the result.
/// \param __b30
/// An 8-bit integral value used to initialize bits [247:240] of the result.
/// \param __b29
/// An 8-bit integral value used to initialize bits [239:232] of the result.
/// \param __b28
/// An 8-bit integral value used to initialize bits [231:224] of the result.
/// \param __b27
/// An 8-bit integral value used to initialize bits [223:216] of the result.
/// \param __b26
/// An 8-bit integral value used to initialize bits [215:208] of the result.
/// \param __b25
/// An 8-bit integral value used to initialize bits [207:200] of the result.
/// \param __b24
/// An 8-bit integral value used to initialize bits [199:192] of the result.
/// \param __b23
/// An 8-bit integral value used to initialize bits [191:184] of the result.
/// \param __b22
/// An 8-bit integral value used to initialize bits [183:176] of the result.
/// \param __b21
/// An 8-bit integral value used to initialize bits [175:168] of the result.
/// \param __b20
/// An 8-bit integral value used to initialize bits [167:160] of the result.
/// \param __b19
/// An 8-bit integral value used to initialize bits [159:152] of the result.
/// \param __b18
/// An 8-bit integral value used to initialize bits [151:144] of the result.
/// \param __b17
/// An 8-bit integral value used to initialize bits [143:136] of the result.
/// \param __b16
/// An 8-bit integral value used to initialize bits [135:128] of the result.
/// \param __b15
/// An 8-bit integral value used to initialize bits [127:120] of the result.
/// \param __b14
/// An 8-bit integral value used to initialize bits [119:112] of the result.
/// \param __b13
/// An 8-bit integral value used to initialize bits [111:104] of the result.
/// \param __b12
/// An 8-bit integral value used to initialize bits [103:96] of the result.
/// \param __b11
/// An 8-bit integral value used to initialize bits [95:88] of the result.
/// \param __b10
/// An 8-bit integral value used to initialize bits [87:80] of the result.
/// \param __b09
/// An 8-bit integral value used to initialize bits [79:72] of the result.
/// \param __b08
/// An 8-bit integral value used to initialize bits [71:64] of the result.
/// \param __b07
/// An 8-bit integral value used to initialize bits [63:56] of the result.
/// \param __b06
/// An 8-bit integral value used to initialize bits [55:48] of the result.
/// \param __b05
/// An 8-bit integral value used to initialize bits [47:40] of the result.
/// \param __b04
/// An 8-bit integral value used to initialize bits [39:32] of the result.
/// \param __b03
/// An 8-bit integral value used to initialize bits [31:24] of the result.
/// \param __b02
/// An 8-bit integral value used to initialize bits [23:16] of the result.
/// \param __b01
/// An 8-bit integral value used to initialize bits [15:8] of the result.
/// \param __b00
/// An 8-bit integral value used to initialize bits [7:0] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_epi8(char __b31, char __b30, char __b29, char __b28,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
char __b27, char __b26, char __b25, char __b24,
char __b23, char __b22, char __b21, char __b20,
char __b19, char __b18, char __b17, char __b16,
char __b15, char __b14, char __b13, char __b12,
char __b11, char __b10, char __b09, char __b08,
char __b07, char __b06, char __b05, char __b04,
char __b03, char __b02, char __b01, char __b00)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256i)(__v32qi){
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__b00, __b01, __b02, __b03, __b04, __b05, __b06, __b07,
__b08, __b09, __b10, __b11, __b12, __b13, __b14, __b15,
__b16, __b17, __b18, __b19, __b20, __b21, __b22, __b23,
__b24, __b25, __b26, __b27, __b28, __b29, __b30, __b31
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
};
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector initialized with the specified
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 64-bit integral values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VPUNPCKLQDQ+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A 64-bit integral value used to initialize bits [255:192] of the result.
/// \param __b
/// A 64-bit integral value used to initialize bits [191:128] of the result.
/// \param __c
/// A 64-bit integral value used to initialize bits [127:64] of the result.
/// \param __d
/// A 64-bit integral value used to initialize bits [63:0] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set_epi64x(long long __a, long long __b, long long __c, long long __d)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256i)(__v4di){ __d, __c, __b, __a };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Create vectors with elements in reverse order */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double],
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// initialized in reverse order with the specified double-precision
/// floating-point values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VUNPCKLPD+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A double-precision floating-point value used to initialize bits [63:0]
/// of the result.
/// \param __b
/// A double-precision floating-point value used to initialize bits [127:64]
/// of the result.
/// \param __c
/// A double-precision floating-point value used to initialize bits [191:128]
/// of the result.
/// \param __d
/// A double-precision floating-point value used to initialize bits [255:192]
/// of the result.
/// \returns An initialized 256-bit floating-point vector of [4 x double].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_pd(double __a, double __b, double __c, double __d)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_pd(__d, __c, __b, __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float],
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// initialized in reverse order with the specified single-precision
/// float-point values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A single-precision floating-point value used to initialize bits [31:0]
/// of the result.
/// \param __b
/// A single-precision floating-point value used to initialize bits [63:32]
/// of the result.
/// \param __c
/// A single-precision floating-point value used to initialize bits [95:64]
/// of the result.
/// \param __d
/// A single-precision floating-point value used to initialize bits [127:96]
/// of the result.
/// \param __e
/// A single-precision floating-point value used to initialize bits [159:128]
/// of the result.
/// \param __f
/// A single-precision floating-point value used to initialize bits [191:160]
/// of the result.
/// \param __g
/// A single-precision floating-point value used to initialize bits [223:192]
/// of the result.
/// \param __h
/// A single-precision floating-point value used to initialize bits [255:224]
/// of the result.
/// \returns An initialized 256-bit floating-point vector of [8 x float].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_ps(float __a, float __b, float __c, float __d,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
float __e, float __f, float __g, float __h)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_ps(__h, __g, __f, __e, __d, __c, __b, __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector, initialized in reverse order
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// with the specified 32-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __i0
/// A 32-bit integral value used to initialize bits [31:0] of the result.
/// \param __i1
/// A 32-bit integral value used to initialize bits [63:32] of the result.
/// \param __i2
/// A 32-bit integral value used to initialize bits [95:64] of the result.
/// \param __i3
/// A 32-bit integral value used to initialize bits [127:96] of the result.
/// \param __i4
/// A 32-bit integral value used to initialize bits [159:128] of the result.
/// \param __i5
/// A 32-bit integral value used to initialize bits [191:160] of the result.
/// \param __i6
/// A 32-bit integral value used to initialize bits [223:192] of the result.
/// \param __i7
/// A 32-bit integral value used to initialize bits [255:224] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_epi32(int __i0, int __i1, int __i2, int __i3,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
int __i4, int __i5, int __i6, int __i7)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_epi32(__i7, __i6, __i5, __i4, __i3, __i2, __i1, __i0);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector, initialized in reverse order
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// with the specified 16-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __w15
/// A 16-bit integral value used to initialize bits [15:0] of the result.
