Commit Graph

1624 Commits

Author SHA1 Message Date
Justin Holewinski c05323dd5c Un-XFAIL CodeGen/builtins-nvptx.c now that the proper changes have
landed in LLVM core

llvm-svn: 157418
2012-05-24 21:39:33 +00:00
John McCall d65dbd8e6a XFAIL this test, which does not pass on trunk since the grand
renaming in r157403.

llvm-svn: 157413
2012-05-24 20:58:21 +00:00
Justin Holewinski 83e9668133 Replace PTX back-end with NVPTX back-end in all places where Clang cares
NV_CONTRIB

llvm-svn: 157403
2012-05-24 17:43:12 +00:00
Richard Smith 01ade177e9 If the first argument of __builtin_object_size can be folded to a constant
pointer, but such folding encounters side-effects, ignore the side-effects
rather than performing them at runtime: CodeGen generates wrong code for
__builtin_object_size in that case.

llvm-svn: 157310
2012-05-23 04:13:20 +00:00
Nuno Lopes a425589fea wire -fbounds-checking to the new LLVM bounds checking pass
llvm-svn: 157262
2012-05-22 17:19:45 +00:00
Nuno Lopes 2b1ff46ed1 revert the usage of the objectsize intrinsic with 3 parameters (to match LLVM r157255)
llvm-svn: 157256
2012-05-22 15:26:48 +00:00
Akira Hatanaka b08cb81f37 XFAIL this test on MIPS.
Since r156650, clang has stopped emitting byval arguments for MIPS targets.

llvm-svn: 156954
2012-05-16 22:06:47 +00:00
Rafael Espindola bac0fdbd0b We were already passing the case in pr9356. Add it to attributes.c to make
sure we don't regress. I will add a warning in a sec.

llvm-svn: 156726
2012-05-13 01:56:10 +00:00
Benjamin Kramer 29c2b4330c Add support for __attribute__((hot)) and __attribute__((cold)).
Currently cold functions are marked with the "optsize" attribute in CodeGen
so they are always optimized for size.  The hot attribute is just ignored,
LLVM doesn't have a way to express hotness at the moment.

llvm-svn: 156723
2012-05-12 21:10:52 +00:00
Akira Hatanaka 8ab86cb344 Coerce byval aggregate arguments to integers whose size matches the integer
register size of the target architecture.
 

llvm-svn: 156650
2012-05-11 21:56:58 +00:00
Akira Hatanaka c37eddf7a7 Fix handling of vector return types.
A vector should be returned via the hidden pointer argument except if its size
is equal to or smaller than 16-bytes and the target ABI is N32 or N64.

llvm-svn: 156642
2012-05-11 21:01:17 +00:00
Nuno Lopes 729e6a056c hopefully fix a bunch of ARM buildbot failures
llvm-svn: 156513
2012-05-09 21:21:49 +00:00
Nuno Lopes ddcce0bb90 update calls to objectsize intrinsic to match LLVM r156473
add a test for -fbounds-checking code generation

llvm-svn: 156474
2012-05-09 15:53:34 +00:00
Eric Christopher c0e7a3d78c The address of a vla is actually complex and requires a dereference.
Part of rdar://11352000

llvm-svn: 156407
2012-05-08 18:56:47 +00:00
Nuno Lopes 9f5616a005 hopefully unbreak some buildbots
llvm-svn: 156357
2012-05-08 01:54:53 +00:00
Nuno Lopes f4c7679f35 make this test a little more useful. I'll add more tests for bound checking once -fbounds-checking patch gets in
llvm-svn: 156321
2012-05-07 20:23:03 +00:00
Alexey Samsonov 74a3868dbb This patch adds a new Clang compiler flag "-gline-tables-only".
It reduces the amount of emitted debug information:
1) DIEs in .debug_info have types DW_TAG_compile_unit, DW_TAG_subprogram,
DW_TAG_inlined_subroutine (for opt builds) and DW_TAG_lexical_block only.
2) .debug_str contains only function names.
3) No debug data for types/namespaces/variables is emitted.
4) The data in .debug_line is enough to produce valid stack traces with
function names and line numbers.

Reviewed by Eric Christopher.

llvm-svn: 156160
2012-05-04 07:39:27 +00:00
John McCall 1db0a2fab2 During block layout, after padding up to the max field alignment,
the alignment might actually exceed the max field alignment;  don't
assert in this case.

llvm-svn: 155937
2012-05-01 20:28:00 +00:00
Chad Rosier 87622b8b84 Get rid of storelv4si builtin as it can be expressed directly. This is general
goodness because it provides opportunites to cleanup things.  For example,

uint64_t t1(__m128i vA)
{
  uint64_t Alo;
  _mm_storel_epi64((__m128i*)&Alo, vA);
  return Alo;
}

was generating 

	movq	%xmm0, -8(%rbp)
	movq	-8(%rbp), %rax

and now generates

	movd	%xmm0, %rax

rdar://11282581

llvm-svn: 155924
2012-05-01 18:11:51 +00:00
Julien Lerouge 4a5b444371 Currently __builtin_annotation() only annotates an i32.
i32 __builtin_annotation(i32, string);

Applying it to i64 (e.g., long long) generates the following IR.

        trunc i64 {{.*}} to i32
        call i32 @llvm.annotation.i32
        zext i32 {{.*}} to i64

The redundant truncation and extension make the result difficult to use.

