symbols dynamically
The CoreFoundation symbols uses by __isOSVersionAtLeast should be loaded at
runtime to ensure that the programs that don't use @available won't have to be
linked to CoreFoundation.
The Clang frontend IRGen library will need to emit a CoreFoundation symbol
when @available is used to ensure that programs that actually use @available
are linked to CoreFoundation.
rdar://31039554
llvm-svn: 297760
This predicate compares the host's marketing OS version to one passed as
argument. Currently, only darwin targets are supported. This is done by parsing
the SystemVersion.plist file.
Also added in this patch is some lit testing infrastructure for builtins, which
previously had none. This part of the patch was written by Alex Lorenz (with
some minor modifications).
This patch is part of a feature I proposed here:
http://lists.llvm.org/pipermail/cfe-dev/2016-July/049851.html
Differential revision: https://reviews.llvm.org/D30136
llvm-svn: 297382
To avoid depending on kernel headers, we just repeat the single define
we need, which is likely never going to change.
Patch by Joakim Sindholt <opensource@zhasha.com>
llvm-svn: 295738
Summary:
This patch implements addsf3/__aeabi_fadd in asm for Thumb1.
Compared with generic C version (lib/fp_add_impl.inc), it
1. all constants are materialized instead of loading from constant pool
2. no stack spills (C version uses 136 bytes stack space)
3. clz() is called only when necessary. (C version always calls it)
Reviewers: compnerd, rengolin, asl
Reviewed By: asl
Subscribers: efriedma, aemerson, mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D29485
llvm-svn: 294172
Summary:
in aeabi_ldivmod and uldivmod, using r6 instead of r12 as the temp reg due to limitation of Thumb1 ISA.
Now, all EABI sources are Thumb1 compatible.
Also added test cases by reusing the test cases from divmodsi4_test.c, udivmodsi4_test and udivmoddi4_test.c
Reviewers: rengolin, compnerd
Reviewed By: rengolin
Subscribers: javed.absar, aemerson, mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D29226
llvm-svn: 293527
syntax fix
Summary:
Make the asm of aeabi_memset be assembled for thumb1.
Also fix some instructions to conform with the syntax of ARM reference manual.
For example, muls requires the form of "Rd, Rn, Rd" and orrs requires
the form of "Rd, Rm". Clang-as is benign but it may fail other assembler
if not in the exact form.
Reviewers: rengolin, compnerd, kubamracek
Reviewed By: rengolin, compnerd
Subscribers: aemerson, llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D28971
llvm-svn: 292727
Summary:
Setting -DCOMPILER_RT_TEST_TARGET_TRIPLE=armv6m-none-eabi will enable the build of builtin functions ARMv6m.
Currently, only those asms that support armv6m are added.
TODO:All asm sin ARM_EABI_Sources are ported for thumb1 so Thumb1_EABI_Sources will be deprecated.
Reviewers: rengolin, compnerd
Reviewed By: compnerd
Subscribers: aemerson, mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D28463
llvm-svn: 292504
The `-target` impacts the CC for the builtins. HF targets (with either
floating point ABI) always use AAPCS VFP for the builtins unless they
are AEABI builtins, in which case they use AAPCS. Non-HF targets (with
either floating point ABI) always use AAPCS for the builtins and AAPCS
for the AEABI builtins. This introduces the thunks necessary to switch
CC for the floating point operations. This is not currently enabled,
and should be dependent on the target being used to build compiler-rt.
However, as a stop-gap, a define can be added for ASFLAGS to get the
thunks.
llvm-svn: 291677
Implement the missing __floattitf() and __floatuntitf() functions, to
convert 128-bit (unsigned) integers to quad-precision floating-point
types. This is needed e.g. on AArch64 where 'long double' is
a quad-precision type.
The code is based on the existing code for __floattixf()
and __floatuntixf(), updated to account for different bit field lengths
of quad-precision float. The tests are also copied, with the rounding
tests adjusted for longer significand.
Differential Revision: https://reviews.llvm.org/D27898
llvm-svn: 291259
Summary:
For idivsi3, convert the Thumb2 only instruction to thumb1.
For aeabi_idivmod, using __divsi3.
Reviewers: rengolin, compnerd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D27472
llvm-svn: 288960
Summary: Since CLZ is not available for Thumb1, we use __ARM_ARCH_ISA_THUMB != 1 as one of the conditions.
Reviewers: rnk, compnerd, rengolin
Subscribers: aemerson, rengolin, llvm-commits
Differential Revision: https://reviews.llvm.org/D27530
llvm-svn: 288954
Summary:
The current uidiv supports archs without clz. However, the asm is for thumb2/arm.
For uidivmod, the existing code calls the C version of uidivmodsi4, which then calls uidiv. The extra push/pop/bl makes it less efficient.
Reviewers: jmolloy, jroelofs, joerg, compnerd, rengolin
Subscribers: llvm-commits, aemerson
Differential Revision: https://reviews.llvm.org/D27309
llvm-svn: 288710
The previous change for enabling MinGW did not preserve the Win32 check and
added the EABI specific routines to a Windows build which does not use the EABI
routines. Correct the conditional check for that.
llvm-svn: 288422
This fixes an incorrect standard usage of GNU99 when the compiler check was for
the ISO standard C99. Furthermore, bump the dependency up to C11. The
motivation for this change is ARM EHABI compatibility with clang 3.8. We rely
on a type definition redefinition which causes an error with -Werror builds.