/// \param __w14
/// A 16-bit integral value used to initialize bits [31:16] of the result.
/// \param __w13
/// A 16-bit integral value used to initialize bits [47:32] of the result.
/// \param __w12
/// A 16-bit integral value used to initialize bits [63:48] of the result.
/// \param __w11
/// A 16-bit integral value used to initialize bits [79:64] of the result.
/// \param __w10
/// A 16-bit integral value used to initialize bits [95:80] of the result.
/// \param __w09
/// A 16-bit integral value used to initialize bits [111:96] of the result.
/// \param __w08
/// A 16-bit integral value used to initialize bits [127:112] of the result.
/// \param __w07
/// A 16-bit integral value used to initialize bits [143:128] of the result.
/// \param __w06
/// A 16-bit integral value used to initialize bits [159:144] of the result.
/// \param __w05
/// A 16-bit integral value used to initialize bits [175:160] of the result.
/// \param __w04
/// A 16-bit integral value used to initialize bits [191:176] of the result.
/// \param __w03
/// A 16-bit integral value used to initialize bits [207:192] of the result.
/// \param __w02
/// A 16-bit integral value used to initialize bits [223:208] of the result.
/// \param __w01
/// A 16-bit integral value used to initialize bits [239:224] of the result.
/// \param __w00
/// A 16-bit integral value used to initialize bits [255:240] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_epi16(short __w15, short __w14, short __w13, short __w12,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
short __w11, short __w10, short __w09, short __w08,
short __w07, short __w06, short __w05, short __w04,
short __w03, short __w02, short __w01, short __w00)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_epi16(__w00, __w01, __w02, __w03,
__w04, __w05, __w06, __w07,
__w08, __w09, __w10, __w11,
__w12, __w13, __w14, __w15);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector, initialized in reverse order
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// with the specified 8-bit integral values.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic is a utility function and does not correspond to a specific
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __b31
/// An 8-bit integral value used to initialize bits [7:0] of the result.
/// \param __b30
/// An 8-bit integral value used to initialize bits [15:8] of the result.
/// \param __b29
/// An 8-bit integral value used to initialize bits [23:16] of the result.
/// \param __b28
/// An 8-bit integral value used to initialize bits [31:24] of the result.
/// \param __b27
/// An 8-bit integral value used to initialize bits [39:32] of the result.
/// \param __b26
/// An 8-bit integral value used to initialize bits [47:40] of the result.
/// \param __b25
/// An 8-bit integral value used to initialize bits [55:48] of the result.
/// \param __b24
/// An 8-bit integral value used to initialize bits [63:56] of the result.
/// \param __b23
/// An 8-bit integral value used to initialize bits [71:64] of the result.
/// \param __b22
/// An 8-bit integral value used to initialize bits [79:72] of the result.
/// \param __b21
/// An 8-bit integral value used to initialize bits [87:80] of the result.
/// \param __b20
/// An 8-bit integral value used to initialize bits [95:88] of the result.
/// \param __b19
/// An 8-bit integral value used to initialize bits [103:96] of the result.
/// \param __b18
/// An 8-bit integral value used to initialize bits [111:104] of the result.
/// \param __b17
/// An 8-bit integral value used to initialize bits [119:112] of the result.
/// \param __b16
/// An 8-bit integral value used to initialize bits [127:120] of the result.
/// \param __b15
/// An 8-bit integral value used to initialize bits [135:128] of the result.
/// \param __b14
/// An 8-bit integral value used to initialize bits [143:136] of the result.
/// \param __b13
/// An 8-bit integral value used to initialize bits [151:144] of the result.
/// \param __b12
/// An 8-bit integral value used to initialize bits [159:152] of the result.
/// \param __b11
/// An 8-bit integral value used to initialize bits [167:160] of the result.
/// \param __b10
/// An 8-bit integral value used to initialize bits [175:168] of the result.
/// \param __b09
/// An 8-bit integral value used to initialize bits [183:176] of the result.
/// \param __b08
/// An 8-bit integral value used to initialize bits [191:184] of the result.
/// \param __b07
/// An 8-bit integral value used to initialize bits [199:192] of the result.
/// \param __b06
/// An 8-bit integral value used to initialize bits [207:200] of the result.
/// \param __b05
/// An 8-bit integral value used to initialize bits [215:208] of the result.
/// \param __b04
/// An 8-bit integral value used to initialize bits [223:216] of the result.
/// \param __b03
/// An 8-bit integral value used to initialize bits [231:224] of the result.
/// \param __b02
/// An 8-bit integral value used to initialize bits [239:232] of the result.
/// \param __b01
/// An 8-bit integral value used to initialize bits [247:240] of the result.
/// \param __b00
/// An 8-bit integral value used to initialize bits [255:248] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_epi8(char __b31, char __b30, char __b29, char __b28,
[x86] Remove tab characters from avxintrin.h. NFC. llvm-svn: 227676
2015-01-31 14:33:59 +08:00
char __b27, char __b26, char __b25, char __b24,
char __b23, char __b22, char __b21, char __b20,
char __b19, char __b18, char __b17, char __b16,
char __b15, char __b14, char __b13, char __b12,
char __b11, char __b10, char __b09, char __b08,
char __b07, char __b06, char __b05, char __b04,
char __b03, char __b02, char __b01, char __b00)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_epi8(__b00, __b01, __b02, __b03, __b04, __b05, __b06, __b07,
__b08, __b09, __b10, __b11, __b12, __b13, __b14, __b15,
__b16, __b17, __b18, __b19, __b20, __b21, __b22, __b23,
__b24, __b25, __b26, __b27, __b28, __b29, __b30, __b31);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector, initialized in reverse order
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// with the specified 64-bit integral values.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VPUNPCKLQDQ+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __a
/// A 64-bit integral value used to initialize bits [63:0] of the result.
/// \param __b
/// A 64-bit integral value used to initialize bits [127:64] of the result.
/// \param __c
/// A 64-bit integral value used to initialize bits [191:128] of the result.
/// \param __d
/// A 64-bit integral value used to initialize bits [255:192] of the result.
/// \returns An initialized 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_setr_epi64x(long long __a, long long __b, long long __c, long long __d)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Fix wrong intrinsic semantic. llvm-svn: 333617
2018-05-31 09:51:07 +08:00
return _mm256_set_epi64x(__d, __c, __b, __a);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Create vectors with repeated elements */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double], with each
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of the four double-precision floating-point vector elements set to the
/// specified double-precision floating-point value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDDUP+VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __w
/// A double-precision floating-point value used to initialize each vector
/// element of the result.
