oota-llvm.git
10 years agoFactor out calls to AA.getDataLayout().
Rafael Espindola [Tue, 25 Feb 2014 15:52:19 +0000 (15:52 +0000)]
Factor out calls to AA.getDataLayout().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202157 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake a few more DataLayout variables const.
Rafael Espindola [Tue, 25 Feb 2014 14:24:11 +0000 (14:24 +0000)]
Make a few more DataLayout variables const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202155 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Use the original load name with the SROA-prefixed IRB rather than
Chandler Carruth [Tue, 25 Feb 2014 11:21:48 +0000 (11:21 +0000)]
[SROA] Use the original load name with the SROA-prefixed IRB rather than
just "load". This helps avoid pointless de-duping with order-sensitive
numbers as we already have unique names from the original load. It also
makes the resulting IR quite a bit easier to read.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202140 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Thread the ability to add a pointer-specific name prefix through
Chandler Carruth [Tue, 25 Feb 2014 11:19:56 +0000 (11:19 +0000)]
[SROA] Thread the ability to add a pointer-specific name prefix through
the pointer adjustment code. This is the primary code path that creates
totally new instructions in SROA and being able to lump them based on
the pointer value's name for which they were created causes
*significantly* fewer name collisions and general noise in the debug
output. This is particularly significant because it is making it much
harder to track down instability in the output of SROA, as name
de-duplication is a totally harmless form of instability that gets in
the way of seeing real problems.

The new fancy naming scheme tries to dig out the root "pre-SROA" name
for pointer values and associate that all the way through the pointer
formation instructions. Digging out the root is important to prevent the
multiple iterative rounds of SROA from just layering too much cruft on
top of cruft here. We already track the layers of SROAs iteration in the
alloca name prefix. We don't need to duplicate it here.

Should have no functionality change, and shouldn't have any really
measurable impact on NDEBUG builds, as most of the complex logic is
debug-only.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202139 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Rather than copying the logic for building a name prefix into the
Chandler Carruth [Tue, 25 Feb 2014 11:12:04 +0000 (11:12 +0000)]
[SROA] Rather than copying the logic for building a name prefix into the
PHI-pointer builder, just copy the builder and clobber the obvious
fields.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202136 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Simplify some of the logic to dig out the old pointer value by
Chandler Carruth [Tue, 25 Feb 2014 11:08:02 +0000 (11:08 +0000)]
[SROA] Simplify some of the logic to dig out the old pointer value by
using OldPtr more heavily. Lots of this code was written before the
rewriter had an OldPtr member setup ahead of time. There are already
asserts in place that should ensure this doesn't change any
functionality.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202135 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Adjust to new clang-format style.
Chandler Carruth [Tue, 25 Feb 2014 11:07:58 +0000 (11:07 +0000)]
[SROA] Adjust to new clang-format style.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202134 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReuse constants for COFF string table entry offsets
Nico Rieck [Tue, 25 Feb 2014 09:50:40 +0000 (09:50 +0000)]
Reuse constants for COFF string table entry offsets

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202130 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Fix a *glaring* bug in r202091: you have to actually *write*
Chandler Carruth [Tue, 25 Feb 2014 09:45:27 +0000 (09:45 +0000)]
[SROA] Fix a *glaring* bug in r202091: you have to actually *write*
the break statement, not just think it to yourself....

No idea how this worked at all, much less survived most bots, my
bootstrap, and some bot bootstraps!