This is problematic for FreeBSD builds. Switching to C11 allows the
compatibility without the unnecessary pedantic warning. The alternative would
be to clutter the support header with a `pragma clang diagnostic ignore`. GCC
4.8+ and the supported clang revisions along with MSVC support enough of C11 to
allow building the builtins in C11 mode. No functional change intended.
llvm-svn: 288099
When building with clang/LLVM in MSVC mode, the msvcrt libraries contain
these functions.
When building in a mingw environment, we need to provide them somehow,
e.g. via compiler-rt.
The aeabi divmod functions work in the same way as the corresponding
__rt_*div* functions for windows, but their parameters are swapped.
The functions for converting float to integer and vice versa are the
same as their aeabi equivalents, only with different function names.
Differential Revision: https://reviews.llvm.org/D26183
llvm-svn: 287465
When the C unwinding personality was corrected to match the ARM EHABI
specification, the unwind header in clang was updated with necessary
declarations. However, when building with an older compiler, we would not have
the necessary declarations. This would result in a build failure. Provide a
supplementary header to ensure that the necessary declarations are present for
the build of the C unwinding personality.
Note that this is NOT an ABI break. It merely is a compile time failure due to
the constants not being present. The constants here are reproduced
equivalently. This header should permit building with clang[<3.9] as well as
gcc.
Addresses PR31035!
llvm-svn: 287359
The CMake build system had missed this macro as part of the build of the
builtins. This would result in the builtins exporting symbols which are
implemented in assembly with global visibility. Ensure that the assembly
optimized routines are given the same visibility as the C routines.
llvm-svn: 285477
clear_cache is using R7 for the SVC call and that's the frame pointer in
GCC, which is only disabled on -O2/3, so Release builds finish, Debug don't.
Fixes PR30797.
llvm-svn: 285204
On Linux ARM, the syscall will take 3 arguments (start, end, flags). Ensure
that we do not pass garbage to the flags, which can cause the cacheflush call to
fail, and therefore cause an abort at runtime.
llvm-svn: 280877
The atomic builtin source is problematic when cross-compiling builtins because it requires a variable and sometimes large set of system headers. This option allows users to optionally prevent it from being built.
llvm-svn: 280416
Summary:
Since we can now build the builtins without a full toolchain these files should no longer be needed.
This is the last vestige of autoconf!
Reviewers: compnerd, iains, jroelofs
Subscribers: dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D23777
llvm-svn: 279539
Summary: This fixes the omission of -fPIC when building the builtins.
Reviewers: compnerd, beanz
Subscribers: dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D23729
llvm-svn: 279469
This fixes a long-standing TODO by implementing a compiler check for supporting the _Atomic keyword. If the _Atomic keyword is supported by the compiler we should include it in the builtin library sources.
llvm-svn: 278454
These are meant to only be included on certain targets. This only disables it
for Windows ARM for now. Ideally these would be conditionally included as
appropriate.
llvm-svn: 277777
We were getting warnings about how 'uint32_t*' is different from
'unsigned long*' even though they are effectively the same on Windows.
llvm-svn: 277363
This reverts commit r276333.
As I commented in the review (https://reviews.llvm.org/D22415), this change isn't needed because CMAKE_C_FLAGS is implicitly added by CMake to the command line for all C source files.
With this patch enabled CMAKE_C_FLAGS is duplicated on all C sources, and applied to ASM sources, which is not ideal.
I sent an email about this to llvm-commits on the commit thread. I suspect the problem the patch author was actually seeing is that CMAKE_C_FLAGS isn't applied to ASM files, and the builtins library has quite a few of those. The correct solution there is to specify CMAKE_ASM_FLAGS with whatever flags need to be passed to the compiler when compiling ASM files.
If there are other problems with flag propagation, please let me know.
llvm-svn: 276683
Summary:
This patch fixes cross-architecture compilation,
by allowing flags like -target and --sysroot to be set for
architecture testing and compilation.
Reviewers: tberghammer, srhines, danalbert, beanz, compnerd
Subscribers: tberghammer, llvm-commits, danalbert
Differential Revision: https://reviews.llvm.org/D22415
llvm-svn: 276333
Summary:
This aims to add support for __cpu_model and address Bug 25510. It uses
the code from lib/Support/Host.cpp for cpu detection, and creates
__cpu_model with that info.
Tested on OSX and built on Linux as well (though libgcc is the default).
The use of "asm" required -std=gnu99, hence the cmake change. Corrections
on better addressing this are welcome.
Previously reverted, up for review again to iron out outstanding issues.
Reviewers: llvm-commits, joerg, echristo, mehdi_amini
Subscribers: mehdi_amini
Differential Revision: http://reviews.llvm.org/D22181
llvm-svn: 275484
Use unified syntax for builtins/arm/aeabi_mem*.S.
This makes these files consistent with the others.