/// \returns An initialized 256-bit floating-point vector of [4 x double].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_pd(double __w)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_pd(__w, __w, __w, __w);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float], with each
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of the eight single-precision floating-point vector elements set to the
/// specified single-precision floating-point value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __w
/// A single-precision floating-point value used to initialize each vector
/// element of the result.
/// \returns An initialized 256-bit floating-point vector of [8 x float].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_ps(float __w)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_ps(__w, __w, __w, __w, __w, __w, __w, __w);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector of [8 x i32], with each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 32-bit integral vector elements set to the specified 32-bit integral
/// value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VPERMILPS+VINSERTF128 </c>
/// instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __i
/// A 32-bit integral value used to initialize each vector element of the
/// result.
/// \returns An initialized 256-bit integer vector of [8 x i32].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_epi32(int __i)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_epi32(__i, __i, __i, __i, __i, __i, __i, __i);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector of [16 x i16], with each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 16-bit integral vector elements set to the specified 16-bit integral
/// value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPSHUFB+VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __w
/// A 16-bit integral value used to initialize each vector element of the
/// result.
/// \returns An initialized 256-bit integer vector of [16 x i16].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_epi16(short __w)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_epi16(__w, __w, __w, __w, __w, __w, __w, __w,
__w, __w, __w, __w, __w, __w, __w, __w);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector of [32 x i8], with each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 8-bit integral vector elements set to the specified 8-bit integral value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VPSHUFB+VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __b
/// An 8-bit integral value used to initialize each vector element of the
/// result.
/// \returns An initialized 256-bit integer vector of [32 x i8].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_epi8(char __b)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_epi8(__b, __b, __b, __b, __b, __b, __b, __b,
__b, __b, __b, __b, __b, __b, __b, __b,
__b, __b, __b, __b, __b, __b, __b, __b,
__b, __b, __b, __b, __b, __b, __b, __b);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector of [4 x i64], with each of the
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 64-bit integral vector elements set to the specified 64-bit integral
/// value.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VMOVDDUP+VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __q
/// A 64-bit integral value used to initialize each vector element of the
/// result.
/// \returns An initialized 256-bit integer vector of [4 x i64].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_set1_epi64x(long long __q)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return _mm256_set_epi64x(__q, __q, __q, __q);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
/* Create __zeroed vectors */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double] with all
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// vector elements initialized to zero.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VXORPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \returns A 256-bit vector of [4 x double] with all elements set to zero.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
_mm256_setzero_pd(void)
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256d){ 0, 0, 0, 0 };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] with all
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// vector elements initialized to zero.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VXORPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \returns A 256-bit vector of [8 x float] with all elements set to zero.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
_mm256_setzero_ps(void)
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256){ 0, 0, 0, 0, 0, 0, 0, 0 };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector initialized to zero.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VXORPS </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \returns A 256-bit integer vector initialized to zero.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
_mm256_setzero_si256(void)
{
[X86] Add __extension__ to a bunch of places in our intrinsic headers that fail if you run it through -pedantic -ansi. All of these are lines that create a 'compound literal' to concatenate elements together. llvm-svn: 333593
2018-05-31 05:08:27 +08:00
return __extension__ (__m256i)(__v4di){ 0, 0, 0, 0 };
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
/* Cast between vector types */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit floating-point vector of [4 x double] into a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// floating-point vector of [8 x float].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [4 x double].
/// \returns A 256-bit floating-point vector of [8 x float] containing the same
/// bitwise pattern as the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castpd_ps(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit floating-point vector of [4 x double] into a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// integer vector.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [4 x double].
/// \returns A 256-bit integer vector containing the same bitwise pattern as the
/// parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castpd_si256(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256i)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit floating-point vector of [8 x float] into a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// floating-point vector of [4 x double].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [8 x float].
/// \returns A 256-bit floating-point vector of [4 x double] containing the same
/// bitwise pattern as the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castps_pd(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256d)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit floating-point vector of [8 x float] into a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// integer vector.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [8 x float].
/// \returns A 256-bit integer vector containing the same bitwise pattern as the
/// parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castps_si256(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256i)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit integer vector into a 256-bit floating-point vector
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of [8 x float].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit integer vector.
/// \returns A 256-bit floating-point vector of [8 x float] containing the same
/// bitwise pattern as the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castsi256_ps(__m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Casts a 256-bit integer vector into a 256-bit floating-point vector
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of [4 x double].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit integer vector.
/// \returns A 256-bit floating-point vector of [4 x double] containing the same
/// bitwise pattern as the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castsi256_pd(__m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
return (__m256d)__a;
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Returns the lower 128 bits of a 256-bit floating-point vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// [4 x double] as a 128-bit floating-point vector of [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [4 x double].
/// \returns A 128-bit floating-point vector of [2 x double] containing the
/// lower 128 bits of the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128d __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castpd256_pd128(__m256d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4df)__a, (__v4df)__a, 0, 1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Returns the lower 128 bits of a 256-bit floating-point vector of
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// [8 x float] as a 128-bit floating-point vector of [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit floating-point vector of [8 x float].
/// \returns A 128-bit floating-point vector of [4 x float] containing the
/// lower 128 bits of the parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128 __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castps256_ps128(__m256 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v8sf)__a, (__v8sf)__a, 0, 1, 2, 3);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Truncates a 256-bit integer vector into a 128-bit integer vector.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 256-bit integer vector.
/// \returns A 128-bit integer vector containing the lower 128 bits of the
/// parameter.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m128i __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castsi256_si128(__m256i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4di)__a, (__v4di)__a, 0, 1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double] from a
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// 128-bit floating-point vector of [2 x double].
///
/// The lower 128 bits contain the value of the source vector. The contents
/// of the upper 128 bits are undefined.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \returns A 256-bit floating-point vector of [4 x double]. The lower 128 bits
/// contain the value of the parameter. The contents of the upper 128 bits
/// are undefined.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castpd128_pd256(__m128d __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v2df)__a, (__v2df)__a, 0, 1, -1, -1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] from a
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// 128-bit floating-point vector of [4 x float].
///
/// The lower 128 bits contain the value of the source vector. The contents
/// of the upper 128 bits are undefined.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \returns A 256-bit floating-point vector of [8 x float]. The lower 128 bits
/// contain the value of the parameter. The contents of the upper 128 bits
/// are undefined.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castps128_ps256(__m128 __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 0, 1, 2, 3, -1, -1, -1, -1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector from a 128-bit integer vector.
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
///
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// The lower 128 bits contain the value of the source vector. The contents
/// of the upper 128 bits are undefined.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit integer vector.