This patch makes __builtin_annotation() generic.

        type __builtin_annotation(type, string);

For the i64 example, it simplifies the generated IR to:

        call i64 @llvm.annotation.i64

Patch by Xi Wang!

llvm-svn: 155764
2012-04-28 17:39:16 +00:00
John McCall ac0350a396 Fix a bug with block layout when the block contains something
more aligned than the block header but also contains something
smaller than the block-header alignment but not exactly half
the difference between the large alignment and the header
alignment.  Got that?

I'm really not sure what I was thinking with the buggy computation
here, but the fix is pretty obvious.

llvm-svn: 155662
2012-04-26 21:14:42 +00:00
Chad Rosier 155005c071 Add atan, atan2, exp, and log to the builtin math library functions.
With -fno-math-errno (the default for Darwin) or -ffast-math these library 
function can be marked readnone enabling more opportunities for CSE and other
optimizations.
rdar://11251464

llvm-svn: 155498
2012-04-24 22:40:01 +00:00
Kostya Serebryany 5dd2cfc81e enable TBAA when -fthread-sanitizer is given, even with -O0 or -relaxed-aliasing
llvm-svn: 155430
2012-04-24 06:57:01 +00:00
Eric Christopher 3cc207bbcc Forward declarations should take a context. This helps the debugger
find forward declarations in the context that the actual definition
will occur.

rdar://11291658

llvm-svn: 155381
2012-04-23 19:00:24 +00:00
Anton Korobeynikov df32b6f2de Do not use stdint.h, driver might provide invalid location for it. Instead, provide the types directly.
This should fix PR12628

llvm-svn: 155342
2012-04-23 09:02:13 +00:00
Craig Topper 26e74e50b6 Convert vperm2f128 and vperm2i128 intrinsics back to using llvm intrinsics. Unfortunately, these instructions have behavior that can't be modeled with shuffle vector.
llvm-svn: 154906
2012-04-17 05:16:56 +00:00
Eli Friedman 114341db6d Attempt to fix test.
llvm-svn: 154897
2012-04-17 01:57:28 +00:00
Chad Rosier a750d46c9f Make sure EmitMoveFromReturnSlot is passing the correct alignment to
EmitFinalDestCopy (and thus pass EmitAggregateCopy the correct alignment).
rdar://11220251

llvm-svn: 154883
2012-04-17 00:35:38 +00:00
Duncan Sands 54c05faade Remove support for -fast-math metadata for the moment.
llvm-svn: 154851
2012-04-16 19:40:49 +00:00
Duncan Sands 8030b810de Generate fpmath metadata when -ffast-math. Note that no optimizations are hooked
up to this yet.

llvm-svn: 154835
2012-04-16 17:24:31 +00:00
Eli Friedman 7f1ff60021 Propagate alignment on lvalues through EmitLValueForField. PR12395.
llvm-svn: 154789
2012-04-16 03:54:45 +00:00
Craig Topper 8e57855ea0 Change _mm256_permute4x64_epi64 and _mm256_permute4x64_pd to use builtin_shufflevector instead of specific builtins. Old builtins will be removed from llvm now that vpermq/vpermpd are supported by shuffle lowering code.
llvm-svn: 154777
2012-04-15 22:18:10 +00:00
Duncan Sands c720e78ebc Use MDBuilder to help with metadata creation.
llvm-svn: 154767
2012-04-15 18:04:54 +00:00
Anton Korobeynikov 20735d2df8 First set of tests for ARM homogenous aggregates. C only. C++ will follow.
llvm-svn: 154666
2012-04-13 11:23:39 +00:00
Richard Smith d65cee9423 Implement __atomic_fetch_nand and __atomic_nand_fetch to complete our set of
GNU __atomic builtins.

llvm-svn: 154659
2012-04-13 06:31:38 +00:00
Richard Smith 01ba47d7b6 Implement the missing pieces needed to support libstdc++4.7's <atomic>:
__atomic_test_and_set, __atomic_clear, plus a pile of undocumented __GCC_*
predefined macros.

Implement library fallback for __atomic_is_lock_free and
__c11_atomic_is_lock_free, and implement __atomic_always_lock_free.

Contrary to their documentation, GCC's __atomic_fetch_add family don't
multiply the operand by sizeof(T) when operating on a pointer type.
libstdc++ relies on this quirk. Remove this handling for all but the
__c11_atomic_fetch_add and __c11_atomic_fetch_sub builtins.

Contrary to their documentation, __atomic_test_and_set and __atomic_clear
take a first argument of type 'volatile void *', not 'void *' or 'bool *',
and __atomic_is_lock_free and __atomic_always_lock_free have an argument
of type 'const volatile void *', not 'void *'.