This fixes a problem on the linker, which can fail with the message
"relocation truncated to fit: R_ARM_THM_JUMP11 against symbol"
Patch by Kor Nielsen.
llvm-svn: 275264
Summary:
This patch is a refactoring of the way cmake 'targets' are grouped.
It won't affect non-UI cmake-generators.
Clang/LLVM are using a structured way to group targets which ease
navigation through Visual Studio UI. The Compiler-RT projects
differ from the way Clang/LLVM are grouping targets.
This patch doesn't contain behavior changes.
Reviewers: kubabrecka, rnk
Subscribers: wang0109, llvm-commits, kubabrecka, chrisha
Differential Revision: http://reviews.llvm.org/D21952
llvm-svn: 275111
Summary:
This aims to add support for __cpu_model and address Bug 25510. It uses
the code from lib/Support/Host.cpp for cpu detection, and creates
__cpu_model with that info.
Tested on OSX, it builts successfully, but the current version does
*not* resolve Bug 25510. The __cpu_model symbol is present in the
library but it only gets loaded with -all_load. This patch will not land
until this issue is clarified.
Built on Linux as well (though libgcc is the default). The use of "asm"
required -std=gnu99, hence the cmake change. Corrections on better
addressing this are welcome.
Note: See additional comments on D20988 (committed as r271921).
Reviewers: llvm-commits, joerg, echristo, mehdi_amini
Subscribers: mehdi_amini
Differential revision: http://reviews.llvm.org/D21033
llvm-svn: 274873
Summary:
This aims to add support for __cpu_model and address Bug 25510. It uses the code from lib/Support/Host.cpp for cpu detection, and creates __cpu_model with that info.
Tested on OSX, it builts successfully, but the current version does *not* resolve Bug 25510. The __cpu_model symbol is present in the library but it only gets loaded with -all_load. This patch will not land until this issue is clarified.
Built on Linux as well (though libgcc is the default). The use of "asm" required -std=gnu99, hence the cmake change. Corrections on better addressing this are welcome.
Note: See additional comments on D20988 (committed as r271921).
Reviewers: llvm-commits, joerg, echristo, mehdi_amini
Subscribers: mehdi_amini
Differential Revision: http://reviews.llvm.org/D21033
llvm-svn: 274865
These routines do not require executable stacks. However, by default ELFish
linkers may assume an executable stack on GNUish environments (and some non-GNU
ones too!). The GNU extension to add a note to indicate a non-executable stack
is honoured by these environments to mark the stack as non-executable (the
compiler normally emits this directive on appropriate targets whenever
possible). This allows normal builds from getting executable stacks due to
linking to the compiler rt builtins.
llvm-svn: 273500
Summary:
This patch adds support for building lib/builtins without a fully functioning toolchain. It allows you to bootstrap a cross-compiler, which previously couldn't be done with CMake.
This patch contains the following specific changes:
* Split builtin-specific code out of config-ix.cmake into builtin-config-ix.cmake
* Split some common CMake functionality needed by both builtins and sanitizers into base-config-ix.cmake
* Made lib/builtins/CMakeLists.txt able to be a top-level CMake configuration
I have tested this on Darwin targeting embedded Darwin, and on FreeBSD x86_64 targeting FreeBSD AArch64.
This patch depends on http://reviews.llvm.org/D19692, and is the last part of http://reviews.llvm.org/D16653.
Reviewers: samsonov, iains, jroelofs
Subscribers: compnerd, aemerson, tberghammer, danalbert, srhines, emaste, llvm-commits
Differential Revision: http://reviews.llvm.org/D19742
llvm-svn: 268977
Windows does not honour the __attribute__((pcs)) on ARM. Although this will
result in ABI mismatches, compiler-rt should largely be unneeded for resolving
dependencies as we generate MS ABI compliant library calls now for the most
part.
llvm-svn: 266891
Clang's StaticAnalyzer seems to (correctly) complain about code like:
T *p = calloc(sizeof(U), N);
...Where T and U are different types.
This patch removes some instances of this pattern from compiler-rt.
Patch by Apelete Seketeli.
Differential Revision: http://reviews.llvm.org/D19085
llvm-svn: 266388
This addresses PR27077. For some historical reason Darwin wasn't shipping multi3 in the compiler builtin library or in the OS builtin library. This caused building ffmpeg to fail because Polly was generating calls to multi3. It is easy enough to just add the builtin.
llvm-svn: 264750
__clear_cache on Android is identical to the version on Linux. Use __linux__
instead of __ANDROID__ as __linux__ is defined for Linux and Android.
llvm-svn: 263833
Until now the only exception APIs supported by gcc_personality_v0
are DWARF EH and SJLJ. This adds support for ARM EHABI as well.
This is achieved by
a) changing the function signature on ARM EHABI,
b) unwinding the stack before returning _URC_CONTINUE_UNWIND.
See "Exception Handling ABI for the ARM Architecture" for details
(http://infocenter.arm.com/help/topic/com.arm.doc.ihi0038b/IHI0038B_ehabi.pdf).
Patch by Timon Van Overveldt.
llvm-svn: 263010
Adjust the clobbers list. This use to work with older versions of gcc, but now
will error on newer versions (tested against 5.3) (as well as clang).