/// \returns A 256-bit integer vector. The lower 128 bits contain the value of
/// the parameter. The contents of the upper 128 bits are undefined.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Avoid names like __in that conflict with SAL in builtin headers Microsoft's Source Annotation Language (SAL) defines a bunch of keywords for annotating the inputs and outputs of functions. Empty definitions for the keywords are provided by <stdlib.h> -> <crtdefs.h> -> <sal.h>. This makes it basically impossible to include MSVC's stdlib.h and Clang's *mmintrin.h headers at the same time if they have variables named __in. As a workaround, I've renamed those variables. This fixes the Modules/compiler_builtins.m test which was XFAILed, presumably due to this conflict. llvm-svn: 179860
2013-04-20 01:00:14 +08:00
_mm256_castsi128_si256(__m128i __a)
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return __builtin_shufflevector((__v2di)__a, (__v2di)__a, 0, 1, -1, -1);
Add AVX intrinsics header llvm-svn: 110253
2010-08-05 06:03:36 +08:00
}
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double] from a
[X86][AVX] Added support for _mm256_zext* helper intrinsics (PR32839) llvm-svn: 301749
2017-04-30 01:17:06 +08:00
/// 128-bit floating-point vector of [2 x double]. The lower 128 bits
/// contain the value of the source vector. The upper 128 bits are set
/// to zero.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit vector of [2 x double].
/// \returns A 256-bit floating-point vector of [4 x double]. The lower 128 bits
/// contain the value of the parameter. The upper 128 bits are set to zero.
static __inline __m256d __DEFAULT_FN_ATTRS
_mm256_zextpd128_pd256(__m128d __a)
{
return __builtin_shufflevector((__v2df)__a, (__v2df)_mm_setzero_pd(), 0, 1, 2, 3);
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] from a
[X86][AVX] Added support for _mm256_zext* helper intrinsics (PR32839) llvm-svn: 301749
2017-04-30 01:17:06 +08:00
/// 128-bit floating-point vector of [4 x float]. The lower 128 bits contain
/// the value of the source vector. The upper 128 bits are set to zero.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit vector of [4 x float].
/// \returns A 256-bit floating-point vector of [8 x float]. The lower 128 bits
/// contain the value of the parameter. The upper 128 bits are set to zero.
static __inline __m256 __DEFAULT_FN_ATTRS
_mm256_zextps128_ps256(__m128 __a)
{
return __builtin_shufflevector((__v4sf)__a, (__v4sf)_mm_setzero_ps(), 0, 1, 2, 3, 4, 5, 6, 7);
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector from a 128-bit integer vector.
[X86][AVX] Added support for _mm256_zext* helper intrinsics (PR32839) llvm-svn: 301749
2017-04-30 01:17:06 +08:00
/// The lower 128 bits contain the value of the source vector. The upper
/// 128 bits are set to zero.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic has no corresponding instruction.
///
/// \param __a
/// A 128-bit integer vector.
/// \returns A 256-bit integer vector. The lower 128 bits contain the value of
/// the parameter. The upper 128 bits are set to zero.
static __inline __m256i __DEFAULT_FN_ATTRS
_mm256_zextsi128_si256(__m128i __a)
{
return __builtin_shufflevector((__v2di)__a, (__v2di)_mm_setzero_si128(), 0, 1, 2, 3);
}
Clean up trailing whitespace in the builtin headers llvm-svn: 247498
2015-09-12 10:55:19 +08:00
/*
[X86, AVX] Replace vinsertf128 intrinsics with generic shuffles. We want to replace as much custom x86 shuffling via intrinsics as possible because pushing the code down the generic shuffle optimization path allows for better codegen and less complexity in LLVM. This is the sibling patch for the LLVM half of this change: http://reviews.llvm.org/D8086 Differential Revision: http://reviews.llvm.org/D8088 llvm-svn: 231792
2015-03-10 23:19:26 +08:00
Vector insert.
We use macros rather than inlines because we only want to accept
invocations where the immediate M is a constant expression.
*/
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a new 256-bit vector of [8 x float] by first duplicating
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// a 256-bit vector of [8 x float] given in the first parameter, and then
/// replacing either the upper or the lower 128 bits with the contents of a
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// 128-bit vector of [4 x float] in the second parameter.
///
/// The immediate integer parameter determines between the upper or the lower
/// 128 bits.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256 _mm256_insertf128_ps(__m256 V1, __m128 V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V1
/// A 256-bit vector of [8 x float]. This vector is copied to the result
/// first, and then either the upper or the lower 128 bits of the result will
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// be replaced by the contents of \a V2.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param V2
/// A 128-bit vector of [4 x float]. The contents of this parameter are
/// written to either the upper or the lower 128 bits of the result depending
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// on the value of parameter \a M.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param M
/// An immediate integer. The least significant bit determines how the values
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// from the two parameters are interleaved: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 0, \a V2 are copied to bits [127:0] of the result,
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// and bits [255:128] of \a V1 are copied to bits [255:128] of the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, \a V2 are copied to bits [255:128] of the
/// result, and bits [127:0] of \a V1 are copied to bits [127:0] of the
/// result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 256-bit vector of [8 x float] containing the interleaved values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_insertf128_ps(V1, V2, M) \
[X86] Add subvector insert and extract builtins to enable target feature checking and immediate range checking. Test changes are due to differences in how we generate undef elements now. We also changed the types used for extractf128_si256/insertf128_si256 to match the signature of the builtin that previously existed which this patch resurrects. This also matches gcc. llvm-svn: 334261
2018-06-08 11:24:47 +08:00
(__m256)__builtin_ia32_vinsertf128_ps256((__v8sf)(__m256)(V1), \
(__v4sf)(__m128)(V2), (int)(M))
[X86, AVX] Replace vinsertf128 intrinsics with generic shuffles. We want to replace as much custom x86 shuffling via intrinsics as possible because pushing the code down the generic shuffle optimization path allows for better codegen and less complexity in LLVM. This is the sibling patch for the LLVM half of this change: http://reviews.llvm.org/D8086 Differential Revision: http://reviews.llvm.org/D8088 llvm-svn: 231792
2015-03-10 23:19:26 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a new 256-bit vector of [4 x double] by first duplicating
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// a 256-bit vector of [4 x double] given in the first parameter, and then
/// replacing either the upper or the lower 128 bits with the contents of a
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// 128-bit vector of [2 x double] in the second parameter.