With this change, libstdc++4.7's <atomic> passes libc++'s atomic test suite,
except for a couple of libstdc++ bugs and some cases where libc++'s test
suite tests for properties which implementations have latitude to vary.

llvm-svn: 154640
2012-04-13 00:45:38 +00:00
Richard Smith feea883de4 Implement support for 18 of the GNU-compatible __atomic builtins.
This is not quite sufficient for libstdc++'s <atomic>: we still need
__atomic_test_and_set and __atomic_clear, and may need a more complete
__atomic_is_lock_free implementation.

We are also missing an implementation of __atomic_always_lock_free,
__atomic_nand_fetch, and __atomic_fetch_nand, but those aren't needed
for libstdc++.

llvm-svn: 154579
2012-04-12 05:08:17 +00:00
Richard Smith b1e36c662b Provide, and document, a set of __c11_atomic_* intrinsics to implement C11's
<stdatomic.h> header.

In passing, fix LanguageExtensions to note that C11 and C++11 are no longer
"upcoming standards" but are now actually standardized.

llvm-svn: 154513
2012-04-11 17:55:32 +00:00
Richard Smith aa22a8cdcf AtomicExpr: make ASTStmtReader a friend and remove setters. Also fix saving
of an uninitialized Stmt* in serialization of __atomic_init and add a test of
atomics serialization.

llvm-svn: 154448
2012-04-10 22:49:28 +00:00
Eric Christopher eacf8056b1 This test actually needs label names.
PR9796

llvm-svn: 154422
2012-04-10 18:36:52 +00:00
Eric Christopher 0241e32304 For debug and coverage analysis if we're not optimizing go ahead
and emit a relatively empty block for a plain break statement. This
enables us to track where we went through a switch.

PR9796 & rdar://11215207

llvm-svn: 154420
2012-04-10 18:20:19 +00:00
Eli Friedman 1fa36050ab Make the variant of __builtin_shufflevector that takes the shuffle indexes as a vector actually usable. Patch by David Neto. PR12465.
llvm-svn: 154128
2012-04-05 21:48:40 +00:00
Eric Christopher b7e28ccb3d Add more constraint registers for mips.
Patch by Jack Carter. Testcase cleanup by me.

llvm-svn: 153921
2012-04-03 01:16:32 +00:00
Craig Topper 74c17c65e4 Correctly check argument types for some vector macros in smmintrin.h. Put parentheses around uses of vector macro arguments.
llvm-svn: 153732
2012-03-30 07:01:17 +00:00
Craig Topper 97f042f2d6 Add _mm_minpos_epu16 to smmintrin.h. Fixes PR12399.
llvm-svn: 153726
2012-03-30 05:41:28 +00:00
Craig Topper 678a53c350 Fix shuffle vector calculation for mm_permute_ps. Fixes PR 12401.
llvm-svn: 153724
2012-03-30 05:09:18 +00:00
Bill Wendling 82b87f19e2 The UTF16 string referenced by a CFString should go into the __TEXT,__ustring
section. A 'normal' string will go into the __TEXT,__const section, but this
isn't good for UTF16 strings. The __ustring section allows for coalescing, among
other niceties (such as allowing the linker to easily split up strings).

Instead of outputting the UTF16 string as a series of bytes, output it as a
series of shorts. The back-end will then nicely place the UTF16 string into the
correct section, because it's a mensch.
<rdar://problem/10655949>

llvm-svn: 153710
2012-03-30 00:26:17 +00:00
David Chisnall abf320d40e Relax the test slightly so that it should pass on other people's computers.
llvm-svn: 153670
2012-03-29 18:41:08 +00:00
David Chisnall db365f38ea Call out to GCC-compatible runtime functions for atomic ops that we can't use
LLVM intrinsics for.

I have an implementation of these functions, which wants to go in a libgcc_s
equivalent in compiler-rt.  It's currently here:

http://people.freebsd.org/~theraven/atomic.c

It will be committed to compiler-rt as soon as I work out where would be a
sensible place to put it...

llvm-svn: 153666
2012-03-29 18:01:11 +00:00
Eric Christopher 8b6d051ea2 Add better support for $fp and $sp for mips inline asm support.
Patch by Jack Carter.

llvm-svn: 153530
2012-03-27 19:56:11 +00:00
Daniel Dunbar eec0abefb6 [tests] Fix test failure in release mode.
llvm-svn: 153460
2012-03-26 20:18:16 +00:00
Kostya Serebryany 141e46faf0 add tbaa metadata to vtable pointer loads/stores
llvm-svn: 153447
2012-03-26 17:03:51 +00:00
Rafael Espindola 5c0034a7c6 Add back r153360 with a fix for enums that cover all the 32 bit values.
Thanks to NAKAMURA Takumi for finding it!

llvm-svn: 153383
2012-03-24 16:50:34 +00:00
NAKAMURA Takumi 2681efcc95 Revert r153360 (and r153380), "Second part of PR12251. Produce the range metadata in clang for booleans and".
For i686 targets (eg. cygwin), I saw "Range must not be empty!" in verifier.