The Polly one didn't survive, and this was filed as PR18959. I don't
have a reduced test case and honestly I'm not seeing the need. What we
probably need here are better asserts / debug-build behavior in
SmallPtrSet so that this madness doesn't make it so far.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202129 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDisable old JIT unittests for AARch64
Renato Golin [Tue, 25 Feb 2014 09:31:05 +0000 (09:31 +0000)]
Disable old JIT unittests for AARch64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202127 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIgnore old JIT tests in AARch64 - CMake style
Renato Golin [Tue, 25 Feb 2014 09:31:00 +0000 (09:31 +0000)]
Ignore old JIT tests in AARch64 - CMake style

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202126 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd aarch64 to config.guess
Renato Golin [Tue, 25 Feb 2014 09:30:54 +0000 (09:30 +0000)]
Add aarch64 to config.guess

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202125 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSilence GCC warning
Alexey Samsonov [Tue, 25 Feb 2014 07:56:00 +0000 (07:56 +0000)]
Silence GCC warning

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202119 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix typos
Alp Toker [Tue, 25 Feb 2014 04:21:15 +0000 (04:21 +0000)]
Fix typos

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202107 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Add a debugging tool which shuffles the slices sequence prior to
Chandler Carruth [Tue, 25 Feb 2014 03:59:29 +0000 (03:59 +0000)]
[SROA] Add a debugging tool which shuffles the slices sequence prior to
sorting it. This helps uncover latent reliance on the original ordering
which aren't guaranteed to be preserved by std::sort (but often are),
and which are based on the use-def chain orderings which also aren't
(technically) guaranteed.

Only available in C++11 debug builds, and behind a flag to prevent noise
at the moment, but this is generally useful so figured I'd put it in the
tree rather than keeping it out-of-tree.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202106 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Use a more direct way of determining whether we are processing
Chandler Carruth [Tue, 25 Feb 2014 03:50:14 +0000 (03:50 +0000)]
[SROA] Use a more direct way of determining whether we are processing
the destination operand or source operand of a memmove.

It so happens that it was impossible for SROA to try to rewrite
self-memmove where the operands are *identical*, because either such
a think is volatile (and we don't rewrite) or it is non-volatile, and we
don't even register it as a use of the alloca.

However, making the 'IsDest' test *rely* on this subtle fact is... Very
confusing for the reader. We should use the direct and readily available
test of the Use* which gives us concrete information about which operand
is being rewritten.

No functionality changed, I hope! ;]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202103 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd some convenience accessors for the underlying Use of an operand.
Chandler Carruth [Tue, 25 Feb 2014 03:34:17 +0000 (03:34 +0000)]
Add some convenience accessors for the underlying Use of an operand.
These complement many of the existing accessors and make it
significantly easier to write code which needs to poke at the underlying
Use without hard coding the operand number at which it resides for
a particular instruction. No functionality changed of course.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202102 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoIndent this continued line.
Nick Lewycky [Tue, 25 Feb 2014 00:43:21 +0000 (00:43 +0000)]
Indent this continued line.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202096 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[SROA] Fix another instability in SROA with respect to the slice
Chandler Carruth [Tue, 25 Feb 2014 00:07:09 +0000 (00:07 +0000)]
[SROA] Fix another instability in SROA with respect to the slice
ordering.

The fundamental problem that we're hitting here is that the use-def
chain ordering is *itself* not a stable thing to be relying on in the
rewriting for SROA. Further, we use a non-stable sort over the slices to
arrange them based on the section of the alloca they're operating on.
With a debugging STL implementation (or different implementations in
stage2 and stage3) this can cause stage2 != stage3.

The specific aspect of this problem fixed in this commit deals with the
rewriting and load-speculation around PHIs and Selects. This, like many
other aspects of the use-rewriting in SROA, is really part of the
"strong SSA-formation" that is doen by SROA where it works very hard to
canonicalize loads and stores in *just* the right way to satisfy the
needs of mem2reg[1]. When we have a select (or a PHI) with 2 uses of the
same alloca, we test that loads downstream of the select are
speculatable around it twice. If only one of the operands to the select
needs to be rewritten, then if we get lucky we rewrite that one first
and the select is immediately speculatable. This can cause the order of
operand visitation, and thus the order of slices to be rewritten, to
change an alloca from promotable to non-promotable and vice versa.

The fix is to defer all of the speculation until *after* the rewrite
phase is done. Once we've rewritten everything, we can accurately test
for whether speculation will work (once, instead of twice!) and the
order ceases to matter.

