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
Jeroen Ketema