Patch by Tee Hao Wei!
llvm-svn: 261821
The type of size and align in struct __emutls_control must be
typedef unsigned int gcc_word __attribute__((mode(word)));
to match GCC. When gcc_word is larger than size_t, which is the case
for x32, the upper extra bits are all zeros. We can use variables of
size_t to operate on size and align.
Fix one trivial C99 warning about mixed declaration and code.
Differential Revision: http://reviews.llvm.org/D16886
llvm-svn: 259824
si_int is already defined in sysroot's siginfo.h
emutls.c includes pthread.h which includes asm/siginfo.h which
in turn includes asm-generic/siginfo.h and that defines si_int.
si_int is also defined in builtin's int_types.h and this leads to
errors. This patch fixes the issue by undefining the si_int in int_types.h
Differential Revision: http://reviews.llvm.org/D15086
llvm-svn: 254472
Summary:
The following tests for 128-bit floating-point type behaved in a strange way, thought it were bugs, but seem to be mistakes in tests:
* `fixtfsi` test checked for `0x80000001` as a value returned for number less than can be represented, while `LONG_MIN` should be returned on saturation;
* `fixunstfdi` wasn't enabled for AArch64, only for PPC, but there is nothing PPC specific in that test;
* `multf3` tried to underflow multiplication by producing result with 16383 exponent, while there are still 112 bits of fraction plus implicit bit, so resultant exponent should be 16497.
Tests for some other builtins didn't exist:
* `fixtfdi`
* `fixtfti`
* `fixunstfti`
They were made by copying similar files and adjusting for wider types and adding/removing some reasonable/extra checks.
Also `__fixuint` seems to have off by one error, updated tests to catch this case.
Reviewers: rengolin, zatrazz, howard.hinnant, t.p.northover, jmolloy, enefaim
Subscribers: aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D14187
llvm-svn: 252180
Add chkstk/alloca for gcc objects.
Replace or instructions with test, the latter should be marginally more
efficent, as it does not write to memory.
Differential Revision: http://reviews.llvm.org/D14044
Patch by vadimcn
llvm-svn: 251928
MSVC 2013 doesnt support C99 fully, including the hexidecimal floating point
representation. Use the expanded value to permit building with it.
Patch by Tee Hao Wei!
llvm-svn: 250365
The assembly implementations use GNU syntax which MSVC doesn't handle. Rather
than duplicate the code in a second syntax, use the C implementations.
Patch by Tee Hao Wei!
llvm-svn: 250360
Revert once more. This seems to fail to build on the buildbots which build with
ninja rather than MSBuild/Visual Studio. This requires further build
infrastructure changes to deal with the assembly routines.
llvm-svn: 250001
__inline is a vendor specific spelling for inline. clang and gcc treat it the
same as inline, and is available in MSVC 2013 which does not implement C99
(VS2015 supports the inline keyword though). This will allow us to build the
builtins using MSVC.
llvm-svn: 249953
cl does not support the same intrinsics as clang. Provide implementations for
the intrinsics using MSVC builtins.
Patch by Tee Hao Wei!
llvm-svn: 249515
Unfortunately, int_utils.h needs to depend on int_lib for the common macro
definitions. However, currently, int_utils.h is included by int_lib.h. Reorder
the inclusion to work around this.
llvm-svn: 249376
We need to make sure that if you change the builtin filter lists CMake re-generates its configurations so it includes the right builtins in the generated libraries.
llvm-svn: 248852
Summary: This ports functionality from the clang_macho_embedded.mk platform makefile over to CMake.
Reviewers: bogner, samsonov, bob.wilson
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D13226
llvm-svn: 248850
If the builtins are built with libgcc as the unwind provider on ARM, the exposed
_Unwind_SetIP is a macro. This results in the following warning due to
expansion of the argument:
warning: suggest parentheses around arithmetic in operand of ‘|’ [-Wparentheses]
Add a no-op set of parenthesis around the argument that will prevent this
warning.
llvm-svn: 248686
This adds a new name for ARM32 (armhf). We now force that the default build for
ARM32 (arm) to be soft-float ABI. This has a corresponding clang change to look
for these names based on the floating point ABI. The functions are built
identically (the functions are marked as AAPCS, so the calling convention does
not change, as per the RTABI specification), however, the object file contains
attributes indicating the build configuration which the linker will ensure are
not mix and matched. We now built the appropriate named archive so that we can
link properly.
llvm-svn: 248648
There are lingering issues building the atomic builtins with various versions of gcc. To unblock people we can only include them on Apple platforms where they are more tested.
llvm-svn: 248386
Summary:
Building the builtins on Darwin platforms is a bit complicated. This is a first-pass implementation of the functionality from clang_darwin.mk into CMake.
When building the builtins on Darwin we have layers of blacklists that we apply based on platform, architecture, and minimum supported OS version.
Reviewers: bogner, filcab, bob.wilson, samsonov
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D13059
llvm-svn: 248383
Summary:
I broke building the builtins with r245967. This fixes them on Linux and builds them properly for Darwin.