This also happens to simplify the other subtlety of speculation -- we
need to *not* speculate anything unless the result of speculating will
make the alloca fully promotable by mem2reg. I had a previous attempt at
simplifying this, but it was still pretty horrible.

There is actually already a *really* nice test case for this in
basictest.ll, but on multiple STL implementations and inputs, we just
got "lucky". Fortunately, the test case is very small and we can
essentially build it in exactly the opposite way to get reasonable
coverage in both directions even from normal STL implementations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202092 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm-dwarfdump: Support for debug_line.dwo section for file names for type units...
David Blaikie [Mon, 24 Feb 2014 23:58:54 +0000 (23:58 +0000)]
llvm-dwarfdump: Support for debug_line.dwo section for file names for type units under fission.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202091 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake some DataLayout pointers const.
Rafael Espindola [Mon, 24 Feb 2014 23:12:18 +0000 (23:12 +0000)]
Make some DataLayout pointers const.

No functionality change. Just reduces the noise of an upcoming patch.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202087 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPermit CMAKE_INSTALL_RPATH to be set on command line
Bernard Ogden [Mon, 24 Feb 2014 22:23:43 +0000 (22:23 +0000)]
Permit CMAKE_INSTALL_RPATH to be set on command line

Commit 201921 overrides setting of CMAKE_INSTALL_RPATH via the
command line. Last time this happened we applied another patch
to only set CMAKE_INSTALL_RPATH if already defined (r197825).
This patch does the same thing again, but only for the UNIX
case - we leave APPLE alone as presumably the original committer
is happy with the non-overriding behaviour.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202085 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agotrivial test commit
Albrecht Kadlec [Mon, 24 Feb 2014 22:18:38 +0000 (22:18 +0000)]
trivial test commit

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202084 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm-objdump: Do not attempt to disassemble symbols outside of section
Simon Atanasyan [Mon, 24 Feb 2014 22:12:11 +0000 (22:12 +0000)]
llvm-objdump: Do not attempt to disassemble symbols outside of section
boundaries.

It is possible to create an ELF executable where symbol from say .text
section 'points' to the address outside the section boundaries. It does
not have a sense to disassemble something outside the section.

Without this fix llvm-objdump prints finite or infinite (depends on
the executable file architecture) number of 'invalid instruction
encoding' warnings.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202083 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDisable an MCJIT test on older Darwins until we have a better interface.
Andrew Trick [Mon, 24 Feb 2014 21:37:30 +0000 (21:37 +0000)]
Disable an MCJIT test on older Darwins until we have a better interface.

See
<rdar://16149106> [MCJIT] provide a platform-independent way to communicate callee-save frame info.
<rdar://16149279> [MCJIT] get the host OS version from a runtime check, not a configure-time check.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202082 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix unused variable
Matt Arsenault [Mon, 24 Feb 2014 21:16:50 +0000 (21:16 +0000)]
Fix unused variable

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202080 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600/SI - Add new CI arithmetic instructions.
Matt Arsenault [Mon, 24 Feb 2014 21:01:28 +0000 (21:01 +0000)]
R600/SI - Add new CI arithmetic instructions.

Does not yet include larger part required
to match v_mad_i64_i32 / v_mad_u64_u32.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202077 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoR600: Make check clearer.
Matt Arsenault [Mon, 24 Feb 2014 21:01:23 +0000 (21:01 +0000)]
R600: Make check clearer.

The check is clearer as southern islands or later,
rather than checking for later than northern islands.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202076 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix DOT4 missing from getTargetOpcodeName
Matt Arsenault [Mon, 24 Feb 2014 21:01:21 +0000 (21:01 +0000)]
Fix DOT4 missing from getTargetOpcodeName

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202075 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd missing const
Matt Arsenault [Mon, 24 Feb 2014 21:01:18 +0000 (21:01 +0000)]
Add missing const

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202074 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTrivial code simplification
Matt Arsenault [Mon, 24 Feb 2014 21:01:15 +0000 (21:01 +0000)]
Trivial code simplification

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202073 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSLPVectorizer: Try vectorizing 'splat' stores
Arnold Schwaighofer [Mon, 24 Feb 2014 19:52:29 +0000 (19:52 +0000)]
SLPVectorizer: Try vectorizing 'splat' stores

Vectorize sequential stores of a broadcasted value.
5% on eon.

radar://16124699

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202067 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[X86][SchedModel] Add missing scheduling model for SSE related instructions.
Quentin Colombet [Mon, 24 Feb 2014 19:33:51 +0000 (19:33 +0000)]
[X86][SchedModel] Add missing scheduling model for SSE related instructions.