It produces (i32)[0x80000000:0x80000000) from (uint64_t)[0xFFFFFFFF80000000ULL:0x0000000080000000ULL), for signed i32 on MDNode::Range.

llvm-svn: 153382
2012-03-24 14:43:42 +00:00
Rafael Espindola 54355820e8 Second part of PR12251. Produce the range metadata in clang for booleans and
c++ enums.

llvm-svn: 153360
2012-03-24 00:28:06 +00:00
Bill Wendling f351774f7b Small cleanup.
llvm-svn: 153292
2012-03-22 23:34:01 +00:00
Bill Wendling 9382fdb163 Use correct FileCheck regexp.
llvm-svn: 153291
2012-03-22 23:32:07 +00:00
Bill Wendling 3b9eaa177d When an MMX output variable is tied to the input variable, we have to implicitly
cast the value to x86_mmx. This gives the ASM string the correct call signature.
<rdar://problem/10919182>

llvm-svn: 153290
2012-03-22 23:25:07 +00:00
Matt Beaumont-Gay 1617ee30fb line endings
llvm-svn: 153289
2012-03-22 23:15:04 +00:00
Eli Friedman 610bb87e17 Make sure we correctly set the alignment for vector loads and stores associated with vector element lvalues. Patch by Kevin Schoedel (with some minor modifications by me).
llvm-svn: 153285
2012-03-22 22:36:39 +00:00
Eli Friedman fefe0d07ea Don't try to create "store atomic" instructions of non-integer types; they aren't supported at the moment. PR12040.
llvm-svn: 152891
2012-03-16 01:48:04 +00:00
Eli Friedman 6f87b44fe3 When a variable has a specified asm name, but isn't using the register storage class, the asm name doesn't specify a register. PR12244.
llvm-svn: 152873
2012-03-15 23:12:51 +00:00
Chad Rosier 9c76d24f9c [frontend] Fix how the frontend handles -fno-inline. AFAIK clang now matches
the behavior of gcc with respect to the -fno-inline and -fno-inline-functions
flags.

llvm-svn: 152861
2012-03-15 22:31:42 +00:00
James Molloy 4813fc8ed6 Fix codegen for vld{3,4}_dup intrinsics.
Patch by Silviu Baranga!

llvm-svn: 152788
2012-03-15 09:12:01 +00:00
Chad Rosier 883b2579aa [frontend] Add support for disabling the "inline" keyword using
-fno-inline-functions.

This behaves much like -fno-inline in gcc, but based on a discussion with 
Daniel it was decided that -fno-inline-functions should subsume -fno-inline.
Please speak up if you object.  The -fno-inline flag remains ignored.
Final part of rdar://10972766

llvm-svn: 152754
2012-03-14 23:32:11 +00:00
Roman Divacky 3ffe746ca8 Fix the long double to be of width/align 64. Rename va_list_test to
powerpc_types and add testing for the (long) double there.

llvm-svn: 152647
2012-03-13 19:20:17 +00:00
James Molloy a7139221f8 Use ZeroLengthBitfieldAlignment for AAPCS, as well as APCS-GNU.
llvm-svn: 152552
2012-03-12 09:14:10 +00:00
Daniel Dunbar f07b5ec0dc IRgen/ABI/x86_64: Avoid passing small structs using byval sometimes.
- We do this when it is easy to determine that the backend will pass them on
   the stack properly by itself.

Currently LLVM codegen is really bad in some cases with byval, for example, on
the test case here (which is derived from Sema code, which likes to pass
SourceLocations around)::

  struct s47 { unsigned a; };
  void f47(int,int,int,int,int,int,struct s47);
  void test47(int a, struct s47 b) { f47(a, a, a, a, a, a, b); }

we used to emit code like this::

  ...
  movl	%esi, -8(%rbp)
  movl	-8(%rbp), %ecx
  movl	%ecx, (%rsp)
  ...

to handle moving the struct onto the stack, which is just appalling.

Now we generate::

  movl	%esi, (%rsp)

which seems better, no?

llvm-svn: 152462
2012-03-10 01:03:58 +00:00
Chad Rosier 64d6be979f [driver] What was implemented in r152130 was actually -fno-inline-functions, not
-fno-inline.
Part of rdar://10972766

llvm-svn: 152145
2012-03-06 21:17:19 +00:00
Chad Rosier 9a099ea3eb Whitespace.
llvm-svn: 152134
2012-03-06 19:14:53 +00:00
Chad Rosier 806031896d [driver] Add support for -fno-inline.
rdar://10972766

llvm-svn: 152130
2012-03-06 18:49:20 +00:00
Chris Lattner 332673d39d From his comment in PR12168, Eli seems confused about the alignment assumptions
we're making.  We only ignore implicit casts.  Add a testcase showing that
we get the right behavior with explicit casts.

llvm-svn: 151994
2012-03-04 00:56:24 +00:00
Chris Lattner aaa18fad7d add a testcase for PR12094 and fix a crash on pointer to incomplete type,
reported by Richard Smith.

llvm-svn: 151993
2012-03-04 00:52:12 +00:00
Chad Rosier ff3f99bf0a I referenced the incorrect rdar in my previous commit (r151745). Add the
correct radar number to the test case for tracking purposes.
rdar://10551066

llvm-svn: 151746
2012-02-29 20:18:57 +00:00
Chad Rosier b1cfc6864f Allocate TargetLibraryInfo for the CodeGen passes. Otherwise, it's instantiated
by the BAA pass, which uses the default TargetLibraryInfo constructor.
Unfortunately, the default TargetLibraryInfo constructor assumes all library
calls are available and thus ignores -fno-builtin.
rdar://10947759

llvm-svn: 151745
2012-02-29 20:14:59 +00:00
James Molloy 6f8780bed1 Reapply r151638 and r151641.
The bug that was caught by Apple's internal buildbots was valid and also showed another bug in my implementation.