///
/// The immediate integer parameter determines between the upper or the lower
/// 128 bits.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256d _mm256_insertf128_pd(__m256d V1, __m128d V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V1
/// A 256-bit vector of [4 x double]. This vector is copied to the result
/// first, and then either the upper or the lower 128 bits of the result will
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// be replaced by the contents of \a V2.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param V2
/// A 128-bit vector of [2 x double]. The contents of this parameter are
/// written to either the upper or the lower 128 bits of the result depending
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// on the value of parameter \a M.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param M
/// An immediate integer. The least significant bit determines how the values
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// from the two parameters are interleaved: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 0, \a V2 are copied to bits [127:0] of the result,
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// and bits [255:128] of \a V1 are copied to bits [255:128] of the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, \a V2 are copied to bits [255:128] of the
/// result, and bits [127:0] of \a V1 are copied to bits [127:0] of the
/// result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 256-bit vector of [4 x double] containing the interleaved values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_insertf128_pd(V1, V2, M) \
[X86] Add subvector insert and extract builtins to enable target feature checking and immediate range checking. Test changes are due to differences in how we generate undef elements now. We also changed the types used for extractf128_si256/insertf128_si256 to match the signature of the builtin that previously existed which this patch resurrects. This also matches gcc. llvm-svn: 334261
2018-06-08 11:24:47 +08:00
(__m256d)__builtin_ia32_vinsertf128_pd256((__v4df)(__m256d)(V1), \
(__v2df)(__m128d)(V2), (int)(M))
[X86, AVX] Replace vinsertf128 intrinsics with generic shuffles. We want to replace as much custom x86 shuffling via intrinsics as possible because pushing the code down the generic shuffle optimization path allows for better codegen and less complexity in LLVM. This is the sibling patch for the LLVM half of this change: http://reviews.llvm.org/D8086 Differential Revision: http://reviews.llvm.org/D8088 llvm-svn: 231792
2015-03-10 23:19:26 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a new 256-bit integer vector by first duplicating a
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// 256-bit integer vector given in the first parameter, and then replacing
/// either the upper or the lower 128 bits with the contents of a 128-bit
[DOXYGEN] Minor improvements in doxygen comments. Separated very long brief sections into two sections. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 303031
2017-05-15 11:25:04 +08:00
/// integer vector in the second parameter.
///
/// The immediate integer parameter determines between the upper or the lower
/// 128 bits.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m256i _mm256_insertf128_si256(__m256i V1, __m128i V2, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V1
/// A 256-bit integer vector. This vector is copied to the result first, and
/// then either the upper or the lower 128 bits of the result will be
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// replaced by the contents of \a V2.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param V2
/// A 128-bit integer vector. The contents of this parameter are written to
/// either the upper or the lower 128 bits of the result depending on the
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// value of parameter \a M.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param M
/// An immediate integer. The least significant bit determines how the values
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// from the two parameters are interleaved: \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 0, \a V2 are copied to bits [127:0] of the result,
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// and bits [255:128] of \a V1 are copied to bits [255:128] of the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, \a V2 are copied to bits [255:128] of the
/// result, and bits [127:0] of \a V1 are copied to bits [127:0] of the
/// result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 256-bit integer vector containing the interleaved values.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_insertf128_si256(V1, V2, M) \
[X86] Fix some typecasts in intrinsic headers that I messed up in r334261. This was caught by the Header tests, but not the CodeGen tests. llvm-svn: 334264
2018-06-08 12:09:14 +08:00
(__m256i)__builtin_ia32_vinsertf128_si256((__v8si)(__m256i)(V1), \
(__v4si)(__m128i)(V2), (int)(M))
[X86, AVX] Replace vinsertf128 intrinsics with generic shuffles. We want to replace as much custom x86 shuffling via intrinsics as possible because pushing the code down the generic shuffle optimization path allows for better codegen and less complexity in LLVM. This is the sibling patch for the LLVM half of this change: http://reviews.llvm.org/D8086 Differential Revision: http://reviews.llvm.org/D8088 llvm-svn: 231792
2015-03-10 23:19:26 +08:00
Clean up trailing whitespace in the builtin headers llvm-svn: 247498
2015-09-12 10:55:19 +08:00
/*
[X86, AVX] replace vextractf128 intrinsics with generic shuffles This is very much like D8088 (checked in at r231792). Now that we've replaced the vinsertf128 intrinsics, do the same for their extract twins. Differential Revision: http://reviews.llvm.org/D8275 llvm-svn: 232052
2015-03-12 23:50:36 +08:00
Vector extract.
We use macros rather than inlines because we only want to accept
invocations where the immediate M is a constant expression.
*/
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Extracts either the upper or the lower 128 bits from a 256-bit vector
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of [8 x float], as determined by the immediate integer parameter, and
/// returns the extracted bits as a 128-bit vector of [4 x float].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm256_extractf128_ps(__m256 V, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V
/// A 256-bit vector of [8 x float].
/// \param M
/// An immediate integer. The least significant bit determines which bits are
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// extracted from the first parameter: \n
/// If bit [0] of \a M is 0, bits [127:0] of \a V are copied to the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, bits [255:128] of \a V are copied to the result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 128-bit vector of [4 x float] containing the extracted bits.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_extractf128_ps(V, M) \
[X86] Add subvector insert and extract builtins to enable target feature checking and immediate range checking. Test changes are due to differences in how we generate undef elements now. We also changed the types used for extractf128_si256/insertf128_si256 to match the signature of the builtin that previously existed which this patch resurrects. This also matches gcc. llvm-svn: 334261
2018-06-08 11:24:47 +08:00
(__m128)__builtin_ia32_vextractf128_ps256((__v8sf)(__m256)(V), (int)(M))
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Extracts either the upper or the lower 128 bits from a 256-bit vector
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// of [4 x double], as determined by the immediate integer parameter, and
/// returns the extracted bits as a 128-bit vector of [2 x double].
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm256_extractf128_pd(__m256d V, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V
/// A 256-bit vector of [4 x double].
/// \param M
/// An immediate integer. The least significant bit determines which bits are
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// extracted from the first parameter: \n
/// If bit [0] of \a M is 0, bits [127:0] of \a V are copied to the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, bits [255:128] of \a V are copied to the result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 128-bit vector of [2 x double] containing the extracted bits.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_extractf128_pd(V, M) \
[X86] Add subvector insert and extract builtins to enable target feature checking and immediate range checking. Test changes are due to differences in how we generate undef elements now. We also changed the types used for extractf128_si256/insertf128_si256 to match the signature of the builtin that previously existed which this patch resurrects. This also matches gcc. llvm-svn: 334261
2018-06-08 11:24:47 +08:00
(__m128d)__builtin_ia32_vextractf128_pd256((__v4df)(__m256d)(V), (int)(M))
[X86, AVX] replace vextractf128 intrinsics with generic shuffles This is very much like D8088 (checked in at r231792). Now that we've replaced the vinsertf128 intrinsics, do the same for their extract twins. Differential Revision: http://reviews.llvm.org/D8275 llvm-svn: 232052
2015-03-12 23:50:36 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Extracts either the upper or the lower 128 bits from a 256-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// integer vector, as determined by the immediate integer parameter, and
/// returns the extracted bits as a 128-bit integer vector.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128i _mm256_extractf128_si256(__m256i V, const int M);
/// \endcode
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param V
/// A 256-bit integer vector.