The old code could not be made to work on Darwin as a result of the refactoring of add_compiler_rt_runtime, so I had to rework the way they are built for Darwin. This solution is not ideal and will be fixed in subsequent commits. I just want to get this in so everything is working again.
Reviewers: samsonov, chh, compnerd, bogner
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D12500
llvm-svn: 246487
Summary: This refactoring moves much of the Apple-specific behavior into a function in AddCompilerRT. The next cleanup patch will remove more of the if(APPLE) checks in the outlying CMakeLists.
This patch adds a bunch of new functionality to add_compiler_rt_runtime so that the target names don't need to be reconstructed outside the call. It also updates some of the call sites to exercise the new functionality, but does not update all uses fully. Subsequent patches will further update call sites and move to using the new features.
Reviewers: filcab, bogner, kubabrecka, zaks.anna, glider, samsonov
Subscribers: beanz, rengolin, llvm-commits
Differential Revision: http://reviews.llvm.org/D12292
llvm-svn: 245970
MachO and COFF do not support aliases. Restrict the alias to ELF targets. This
should also fix the Darwin build. Make the FNALIAS usage an error on non-ELF
targets.
llvm-svn: 245669
Summary:
Currently CMake doesn't build builtins for AArch64 and if one does this anyway
it's likely that at least `__multc3`, `__floatditf` and `__floatunditf` will be
missing. There is actually more builtins to add, but these come from
different libc implementations, thus providing them makes compiler-rt for
AArch64 good enough at least for basic usage.
Builtins implementation were originally taken from FreeBSD project:
* [[ https://reviews.freebsd.org/D2173 | __multc3 ]]
* [[ https://reviews.freebsd.org/D2174 | __floatditf and __floatunditf ]]
Until they have been tested to find mistakes in `__float*` functions.
`__floatditf` was based on `__floatsitf`, which had the same mistakes
(fixed it in r243746).
Version of the builtins in this patch are fixed and complemented with basic
tests. Additionally they were tested via GCC's torture (this is what revealed
these issues).
P.S. Ed (author of FreeBSD patches) asked for feedback on the list some time ago (here [[ http://lists.cs.uiuc.edu/pipermail/llvmdev/2015-March/084064.html | here ]])
and got no response, but it seems to be worth adding these builtins as is and
extracting common part later.
Reviewers: howard.hinnant, t.p.northover, jmolloy, enefaim, rengolin, zatrazz
Subscribers: asl, emaste, samsonov, aemerson, llvm-commits, rengolin
Differential Revision: http://reviews.llvm.org/D11679
llvm-svn: 245296
Negative numbers were handled properly initially, but got broken
during addressing review, so none of them did actually work. Issues:
* Wrong negation.
* Wrong exponent calculation.
llvm-svn: 243746
This applies default compiler flags to .S files, in particular removing
the "-pedantic" option, which is desirable because there is nothing to
reasonably warn about; and the only thing that gcc warns about is that
you allegedly can't correctly invoke GLUE2 in lib/builtins/assembly.h
on platforms for which USER_LABEL_PREFIX is the empty string.
In the gcc bug https://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305 that
added the warning, a commenter notes that giving a macro of zero characters
to another macro is not precisely the same as failing to supply an argument,
and "there is a widespread belief in C++ community that such usage is valid".
Unfortunately the only way to silence the warning is to avoid -pedantic.
Differential Revision: http://reviews.llvm.org/D10713
llvm-svn: 243446
The is so that we can avoid using libgcc and use compiler-rt with
mingw-w64.
Related driver patch
http://reviews.llvm.org/D11077
I have tested this with mingw-w64 and everything seems to be in order.
I also sent this patch to the mingw-w64 mailing list for them to look at.
Patch by Martell Malone.
Differential Revision: http://reviews.llvm.org/D11085
llvm-svn: 242539
Summary:
This patch fixes incorrect truncation when the input wider value is
exactly 2^dstBits. For that value, the overflow to infinity is not
correctly handled. The fix is to replace a strict '>' with '>='.
Currently,
__truncdfsf2(340282366900000000000000000000000000000.0) returns infinity
__truncdfsf2(340282366920938463463374607431768211456.0) returns 0
__truncdfsf2(400000000000000000000000000000000000000.0) returns infinity
Likewise, __truncdfhf2 and __truncsfhf2 (and consequently gnu_f2h_ieee)
are discontinuous at 65536.0.
This patch adds tests for all three cases, along with adding a missing
header include to fp_test.h.
Reviewers: joerg, ab, srhines
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D10594
llvm-svn: 240450
Windows does not use AAPCS, but rather AAPCS-VFP, and thus the functions which
are assumed to be AAPCS will cause invalid argument setup. Ensure that the
functions are marked as AAPCS.
llvm-svn: 238056
Follow-up to r237161; seems like we can't use aliases, but we
can do better than duplicating the bodies, especially when that
body, after inlining, isn't as small as it looks.
Better approaches welcome. Perhaps the best thing is just to have
an #ifndef __APPLE__ over the GNUEABI names, since they're not used
there.
llvm-svn: 237323
Mostly uninteresting, except:
- in __extendXfYf2, when checking if the number is normal, the old
code relied on the unsignedness of src_rep_t, which is a problem
when sizeof(src_rep_t) < sizeof(int): the result gets promoted to
int, the signedness of which breaks the comparison.