These are now fixed, with regression tests added to catch them both (not Darwin-specific).

Original log:
====================

Revert r151638 because it causes assertion hit on PCH creation for Cocoa.h

Original log:
---------------------
Correctly track tags and enum members defined in the prototype of a function, and ensure they are properly scoped.

This fixes code such as:

enum e {x, y};
int f(enum {y, x} n) {
 return 0;
}

This finally fixes PR5464 and PR5477.
---------------------

I also reverted r151641 which was enhancement on top of r151638.

====================

llvm-svn: 151712
2012-02-29 10:24:19 +00:00
Argyrios Kyrtzidis 5929ef2ee7 Revert r151638 because it causes assertion hit on PCH creation for Cocoa.h
Original log:
---------------------
Correctly track tags and enum members defined in the prototype of a function, and ensure they are properly scoped.

This fixes code such as:

enum e {x, y};
int f(enum {y, x} n) {
 return 0;
}

This finally fixes PR5464 and PR5477.
---------------------

I also reverted r151641 which was enhancement on top of r151638.

llvm-svn: 151667
2012-02-28 23:39:14 +00:00
James Molloy 051390fffa Correctly track tags and enum members defined in the prototype of a function, and ensure they are properly scoped.
This fixes code such as:

enum e {x, y};
int f(enum {y, x} n) {
  return 0;
}

This finally fixes PR5464 and PR5477.

llvm-svn: 151638
2012-02-28 18:12:11 +00:00
Eli Friedman afa84ae85b Add missing code for compound literals of complex type. <rdar://problem/10938628>
llvm-svn: 151549
2012-02-27 20:26:13 +00:00
Chad Rosier d6a56db7b6 Test case for r151429/r151430, which ensures llvm.lifetime intrinsics are not
being emitted at -O0, but are emitted when optimizations are enabled.

llvm-svn: 151533
2012-02-27 18:45:03 +00:00
Eli Friedman cb3785e450 Fix a stupid mistake in r151133. Reported to me by Joerg Sonnenberger.
llvm-svn: 151407
2012-02-24 23:53:49 +00:00
Aaron Ballman 0c84ebb539 Turned on support for __declspecs: noreturn, noinline, nothrow and naked in MS compatibility mode.
llvm-svn: 151295
2012-02-23 22:46:33 +00:00
Chandler Carruth a2a5410e6d Add 3dNOW intrinsic header to x86intrin.h, conditioned on __3dNOW__ to
match the behavior of GCC. Also add a test for these intrinsics, which
apparently have *zero* tests. =[ Not surprisingly, Clang crashed when
compiling these.

Fix the bug in CodeGen where we failed to bitcast the argument type to
x86mmx prior to calling the LLVM intrinsic. This fixes an assert on the
new 3dnow-builtins.c test.

This is one issue impacting the efforts to get Clang to emulate the
Microsoft intrinsics headers -- 3dnow intrinsics are implictitly made
available there.

llvm-svn: 150948
2012-02-20 07:35:45 +00:00
David Chisnall 0867d9cfbc Implement #pragma redefine_extname.
This fixes PR5172 and allows clang to compile C++ programs on Solaris using the system headers.

llvm-svn: 150881
2012-02-18 16:12:34 +00:00
Evgeniy Stepanov 0247b99171 Loosen the test from r150238 a bit to make some of our bots happy.
llvm-svn: 150242
2012-02-10 10:04:01 +00:00
Evgeniy Stepanov 3fae4aeae4 Fix function prolog codegen whe coerce-to type is a struct.
This changes function prolog in such a way as to avoid out-of-bounds
stack store in the case when coerce-to type has a larger storage size
than the real argument type.

Fixes PR11905.

llvm-svn: 150238
2012-02-10 09:30:15 +00:00
Akira Hatanaka 4984f5dbbe Class objects passed by value follow the same rules as structure objects.
Double fields of by-value class objects should be passed in floating point
registers.

llvm-svn: 150200
2012-02-09 19:54:16 +00:00
Akira Hatanaka b6f7443c3c Fix bugs in function MipsABIInfo::returnAggregateInRegs. Functions returning
class objects follow the same rules as those returning struct objects.

llvm-svn: 150196
2012-02-09 18:49:26 +00:00
Craig Topper e5ea3b0239 Remove vperm2f* and vperm2i builtins. Same effect can be achieved with builtin_shufflevector.
llvm-svn: 150064
2012-02-08 07:33:36 +00:00
Craig Topper fec9f8edb7 Remove vpermilp* builtins. Same effect can be achieved with builtin_shufflevector.
llvm-svn: 150056
2012-02-08 05:16:54 +00:00
Akira Hatanaka c07c4655f1 Do not return records with non trivial destructors or copy constructors in
registers.