The patch defines new or refines existing generic scheduling classes to match
the behavior of the SSE instructions.
It also maps those scheduling classes on the related SSE instructions.

<rdar://problem/15607571>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202065 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd a dwarf number to the Y register.
Roman Divacky [Mon, 24 Feb 2014 18:41:31 +0000 (18:41 +0000)]
Add a dwarf number to the Y register.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202057 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoReplace the F_Binary flag with a F_Text one.
Rafael Espindola [Mon, 24 Feb 2014 18:20:12 +0000 (18:20 +0000)]
Replace the F_Binary flag with a F_Text one.

After this I will set the default back to F_None. The advantage is that
before this patch forgetting to set F_Binary would corrupt a file on windows.
Forgetting to set F_Text produces one that cannot be read in notepad, which
is a better failure mode :-)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202052 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoLTO: Add the loop vectorizer to the LTO pipeline.
Arnold Schwaighofer [Mon, 24 Feb 2014 18:19:31 +0000 (18:19 +0000)]
LTO: Add the loop vectorizer to the LTO pipeline.

During the LTO phase LICM will move loop invariant global variables out of loops
(informed by GlobalModRef). This makes more loops countable presenting
opportunity for the loop vectorizer.

Adding the loop vectorizer improves some TSVC benchmarks and twolf/ref dataset
(5%) on x86-64.

radar://15970632

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202051 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix windows unittest I missed in the raw_fd_ostream constructor change.
Rafael Espindola [Mon, 24 Feb 2014 16:40:34 +0000 (16:40 +0000)]
Fix windows unittest I missed in the raw_fd_ostream constructor change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202050 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFor lcov tests, don't Xfail mips littl endian (mipsel-... and mip64el-...)
Reed Kotler [Mon, 24 Feb 2014 16:33:56 +0000 (16:33 +0000)]
For lcov tests, don't Xfail mips littl endian (mipsel-... and mip64el-...)
targets. Just big endian (mips-... and mips64-...)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202049 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't make F_None the default.
Rafael Espindola [Mon, 24 Feb 2014 15:07:20 +0000 (15:07 +0000)]
Don't make F_None the default.

This will make it easier to switch the default to being binary files.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202042 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Remove dependency on non-existing profile_rt-shared. Patch by Brad King.
Alexey Samsonov [Mon, 24 Feb 2014 15:07:06 +0000 (15:07 +0000)]
[CMake] Remove dependency on non-existing profile_rt-shared. Patch by Brad King.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202041 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[asan] remove test that should have been removed in r202033
Kostya Serebryany [Mon, 24 Feb 2014 13:44:24 +0000 (13:44 +0000)]
[asan] remove test that should have been removed in r202033

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202034 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[asan] simplify the code that compute the shadow offset; get rid of two internal...
Kostya Serebryany [Mon, 24 Feb 2014 13:40:24 +0000 (13:40 +0000)]
[asan] simplify the code that compute the shadow offset; get rid of two internal flags that allowed to override it. The tests pass, but still this change might break asan on some platform not covered by tests. If you see this, please submit a fix with a test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202033 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd AArch64 big endian Target (aarch64_be)
Christian Pirker [Mon, 24 Feb 2014 11:34:50 +0000 (11:34 +0000)]
Add AArch64 big endian Target (aarch64_be)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202024 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAVX-512: Fixed encoding of VPCMPEQ and VPCMPGT
Elena Demikhovsky [Mon, 24 Feb 2014 10:08:30 +0000 (10:08 +0000)]
AVX-512: Fixed encoding of VPCMPEQ and VPCMPGT

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202015 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoShare a createUniqueEntity implementation between unix and windows.
Rafael Espindola [Mon, 24 Feb 2014 03:07:41 +0000 (03:07 +0000)]
Share a createUniqueEntity implementation between unix and windows.