/// \param M
/// An immediate integer. The least significant bit determines which bits are
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// extracted from the first parameter: \n
/// If bit [0] of \a M is 0, bits [127:0] of \a V are copied to the
/// result. \n
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// If bit [0] of \a M is 1, bits [255:128] of \a V are copied to the result.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \returns A 128-bit integer vector containing the extracted bits.
[X86] Remove __extension__ from macro intrinsics when its not needed. I think this is a holdover from when we used to declare variables inside the macros. And then its been copy and pasted forward for years every time a new macro intrinsic gets added. Interestingly this caused some tests for IRGen to be slightly more optimized. We now return a zeroinitializer directly instead of going through a store+load. It also removed a bogus error message on another test. llvm-svn: 333613
2018-05-31 08:51:20 +08:00
#define _mm256_extractf128_si256(V, M) \
[X86] Fix some typecasts in intrinsic headers that I messed up in r334261. This was caught by the Header tests, but not the CodeGen tests. llvm-svn: 334264
2018-06-08 12:09:14 +08:00
(__m128i)__builtin_ia32_vextractf128_si256((__v8si)(__m256i)(V), (int)(M))
[X86, AVX] replace vextractf128 intrinsics with generic shuffles This is very much like D8088 (checked in at r231792). Now that we've replaced the vinsertf128 intrinsics, do the same for their extract twins. Differential Revision: http://reviews.llvm.org/D8275 llvm-svn: 232052
2015-03-12 23:50:36 +08:00
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
/* SIMD load ops (unaligned) */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads two 128-bit floating-point vectors of [4 x float] from
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// unaligned memory locations and constructs a 256-bit floating-point vector
/// of [8 x float] by concatenating the two 128-bit vectors.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to load instructions followed by the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
/// A pointer to a 128-bit memory location containing 4 consecutive
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// single-precision floating-point values. These values are to be copied to
/// bits[255:128] of the result. The address of the memory location does not
/// have to be aligned.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param __addr_lo
/// A pointer to a 128-bit memory location containing 4 consecutive
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// single-precision floating-point values. These values are to be copied to
/// bits[127:0] of the result. The address of the memory location does not
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// have to be aligned.
/// \returns A 256-bit floating-point vector of [8 x float] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu2_m128(float const *__addr_hi, float const *__addr_lo)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
[X86] Use unaligned load intrinsics to implement other intrinsics instead of manually creating the unaligned load. llvm-svn: 271250
2016-05-31 13:49:13 +08:00
__m256 __v256 = _mm256_castps128_ps256(_mm_loadu_ps(__addr_lo));
return _mm256_insertf128_ps(__v256, _mm_loadu_ps(__addr_hi), 1);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads two 128-bit floating-point vectors of [2 x double] from
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// unaligned memory locations and constructs a 256-bit floating-point vector
/// of [4 x double] by concatenating the two 128-bit vectors.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to load instructions followed by the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
/// A pointer to a 128-bit memory location containing two consecutive
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// double-precision floating-point values. These values are to be copied to
/// bits[255:128] of the result. The address of the memory location does not
/// have to be aligned.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// \param __addr_lo
/// A pointer to a 128-bit memory location containing two consecutive
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// double-precision floating-point values. These values are to be copied to
/// bits[127:0] of the result. The address of the memory location does not
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// have to be aligned.
/// \returns A 256-bit floating-point vector of [4 x double] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu2_m128d(double const *__addr_hi, double const *__addr_lo)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
[X86] Use unaligned load intrinsics to implement other intrinsics instead of manually creating the unaligned load. llvm-svn: 271250
2016-05-31 13:49:13 +08:00
__m256d __v256 = _mm256_castpd128_pd256(_mm_loadu_pd(__addr_lo));
return _mm256_insertf128_pd(__v256, _mm_loadu_pd(__addr_hi), 1);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Loads two 128-bit integer vectors from unaligned memory locations and
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// constructs a 256-bit integer vector by concatenating the two 128-bit
/// vectors.
///
/// \headerfile <x86intrin.h>
///
/// This intrinsic corresponds to load instructions followed by the
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
/// A pointer to a 128-bit memory location containing a 128-bit integer
/// vector. This vector is to be copied to bits[255:128] of the result. The
/// address of the memory location does not have to be aligned.
/// \param __addr_lo
/// A pointer to a 128-bit memory location containing a 128-bit integer
/// vector. This vector is to be copied to bits[127:0] of the result. The
/// address of the memory location does not have to be aligned.
/// \returns A 256-bit integer vector containing the concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_loadu2_m128i(__m128i const *__addr_hi, __m128i const *__addr_lo)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
[X86] Use unaligned load intrinsics to implement other intrinsics instead of manually creating the unaligned load. llvm-svn: 271250
2016-05-31 13:49:13 +08:00
__m256i __v256 = _mm256_castsi128_si256(_mm_loadu_si128(__addr_lo));
return _mm256_insertf128_si256(__v256, _mm_loadu_si128(__addr_hi), 1);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
/* SIMD store ops (unaligned) */
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores the upper and lower 128 bits of a 256-bit floating-point
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// vector of [8 x float] into two different unaligned memory locations.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction and the
/// store instructions.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[255:128] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __addr_lo
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[127:0] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __a
/// A 256-bit floating-point vector of [8 x float].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu2_m128(float *__addr_hi, float *__addr_lo, __m256 __a)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__m128 __v128;
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_castps256_ps128(__a);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_ps(__addr_lo, __v128);
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_extractf128_ps(__a, 1);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_ps(__addr_hi, __v128);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores the upper and lower 128 bits of a 256-bit floating-point
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// vector of [4 x double] into two different unaligned memory locations.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction and the
/// store instructions.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[255:128] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __addr_lo
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[127:0] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __a
/// A 256-bit floating-point vector of [4 x double].
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu2_m128d(double *__addr_hi, double *__addr_lo, __m256d __a)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__m128d __v128;
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_castpd256_pd128(__a);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_pd(__addr_lo, __v128);
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_extractf128_pd(__a, 1);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_pd(__addr_hi, __v128);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Stores the upper and lower 128 bits of a 256-bit integer vector into
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// two different unaligned memory locations.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for tmmintrin.h intrinsics. Added \n commands to insert a line breaks where necessary to make the documentation more readable. Formatted comments to fit into 80 chars. llvm-svn: 290458
2016-12-24 07:36:26 +08:00
/// This intrinsic corresponds to the <c> VEXTRACTF128 </c> instruction and the
/// store instructions.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __addr_hi
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[255:128] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __addr_lo
[DOXYGEN] Improved doxygen comments for avxintrin.h intrinsics. Tagged parameter names with \a doxygen command to display them in italics. Formatted comments to fit into 80 chars. llvm-svn: 289022
2016-12-08 12:09:17 +08:00
/// A pointer to a 128-bit memory location. Bits[127:0] of \a __a are to be
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// copied to this memory location. The address of this memory location does
/// not have to be aligned.