llvm-svn: 150035
2012-02-08 01:31:22 +00:00
Eli Friedman 1b125c3e31 Make FunctionDecl::doesDeclarationForceExternallyVisibleDefinition use the same logic as FunctionDecl::isInlineDefinitionExternallyVisible to figure out whether to emit a definition. Based on work by Anton Yartsev.
llvm-svn: 149963
2012-02-07 03:50:18 +00:00
Bob Wilson 49708d41a6 Preserve alignment for Neon vld1_lane/dup and vst1_lane intrinsics.
We had been generating load/store instructions with the default alignment
for the vector element type, even when the pointer argument had less alignment.
<rdar://problem/10538555>

llvm-svn: 149794
2012-02-04 23:58:08 +00:00
Devang Patel fb04ece198 Update tests so that they don't rely upon LLVMDebugVersion number.
llvm-svn: 149726
2012-02-03 23:58:39 +00:00
Chad Rosier d1956e46dd [frontend] Don't allow a mapping to a warning override an error/fatal mapping.
rdar://10736625

llvm-svn: 149662
2012-02-03 01:49:51 +00:00
Anton Yartsev 79de4d49a6 r149587 reverted
llvm-svn: 149594
2012-02-02 06:06:34 +00:00
Anton Yartsev 3da212585e Fix for PR10657 (http://llvm.org/bugs/show_bug.cgi?id=10657)
extern inline case considered

llvm-svn: 149587
2012-02-02 05:13:59 +00:00
Craig Topper d6d3a05b4f Cleanup 3dnow builtin handling. Most of them were already handled by LLVM connecting intrinsics and builtins in IntrinsicsX86.td.
llvm-svn: 149233
2012-01-30 08:18:19 +00:00
Benjamin Kramer 1412816686 Make the __builtin_c[lt]zs builtins target independent.
There is really no reason to have these only available on x86. It's
just __builtin_c[tl]z for shorts.

Modernize the test while at it.

llvm-svn: 149183
2012-01-28 18:42:57 +00:00
Bob Wilson a7a61e2701 Make clz/ctz builtins defined for zero on ARM targets. rdar://10732455
ARM supports clz and ctz directly and both operations have well-defined
results for zero.  There is no disadvantage in performance to using the
defined-at-zero versions of llvm.ctlz/cttz intrinsics.  We're running into
ARM-specific code written with the assumption that __builtin_clz(0) == 32,
even though that value is technically undefined.  The code is failing now
because of llvm optimizations that are taking advantage of the undef
behavior (specifically svn r147255).  There's nothing wrong with that
optimization on x86 where any incorrect assumptions about __builtin_clz(0)
will quickly be exposed.  For ARM, though, optimizations based on that undef
behavior are likely to cause subtle bugs.  Other targets with defined-at-zero
clz/ctz support may want to override the default behavior as well.

llvm-svn: 149086
2012-01-26 22:14:27 +00:00
Douglas Gregor 35696d1a88 Try harder to get X7 defined
llvm-svn: 149072
2012-01-26 18:38:13 +00:00
Douglas Gregor e6e5406122 Force layout of more of the unions and structures in this test
llvm-svn: 149066
2012-01-26 17:23:00 +00:00
Chandler Carruth d0a41fead3 Suppress any warnings from this test. They aren't interesting, and they
end up in the same output file as the layout stuff. There may even be
a race condition which is causing this output to confuse the FileCheck
in some cases. I actually don't know how on earth the parsing of the
layout file even works given that there are diagnostics in the middle of
it. ;]

llvm-svn: 149058
2012-01-26 09:28:50 +00:00
Douglas Gregor e9fc377a44 Extend the ExternalASTSource interface to allow the AST source to
provide the layout of records, rather than letting Clang compute
the layout itself. LLDB provides the motivation for this feature:
because various layout-altering attributes (packed, aligned, etc.)
don't get reliably get placed into DWARF, the record layouts computed
by LLDB from the reconstructed records differ from the actual layouts,
and badness occurs. This interface lets the DWARF data drive layout,
so we don't need the attributes preserved to get the answer write.

The testing methodology for this change is fun. I've introduced a
variant of -fdump-record-layouts called -fdump-record-layouts-simple
that always has the simple C format and provides size/alignment/field
offsets. There is also a -cc1 option -foverride-record-layout=<file>
to take the output of -fdump-record-layouts-simple and parse it to
produce a set of overridden layouts, which is introduced into the AST
via a testing-only ExternalASTSource (called
LayoutOverrideSource). Each test contains a number of records to lay
out, which use various layout-changing attributes, and then dumps the
layouts. We then run the test again, using the preprocessor to
eliminate the layout-changing attributes entirely (which would give us
different layouts for the records), but supplying the
previously-computed record layouts. Finally, we diff the layouts
produced from the two runs to be sure that they are identical.

Note that this code makes the assumption that we don't *have* to
provide the offsets of bases or virtual bases to get the layout right,
because the alignment attributes don't affect it. I believe this
assumption holds, but if it does not, we can extend
LayoutOverrideSource to also provide base offset information.