I added an explicit cast; it shouldn't affect other types.
- we can't pass __fp16, so src_t and src_rep_t are the same.
- the gnu_*_ieee symbols are simply duplicated definitions, as aliases
are problematic on mach-o (where only weak aliases are supported;
that's not what we want).
Differential Revision: http://reviews.llvm.org/D9693
llvm-svn: 237161
Talking to John and Doug, we concluded that these functions from
stdatomic really do belong here in compiler-rt rather than in libc,
since the compiler owns stdatomic.h and these need to refer to
clang-specific builtins. Nonetheless, I've only added these on darwin
for now - other platforms should probably do the same unless their
libc does implement these functions.
llvm-svn: 236805
Summary:
The spec for these functions says that they should return 0 in this case but
this regressed in r234148. That revision essentially delegates the conversion
to the hardware, but that has different behavior on different platforms (e.g.
it is wrong on x86).
Also fix a typo in the name of __fixunsdfti
Test Plan: The existing unit tests now pass
Reviewers: joerg, howard.hinnant
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D9305
llvm-svn: 236319
This adds hard-float implementation for the following builtins:
* __fixdfdi()
* __fixsfdi()
* __fixunsdfdi()
* __fixunssfdi()
The soft-float implementation does never raise floating point
exceptions, which doesn't allow clients to detect floating point
conversion errors.
I must mention that I had to refer to libgcc's implementation to
write these functions.
Related unit-tests of compiler-rt passed with these changes.
Patch was somewhat out-dated, so was updated locally without any
functional changes.
Differential Revision: http://reviews.llvm.org/D5376
llvm-svn: 234148
80bit Intel/PPC long double is excluded due to lacking support
for the abstraction. Consistently provide saturation logic.
Extend to long double on 128bit IEEE extended platforms.
Initial patch with test cases from GuanHong Liu.
Reviewed by Steve Canon.
Differential Revision: http://reviews.llvm.org/D2804
llvm-svn: 231965
The syscall function itself is declared in <unistd.h>
Compiling compiler-rt with -Werror-implicit-function-declaration will
cause a build failure. This change fixes it.
Differential Revision: http://reviews.llvm.org/D7949
llvm-svn: 230838
This makes int_endianness.h work on operating systems for which we
haven't written explicit constructs, as long as GCC or Clang are being
used.
llvm-svn: 230208
Make sure clear_cache() builtin has an appropriate definition for Linux.
Call syscall(NR_cacheflush, ...).
Differential Revision: http://reviews.llvm.org/D7205
llvm-svn: 228767
Summary: This causes an invalid lvalue error when building for Android.
Reviewers: danalbert
Reviewed By: danalbert
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D7306
llvm-svn: 227907
Use synci implementation of clear_cache for short address ranges.
For long address ranges, make a kernel call.
Differential Revision: http://reviews.llvm.org/D6661
llvm-svn: 226567
Use unwind.h to get the declarations for unwinding interfaces. This header is
already provided by clang and gcc, so this adds no additional dependencies for
building the builtins library. It avoids the duplication which may drift over
time though.
llvm-svn: 225990
clang does not like the definition of builtins. In order to work around this,
we use a SUN CC to redefine the generated name. However, this requires that we
account for the user label prefix. Rather than hard coding that into the file,
rely on the compiler to tell us the information and use the preprocessor to
generate the name as we do in the assembly routines. NFC.
llvm-svn: 224597
Each of the object formats use a different directive for selecting the constant
section. Use a macro to avoid the duplication across a number of files. Also
correct a small macro mismatch on the Windows case (HIDDEN_DIRECTIVE -> HIDDEN).
Patch by Vadim Chugunov!
llvm-svn: 223910
The code in {add,sub}v.i3 routines does not trap when it should, because
it performs the actual add/subtract operation in signed arithmetic,
rather than unsigned.
Patch by Francois-Xavie Coudert!
llvm-svn: 221826
__ANDROID__ is a define that comes from the toolchain when building
for Android targets. ANDROID has a different meaning. ANDROID is
defined for _every_ Android build, including those done for host
modules. For host modules, we want to build the regular Linux
sanitizers and builtins, not the one for Android devices. This hasn't
been a problem until now because we only just started building the
sanitizers for the host.
llvm-svn: 220203
Clang r219557 introduces libcalls to complex multiply/divide functions.
Since these functions are not available in iOS for arm64 devices, add them to
the static libraries.
llvm-svn: 219715
This is simply to help clarity of the code. The functions are built as thumb
only if Thumb2 is available (__ARM_ARCH_ISA_THUMB == 2). Sink the selection
into the location of the definition and make DEFINE_COMPILERRT_THUMB_FUNCTION
always define a thumb function while DEFINE_COMPILERRT_FUNCTION always selects
the default.
Since the .thumb_func directive is always available (at least on Linux, Windows,
and BSD), sinking the macro right into the macro works just as well.