/// \param __a
/// A 256-bit integer vector.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline void __DEFAULT_FN_ATTRS
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
_mm256_storeu2_m128i(__m128i *__addr_hi, __m128i *__addr_lo, __m256i __a)
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
{
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__m128i __v128;
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_castsi256_si128(__a);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_si128(__addr_lo, __v128);
PR14964: intrinsic headers using non-reserved identifiers Several of the intrinsic headers were using plain non-reserved identifiers. C++11 17.6.4.3.2 [global.names] p1 reservers names containing a double begining with an underscore followed by an uppercase letter for any use. I think I got them all, but open to being corrected. For the most part I didn't bother updating function-like macro parameter names because I don't believe they're subject to any such collission - though some function-like macros already follow this convention (I didn't update them in part because the churn was more significant as several function-like macros use the double underscore prefixed version of the same name as a parameter in their implementation) llvm-svn: 172666
2013-01-17 07:08:36 +08:00
__v128 = _mm256_extractf128_si256(__a, 1);
[X86] Replace unaligned store builtins in SSE/AVX intrinsic files with code that will compile to a native unaligned store. Remove the builtins since they are no longer used. Intrinsics will be removed from llvm in a future commit. llvm-svn: 271214
2016-05-31 01:10:30 +08:00
_mm_storeu_si128(__addr_hi, __v128);
[avx] Define the _mm256_loadu2_xxx and _mm256_storeu2_xxx intrinsics. From the Intel Optimization Reference Manual, Section 11.6.2. When data cannot be aligned or alignment is not known, 16-byte memory accesses may provide better performance. rdar://11076953 llvm-svn: 153091
2012-03-21 00:40:00 +08:00
}
Add missing include guards into headers in lib/Headers. While it may appear that these headers should not be included more than once, they are in fact included twice when building our builtins module (in order for it to generate submodules for them), and without this, any modular build enabling AVX and including any builtin header fails. Testing this is tricky because including any of these headers in a modular build is liable to fail, due to unrelated builtin headers in the same module including headers which might not be available on the system running the tests. Suggestion on that front are welcome (but we're getting close to being able to run a buildbot that has modules enabled for all tests, which would nicely solve the testing problem). llvm-svn: 186275
2013-07-14 13:41:45 +08:00
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] by
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// concatenating two 128-bit floating-point vectors of [4 x float].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __hi
/// A 128-bit floating-point vector of [4 x float] to be copied to the upper
/// 128 bits of the result.
/// \param __lo
/// A 128-bit floating-point vector of [4 x float] to be copied to the lower
/// 128 bits of the result.
/// \returns A 256-bit floating-point vector of [8 x float] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_set_m128 (__m128 __hi, __m128 __lo)
{
[X86] Add typecasts to remove most assumptions about what __m128i/__m256i is defined as. Add similar typecasts for the fp types as well. llvm-svn: 269632
2016-05-16 14:38:42 +08:00
return (__m256) __builtin_shufflevector((__v4sf)__lo, (__v4sf)__hi, 0, 1, 2, 3, 4, 5, 6, 7);
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double] by
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// concatenating two 128-bit floating-point vectors of [2 x double].
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __hi
/// A 128-bit floating-point vector of [2 x double] to be copied to the upper
/// 128 bits of the result.
/// \param __lo
/// A 128-bit floating-point vector of [2 x double] to be copied to the lower
/// 128 bits of the result.
/// \returns A 256-bit floating-point vector of [4 x double] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_set_m128d (__m128d __hi, __m128d __lo)
{
[X86] Fix various type mismatches in intrinsic headers and intrinsic tests that cause extra bitcasts to be emitted in the IR. Found via imprecise grepping of the -O0 IR. There could still be more bugs out there. llvm-svn: 336487
2018-07-08 01:03:32 +08:00
return (__m256d) __builtin_shufflevector((__v2df)__lo, (__v2df)__hi, 0, 1, 2, 3);
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector by concatenating two 128-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// integer vectors.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __hi
/// A 128-bit integer vector to be copied to the upper 128 bits of the
/// result.
/// \param __lo
/// A 128-bit integer vector to be copied to the lower 128 bits of the
/// result.
/// \returns A 256-bit integer vector containing the concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_set_m128i (__m128i __hi, __m128i __lo)
{
[X86] Fix various type mismatches in intrinsic headers and intrinsic tests that cause extra bitcasts to be emitted in the IR. Found via imprecise grepping of the -O0 IR. There could still be more bugs out there. llvm-svn: 336487
2018-07-08 01:03:32 +08:00
return (__m256i) __builtin_shufflevector((__v2di)__lo, (__v2di)__hi, 0, 1, 2, 3);
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [8 x float] by
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// concatenating two 128-bit floating-point vectors of [4 x float]. This is
/// similar to _mm256_set_m128, but the order of the input parameters is
/// swapped.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __lo
/// A 128-bit floating-point vector of [4 x float] to be copied to the lower
/// 128 bits of the result.
/// \param __hi
/// A 128-bit floating-point vector of [4 x float] to be copied to the upper
/// 128 bits of the result.
/// \returns A 256-bit floating-point vector of [8 x float] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256 __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_setr_m128 (__m128 __lo, __m128 __hi)
{
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
return _mm256_set_m128(__hi, __lo);
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit floating-point vector of [4 x double] by
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// concatenating two 128-bit floating-point vectors of [2 x double]. This is
/// similar to _mm256_set_m128d, but the order of the input parameters is
/// swapped.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __lo
/// A 128-bit floating-point vector of [2 x double] to be copied to the lower
/// 128 bits of the result.
/// \param __hi
/// A 128-bit floating-point vector of [2 x double] to be copied to the upper
/// 128 bits of the result.
/// \returns A 256-bit floating-point vector of [4 x double] containing the
/// concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256d __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_setr_m128d (__m128d __lo, __m128d __hi)
{
[X86] Fix various type mismatches in intrinsic headers and intrinsic tests that cause extra bitcasts to be emitted in the IR. Found via imprecise grepping of the -O0 IR. There could still be more bugs out there. llvm-svn: 336487
2018-07-08 01:03:32 +08:00
return (__m256d)_mm256_set_m128d(__hi, __lo);
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
}
Remove \brief commands from doxygen comments. This is similar to the LLVM change https://reviews.llvm.org/D46290. We've been running doxygen with the autobrief option for a couple of years now. This makes the \brief markers into our comments redundant. Since they are a visual distraction and we don't want to encourage more \brief markers in new code either, this patch removes them all. Patch produced by for i in $(git grep -l '\@brief'); do perl -pi -e 's/\@brief //g' $i & done for i in $(git grep -l '\\brief'); do perl -pi -e 's/\\brief //g' $i & done Differential Revision: https://reviews.llvm.org/D46320 llvm-svn: 331834
2018-05-09 09:00:01 +08:00
/// Constructs a 256-bit integer vector by concatenating two 128-bit
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
/// integer vectors. This is similar to _mm256_set_m128i, but the order of
/// the input parameters is swapped.