Fixes the Clang side of <rdar://problem/10169539>.

llvm-svn: 149055
2012-01-26 07:55:45 +00:00
Eli Friedman a98d1f8ca6 Add some ABI tweaks for i386-pc-win32 triple so that we return structs in an MSVC-compatible way. Patch by Joe Groff.
llvm-svn: 148992
2012-01-25 22:46:34 +00:00
NAKAMURA Takumi 509b7af017 test/CodeGen/avx-builtins.c: Fix more for -Asserts.
llvm-svn: 148944
2012-01-25 09:11:21 +00:00
Craig Topper 80df922f2f Re-enable test that was broken by r148919
llvm-svn: 148932
2012-01-25 06:23:23 +00:00
Chris Lattner f07b612313 disable this test for now.
llvm-svn: 148928
2012-01-25 05:38:06 +00:00
Chris Lattner 29969cd86a fix broken testcase.
llvm-svn: 148925
2012-01-25 05:29:46 +00:00
Craig Topper 9e9301a83a 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 04:26:17 +00:00
Kostya Serebryany 588d6abf7d The following patch adds __attribute__((no_address_safety_analysis)) which will allow to disable
address safety analysis (such as e.g. AddressSanitizer or SAFECode) for a specific function.

When building with AddressSanitizer, add AddressSafety function attribute to every generated function
except for those that have __attribute__((no_address_safety_analysis)).

With this patch we will be able to
1. disable AddressSanitizer for a particular function
2. disable AddressSanitizer-hostile optimizations (such as some cases of load widening) when AddressSanitizer is on.

llvm-svn: 148842
2012-01-24 19:25:38 +00:00
Seth Cantrell a7956f92df add tests for wide character encodings
and fix typo

llvm-svn: 148603
2012-01-21 00:16:11 +00:00
Sebastian Pop 422377cfd3 rename -ccc-host-triple into -target
llvm-svn: 148582
2012-01-20 22:01:23 +00:00
Seth Cantrell 6255c257e9 Add and update tests for character literals
llvm-svn: 148392
2012-01-18 12:27:10 +00:00
Nick Lewycky e2367d2508 Add testcase for r148375!
llvm-svn: 148378
2012-01-18 04:28:31 +00:00
Tanya Lattner 81ce08c983 Add new line.
llvm-svn: 148255
2012-01-16 21:15:33 +00:00
Tanya Lattner 20248226a9 Add support for OpenCL 1.1 logical operations.
llvm-svn: 148254
2012-01-16 21:02:28 +00:00
David Chisnall fa35df628a Some improvements to the handling of C11 atomic types:
- Add atomic-to/from-nonatomic cast types
- Emit atomic operations for arithmetic on atomic types
- Emit non-atomic stores for initialisation of atomic types, but atomic stores and loads for every other store / load
- Add a __atomic_init() intrinsic which does a non-atomic store to an _Atomic() type.  This is needed for the corresponding C11 stdatomic.h function.
- Enables the relevant __has_feature() checks.  The feature isn't 100% complete yet, but it's done enough that we want people testing it.

Still to do:

- Make the arithmetic operations on atomic types (e.g. Atomic(int) foo = 1; foo++;) use the correct LLVM intrinsic if one exists, not a loop with a cmpxchg.
- Add a signal fence builtin
- Properly set the fenv state in atomic operations on floating point values
- Correctly handle things like _Atomic(_Complex double) which are too large for an atomic cmpxchg on some platforms (this requires working out what 'correctly' means in this context)
- Fix the many remaining corner cases

llvm-svn: 148242
2012-01-16 17:27:18 +00:00
Richard Smith dafff94759 constexpr irgen: Add irgen support for APValue::Struct, APValue::Union,
APValue::Array and APValue::MemberPointer. All APValue values can now be emitted
as constants.

Add new CGCXXABI entry point for emitting an APValue MemberPointer. The other
entrypoints dealing with constant member pointers are no longer necessary and
will be removed in a later change.

Switch codegen from using EvaluateAsRValue/EvaluateAsLValue to
VarDecl::evaluateValue. This performs caching and deals with the nasty cases in
C++11 where a non-const object's initializer can refer indirectly to
previously-initialized fields within the same object.

Building the intermediate APValue object incurs a measurable performance hit on
pathological testcases with huge initializer lists, so we continue to build IR
directly from the Expr nodes for array and record types outside of C++11.

llvm-svn: 148178
2012-01-14 04:30:29 +00:00
Eli Friedman d749c6bf2e Revert r148138; it's causing test failures.
llvm-svn: 148141
2012-01-13 21:33:06 +00:00
Sebastian Pop 9a8d528ddf rename -ccc-host-triple into -target
llvm-svn: 148138
2012-01-13 20:37:02 +00:00
Akira Hatanaka acf9896ca1 Fix test case committed in r147986.
llvm-svn: 147993
2012-01-12 01:27:24 +00:00
Akira Hatanaka 32604a9708 Take into account the pointer to an aggregate that is passed as a hidden
argument when Offset is initialized.

llvm-svn: 147986
2012-01-12 01:10:09 +00:00
NAKAMURA Takumi 68a75e879d clang/test/CodeGen/vla-4.c: Relax expression for -Asserts.
llvm-svn: 147934
2012-01-11 08:34:52 +00:00
Abramo Bagnara fa95ed4074 Fixed VLA code generation.
llvm-svn: 147933
2012-01-11 08:19:46 +00:00
Akira Hatanaka 428f5b2008 Add field IsIEEE in FloatingLiteral to distinguish between different 128-bit
floating point formats.

llvm-svn: 147887
2012-01-10 22:40:09 +00:00
Bill Wendling 86131f2c8d The `-fwritable-strings' flag doesn't make the backing store strings of all
CFStrings writable.