No functional change intended.
llvm-svn: 219182
The arm builtins converted into thumb in r213481 are not working
on darwin. On apple platforms, .thumb_func directive is required
to generated correct symbols for thumb functions.
<rdar://problem/18523605>
llvm-svn: 219040
This is needed so we can produce -i686- named libraries for
x86 Android (which is i686-linux-android).
An alternative solution would be keeping the "i386" name internally and
tweaking the OUTPUT_NAME of compiler-rt libraries.
llvm-svn: 218761
The ldrexd and strexd instructions are undefined for the ARMv7M
architecture, so we cannot use them to implement the
__sync_fetch_and_*_8 builtins. There is no other way to implement
these without OS support, so this patch #ifdef's these functions out
for M-class architectures.
There are no tests as I cannot find any existing tests for these
builtins.
I used the __ARM_ARCH_PROFILE predefine because __ARM_FEATURE_LDREX is
deprecated and not set by clang.
llvm-svn: 218601
__ARM_ARCH is part of the ACLE specification. At least clang and GCC have
supported this part of the ACLE for some time now. Let the compiler provide the
proper definition for the macro rather than try to guess it.
llvm-svn: 218095
Add the missing AEABI functions that are part of the base platform ABI
specification. The provided implementation does the bare minimum to avoid
requiring libc headers. This permits the use of compiler-rt on bare-metal
environments which conform to EABI.
llvm-svn: 217322
This macro did not do anything at this point, and is not particularly needed for
Windows unless building the builtins as a shared library. NFC.
llvm-svn: 217321
When building the builtins for a modern CPU (idiv support), __umodsi3 was
completely incorrect as it would behave as __udivmosi3, which takes a tertiary
parameter which is a pointer.
__udivsi3 was also incorrect, returning the remainder in r1. Although this
would not result in any crash or invalid behaviour as r1 is a caller saved
register in AAPCS, this is unnecessary. Simply perform the division ignoring
the remainder.
llvm-svn: 215295
The LLVM IAS seems to accept wide instructions for add and sub in ARM mode even
though it is not permitted. This uses a macro to ensure that the wide modifier
is only applied when building in THUMB mode.
This repairs building with GCC/GAS in ARM mode.
llvm-svn: 214046
The macro definitions are shared across multiple files. Define them once in the
assembly.h header rather than redefining it in each file.
llvm-svn: 214045
The architecture specific implementation of routines would be built and included
along with the generic implementation. This would result in multiple
definitions of those symbols.
The linker is free to select either of the two. Most of the time, this
shouldn't be too terrible as the forward iteration should catch the architecture
version due to the ordering. Rather than relying on the linker and build
infrastructure ordering things in a specific manner, only provide the
architecture version when available.
This reduces the size of compiler-rt, simplifies inspection of the library
implementations, and guarantees that the desired version is selected at a
slightly complex build system.
llvm-svn: 214040
Place the floating point constants into the read-only data section. This was
already being done for x86_64, this simply mirrors the behaviour for i686.
llvm-svn: 214034
MMX/SSE instructions expect 128-bit alignment (16-byte) for constants that they
reference. Correct the alignment on the constant values. Although it is quite
possible for the data to end up aligned, there is no guarantee that this will
occur unless it is explicitly aligned to the desired location. If the data ends
up being unaligned, the resultant binary would fault at runtime due to the
unaligned access.
As an example, the follow would fault previously:
cc -c lib/builtins/x86_64/floatundidf.S -o floatundidf.o
cc -c test/builtins/Unit/floatundidf_test.c -o floatundidf_test.c
ld -m elf_x86_64 floatundidf.o floatundidf_test.o -lc -o floatundidf
However, if the object files were reversed, the data would end up aligned and
the problem would go unnoticed.
llvm-svn: 214033
Re-apply SVN r213684 which was reverted in SVN r213724 since it broke the
build bots. Add a tweak to enable inclusion of the assembly sources in
standalone build as well.
Original commit message:
This patch address the PR20360. The CMake assembler build system
ignores the .S assembly files in builtins library build. This patch
fixes the issue.
llvm-svn: 214013
The .rodata directive was added on the IA-64 (Itanium) platform. The LLVM IAS
supports the .rodata on i386 and x86_64 as well. There is no reason to really
restrict compilation of the builtins to just clang. By explicitly indicating
that the data is meant to be pushed into the .rodata section via the .section
.rodata, the assembly is made compatible with clang and gcc (with GAS).
This will enable building these routines on the Linux buildbots via CMake.
llvm-svn: 214012
Convert the CBNZ backward branch instruction to CMP and BNE
avoiding illegal backwards branch and making the assembly code
in synh-ops.h to be UAL compliant.
Patch by: Sumanth Gundapaneni
llvm-svn: 213685
The CMake assembler build system ignores the .S assembly files in builtins
library build. This patch fixes the issue.
Patch by: Sumanth Gundapaneni
llvm-svn: 213684
For ARM cores that are ARMv6T2+ but not ARMv7ve or ARMv7-r and not an updated
ARMv7-a that has the idiv extension (chips with clz but not idiv), an incorrect
jump would be calculated due to the preference to thumb instructions over ARM.