///
/// \headerfile <x86intrin.h>
///
[DOXYGEN] Improved doxygen comments for x86 intrinsics headers. Tagged instruction names with <c> INSTR_NAME </c> to display them in typewriter font. In the past, \c command was used, unfortunately it applied to only one word. <c> .. </c> has the same meaning, but applies to all words in between the tags. llvm-svn: 289249
2016-12-10 02:35:50 +08:00
/// This intrinsic corresponds to the <c> VINSERTF128 </c> instruction.
Doxygen comments for avxintrin.h. Added doxygen comments to avxintrin.h's intrinsics. As of now, around 75% of the intrinsics in this file are documented here. The patches for the other 25% will be se nt out later. Removed extra spaces in emmitrin.h. Note: The doxygen comments are automatically generated based on Sony's intrinsics document. I got an OK from Eric Christopher to commit doxygen comments without prior code review upstream. llvm-svn: 286336
2016-11-09 11:58:30 +08:00
///
/// \param __lo
/// A 128-bit integer vector to be copied to the lower 128 bits of the
/// result.
/// \param __hi
/// A 128-bit integer vector to be copied to the upper 128 bits of the
/// result.
/// \returns A 256-bit integer vector containing the concatenated result.
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
static __inline __m256i __DEFAULT_FN_ATTRS
Minor changes in x86 intrinsics headers; NFC I made several changes for consistency with the rest of x86 instrinsics header files. Some of these changes help to render doxygen comments better. 1. avxintrin.h – Moved the opening bracket on a separate line for several intrinsics (for consistency with the rest of the intrinsics). 2. emmintrin.h - Moved the doxygen comment next to the body of the function; - Added braces after extern "C" even though there is only one declaration each time 3. xmmintrin.h - Moved the doxygen comment next to the body of the function; - Added intrinsic prototypes for a couple of macro definitions into the doxygen comment; - Added braces after extern "C" even though there is only one declaration each time 4. ammintrin.h – Removed extra line between the doxygen comment and the body of the functions (for consistency with the rest of the files). Desk reviewed by Paul Robinson. llvm-svn: 287278
2016-11-18 07:02:00 +08:00
_mm256_setr_m128i (__m128i __lo, __m128i __hi)
{
[X86] Fix various type mismatches in intrinsic headers and intrinsic tests that cause extra bitcasts to be emitted in the IR. Found via imprecise grepping of the -O0 IR. There could still be more bugs out there. llvm-svn: 336487
2018-07-08 01:03:32 +08:00
return (__m256i)_mm256_set_m128i(__hi, __lo);
[X86] Add _mm256_set_m128 and its 5 variants. Differential Revision: http://reviews.llvm.org/D9855 llvm-svn: 237778
2015-05-20 15:46:52 +08:00
}
[X86] Rename DEFAULT_FN_ATTR macro to __DEFAULT_FN_ATTR llvm-svn: 241065
2015-06-30 21:36:19 +08:00
#undef __DEFAULT_FN_ATTRS
[Builtins][Attributes][X86] Tag all X86 builtins with their required vector width. Add a min_vector_width function attribute and tag all x86 instrinsics with it This is part of an ongoing attempt at making 512 bit vectors illegal in the X86 backend type legalizer due to CPU frequency penalties associated with wide vectors on Skylake Server CPUs. We want the loop vectorizer to be able to emit IR containing wide vectors as intermediate operations in vectorized code and allow these wide vectors to be legalized to 256 bits by the X86 backend even though we are targetting a CPU that supports 512 bit vectors. This is similar to what happens with an AVX2 CPU, the vectorizer can emit wide vectors and the backend will split them. We want this splitting behavior, but still be able to use new Skylake instructions that work on 256-bit vectors and support things like masking and gather/scatter. Of course if the user uses explicit vector code in their source code we need to not split those operations. Especially if they have used any of the 512-bit vector intrinsics from immintrin.h. And we need to make it so that merely using the intrinsics produces the expected code in order to be backwards compatible. To support this goal, this patch adds a new IR function attribute "min-legal-vector-width" that can indicate the need for a minimum vector width to be legal in the backend. We need to ensure this attribute is set to the largest vector width needed by any intrinsics from immintrin.h that the function uses. The inliner will be reponsible for merging this attribute when a function is inlined. We may also need a way to limit inlining in the future as well, but we can discuss that in the future. To make things more complicated, there are two different ways intrinsics are implemented in immintrin.h. Either as an always_inline function containing calls to builtins(can be target specific or target independent) or vector extension code. Or as a macro wrapper around a taget specific builtin. I believe I've removed all cases where the macro was around a target independent builtin. To support the always_inline function case this patch adds attribute((min_vector_width(128))) that can be used to tag these functions with their vector width. All x86 intrinsic functions that operate on vectors have been tagged with this attribute. To support the macro case, all x86 specific builtins have also been tagged with the vector width that they require. Use of any builtin with this property will implicitly increase the min_vector_width of the function that calls it. I've done this as a new property in the attribute string for the builtin rather than basing it on the type string so that we can opt into it on a per builtin basis and avoid any impact to target independent builtins. There will be future work to support vectors passed as function arguments and supporting inline assembly. And whatever else we can find that isn't covered by this patch. Special thanks to Chandler who suggested this direction and reviewed a preview version of this patch. And thanks to Eric Christopher who has had many conversations with me about this issue. Differential Revision: https://reviews.llvm.org/D48617 llvm-svn: 336583
2018-07-10 03:00:16 +08:00
#undef __DEFAULT_FN_ATTRS128
Use a define for per-file function attributes for the Intel intrinsic headers. This is a precursor to changing them to use the new target attribute code. llvm-svn: 239882
2015-06-17 15:09:20 +08:00
Add missing include guards into headers in lib/Headers. While it may appear that these headers should not be included more than once, they are in fact included twice when building our builtins module (in order for it to generate submodules for them), and without this, any modular build enabling AVX and including any builtin header fails. Testing this is tricky because including any of these headers in a modular build is liable to fail, due to unrelated builtin headers in the same module including headers which might not be available on the system running the tests. Suggestion on that front are welcome (but we're getting close to being able to run a buildbot that has modules enabled for all tests, which would nicely solve the testing problem). llvm-svn: 186275
2013-07-14 13:41:45 +08:00
#endif /* __AVXINTRIN_H */