The strings (both Unicode and ASCII) should reside in a read-only section. E.g.,
__TEXT,__cstring instead of __DATA,__data. This is done by making the global
variable created for the strings constant despite the value of that flag.
<rdar://problem/10657500>

llvm-svn: 147845
2012-01-10 08:46:39 +00:00
Akira Hatanaka 1632af603d Insert padding before unaligned long double arguments.
llvm-svn: 147791
2012-01-09 19:31:25 +00:00
Abramo Bagnara 1cd8368eb2 Fixed TypeofExpr AST and code generation.
llvm-svn: 147730
2012-01-07 10:52:36 +00:00
Akira Hatanaka f64e1adcbb Add field PaddingType to ABIArgInfo which specifies the type of padding that
is inserted before the real argument. Padding is needed to ensure the backend
reads from or writes to the correct argument slots when the original alignment
of a byval structure is unavailable due to flattening.

llvm-svn: 147699
2012-01-07 00:25:33 +00:00
Eli Friedman df88c54f8d Revert r147655; it's breaking the compiler_rt build on OSX.
llvm-svn: 147677
2012-01-06 20:03:09 +00:00
David Chisnall 9217435a33 If we are compiling with -fno-builtin then don't do constant folding of
builtins.

This fixes PR11711.

llvm-svn: 147655
2012-01-06 12:20:19 +00:00
Chris Lattner 3ad9042456 fix test on -Asserts builds.
llvm-svn: 147558
2012-01-04 22:51:21 +00:00
Chris Lattner 36bc4f41bc implement rdar://10639962 by keeping track of increased alignment
information even in subscripting operations.

llvm-svn: 147557
2012-01-04 22:35:55 +00:00
Craig Topper b4ceb6fd52 Add FMA4 intrinsics.
llvm-svn: 147372
2011-12-30 09:15:03 +00:00
Craig Topper 1de8348db7 Add popcnt feature flag to match gcc. This flag is implied when sse42 is enabled, but can be disabled separately. Move popcnt intrinsics to popcntintrin.h to match gcc.
llvm-svn: 147340
2011-12-29 16:10:46 +00:00
Craig Topper c334dd68a7 Add BMI2 intrinsics.
llvm-svn: 147275
2011-12-26 02:31:10 +00:00
Craig Topper 03348d09e3 Change LZCNT and BMI tests to check IR instead of X86 assembly.
llvm-svn: 147267
2011-12-25 15:20:31 +00:00
Craig Topper a06d4a1c40 Add the rest of the BMI intrinsics.
llvm-svn: 147265
2011-12-25 07:27:12 +00:00
Craig Topper f2855ade2b Add intrinsics for lzcnt and tzcnt instructions.
llvm-svn: 147263
2011-12-25 06:25:37 +00:00
Craig Topper 175543ac78 Add last of the AVX2 intrinsics except for gather.
llvm-svn: 147253
2011-12-24 17:20:15 +00:00
Craig Topper a9387739aa Add AVX2 permute intrinsics. Also add parentheses on some macro arguments in other intrinsic headers.
llvm-svn: 147242
2011-12-24 07:55:25 +00:00
Craig Topper 9a01e58447 Add AVX2 intrinsics for FP vbroadcast, vbroadcasti128, and vpblendd.
llvm-svn: 147240
2011-12-24 05:19:47 +00:00
Craig Topper a6fdbd1807 Intrinsics for AVX2 unpack instructions.
llvm-svn: 147237
2011-12-24 03:58:43 +00:00
Craig Topper f4bb952533 More AVX2 intrinsics for shift, psign, some shuffles, and psadbw.
llvm-svn: 147236
2011-12-24 03:28:57 +00:00
Craig Topper 235a365d58 Add AVX2 multiply intrinsics.
llvm-svn: 147219
2011-12-23 08:31:16 +00:00
Craig Topper 1f2460ad43 Add AVX2 intrinsics for max, min, sign extend, and zero extend.
llvm-svn: 147141
2011-12-22 09:18:58 +00:00
Eli Friedman 803acb3ff2 Add support for bitcasts to vector type in Evaluate.
llvm-svn: 147137
2011-12-22 03:51:45 +00:00
Eli Friedman 1148b0f13c Fix a silly mistake in this test that somehow slipped into my last commit.
llvm-svn: 147112
2011-12-22 00:06:39 +00:00
Eli Friedman 5c9cd7a8a3 Fix a failure (which led to a crash) in constant emission code with vector compound literals.
llvm-svn: 147111
2011-12-22 00:04:00 +00:00