Rather than computing the target at runtime, use a jumptable instead. This
trades a bit of storage for performance. The overhead is 32-bytes for each of
the three routines, but avoid the calculation of the offset.
Because clz was introduced in ARMv6T2 and idiv in certain versions of ARMv7,
the non-clz, non-idiv case implies a target which does not support Thumb-2, and
thus we cannot use Thumb on those targets (as it is unlikely that the assembly
will assemble).
Take the opportunity to refactor the IT block macros into assembly.h rather than
redefining them in the TUs where they are used.
Existing tests cover the full change already, so no new tests are added.
This effectively reverts SVN r213309.
llvm-svn: 213467
The udivmodsi4/modsi3/umodsi3 code computes jump targets based on ARM encodings
(if CLZ is present and IDIV is not present).
Reverts parts of r211032 and r211035.
llvm-svn: 213309
Without some mention of armv6m in a subdirectory of builtins, the make code
doesn't even know that armv6m exists and is something it should be looking for
in the platform-specific Makefiles. This means that none of the functions
listed actually get built and we end up with an almost entirely empty
libclang_rt.a for armv6m.
Unfortunately, the assembly code in the usual arm directory has no hope of
running on armv6m, which only supports Thumb-1 (not even ARM mode), so adding
it there won't work. Realistically, we probably *will* want to put any
optimised versions in a separate directory, so creating it now is harmless.
rdar://problem/17613576
llvm-svn: 212696
Storage Class 3 is static storage. These symbols need to be marked as external
(storage class 2) so that they can be referenced. Note that this external is
not the same as ELF "external" visibility, which is indicated by DLL Storage
Class (i.e. __declspec(dllexport) or __declspec(dllimport)).
llvm-svn: 211428
When possible, use Thumb or Thumb-2 over ARM instructions. This is particularly
important for pure-Thumb environments (e.g. Windows on ARM). Although, it is
possible to conditionalise this for that target specifically, this is available
on most newer ARM CPUs, and the code remains compatible with older CPUs with no
adverse effects. It therefore feels better to always prefer Thumb when
possible.
llvm-svn: 211032
Place constants into .rdata if targeting ELF or COFF/PE. This should be
functionally identical, however, the data would be placed into a different
section. This is purely a cleanup change.
llvm-svn: 209986
Make the whitespace a bit more uniform in the various assembly routines. This
also makes the assembly files a bit more uniform on the ARM side by explicitly
stating that it is using the unified syntax and that the contents of the code is
in the text section (or segment). No functional change.
llvm-svn: 209985
Call it "libclang_rt.builtins-<arch>.a" to be consistent
with sanitizers/profile libraries naming. Modify Makefile
and CMake build systems and Clang driver accordingly.
Fixes PR19822.
llvm-svn: 209473
Extend the function definition macros further to support COFF object emission.
The function definition in COFF includes the type and storage class in the
symbol definition context. This is needed to make the assembly routines
possible to be built for COFF environments (i.e. Windows).
llvm-svn: 209095
Rename the HIDDEN_DIRECTIVE macro to HIDDEN and give it a parameter providing
the name of the symbol to be given hidden visibility. This makes the macros
more amenable to COFF.
llvm-svn: 209094
Add (missing) definition of COMPILER_RT_EXPORT which is meant to be used for
decorating functions that are meant to be exported. This is useful for
platforms where exports and imports must be decorated explicitly (i.e. Windows).
llvm-svn: 208593
Use COMPILER_RT_EXPORT rather than COMPILER_RT_ABI for this function. Adding an
explicit PCS standard to the function causes a mismatch between the
declarations. Furthermore, the function is implemented in C, and should take
the CC based on the target triple.
llvm-svn: 208591
The .align statements in ARM assembly routines is actually meant to be a power
of 2 alignment (e.g. .align 2 == 4 byte alignment, not 2). Switch to using
.p2align. .p2align is guaranteed to be a power-of-two alignment always and much
more explicit.
The .align in the case of x86_64 is byte alignment, use .balign instead of
.align.
llvm-svn: 208578
r201909, which introduced CRT_HAS_128BIT, unintentionally broke self-hosting on
PPC32. We used to define CRT_HAS_128BIT only on LP64 systems, but this is not
quite right (at least for Clang-compiled code). Even though __int128 is not
supported on PPC32, SROA can (and does) still form i128 variables at the IR
level, and operations on those variables may turn into the associated runtime
calls. As a result, we still need to compile __ashlti3, __ashrti3, __lshrti3,
and perhaps others, on PPC32.
llvm-svn: 208560
Soon there will be an option to build compiler-rt parts as shared libraries
on Linux. Extracted from http://llvm-reviews.chandlerc.com/D3042
by Yuri Gribov.
llvm-svn: 205183
negative shift amounts and/or shifts wider than the type. VAX traps on
the former, X86 and other platforms produce incorrect results on the
latter.
llvm-svn: 204193
Since these are primarily useful for deeply embedded targets where code size is
very important, they are each in a separate file making use of infrastructure
in sync-ops.h. This allows a linker to include just the functions that are
actually used.
rdar://problem/14736665
llvm-svn: 202812
Alex Lorenz