The only extra bit of functionality that had to be exposed for this be be
implemented in Path.cpp is opening a file in rw mode.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202005 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDelete dead code.
Rafael Espindola [Mon, 24 Feb 2014 01:07:38 +0000 (01:07 +0000)]
Delete dead code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@202001 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAsm Parser: support .error directive
Saleem Abdulrasool [Sun, 23 Feb 2014 23:02:23 +0000 (23:02 +0000)]
Asm Parser: support .error directive

The .error directive is similar to .err in that it will halt assembly if it is
evaluated for assembly.  However, it permits a user supplied message to be
rendered.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201999 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAsmParser: support .ifeqs directive
Saleem Abdulrasool [Sun, 23 Feb 2014 23:02:18 +0000 (23:02 +0000)]
AsmParser: support .ifeqs directive

The .ifeqs directive assembles the following code if the quoted string
parameters are equal.  The strings must be quoted using double quotes.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201998 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSPARC: Implement TRAP lowering. Matches what GCC emits.
Benjamin Kramer [Sun, 23 Feb 2014 21:43:52 +0000 (21:43 +0000)]
SPARC: Implement TRAP lowering. Matches what GCC emits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201994 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARMAsmParser: whitespace
Saleem Abdulrasool [Sun, 23 Feb 2014 17:45:36 +0000 (17:45 +0000)]
ARMAsmParser: whitespace

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201989 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM IAS: support .align without parameters
Saleem Abdulrasool [Sun, 23 Feb 2014 17:45:32 +0000 (17:45 +0000)]
ARM IAS: support .align without parameters

.align is handled specially on certain targets.  .align without any parameters
on ARM indicates a default alignment (4).  Handle the special case in the target
parser, but fall back to the generic parser for the normal version.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201988 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMCAsmParser: support .ifne
Saleem Abdulrasool [Sun, 23 Feb 2014 15:53:41 +0000 (15:53 +0000)]
MCAsmParser: support .ifne

The .ifne directive assembles the following section of code if the argument
expression is non-zero.  Effectively, it is equivalent to if.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201986 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMCAsmParser: handle space properly for .ifc/.ifnc
Saleem Abdulrasool [Sun, 23 Feb 2014 15:53:36 +0000 (15:53 +0000)]
MCAsmParser: handle space properly for .ifc/.ifnc

If the strings are not quoted, the first string stops at the first comma, and
the second string stops at the end of the line.  Strings which contain
whitespace should be quoted.  Unquoted space is to be discarded.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201985 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMCAsmParser: add support for .err directive
Saleem Abdulrasool [Sun, 23 Feb 2014 15:53:30 +0000 (15:53 +0000)]
MCAsmParser: add support for .err directive

The .err directive produces an error whenever it is assembled.  This can be
useful for preventing assembly when an unexpected condition occurs.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201984 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse static instead of an anonymous namespace.
Rafael Espindola [Sun, 23 Feb 2014 15:16:03 +0000 (15:16 +0000)]
Use static instead of an anonymous namespace.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201983 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAVX-512: Fixed encoding of VPTESTMQ
Elena Demikhovsky [Sun, 23 Feb 2014 14:28:35 +0000 (14:28 +0000)]
AVX-512: Fixed encoding of VPTESTMQ

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201980 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSimplify remove, create_directory and create_directories.
Rafael Espindola [Sun, 23 Feb 2014 13:56:14 +0000 (13:56 +0000)]
Simplify remove, create_directory and create_directories.

Before this patch they would take an boolean argument to say if the path
already existed. This was redundant with the returned error_code which is able
to represent that. This allowed for callers to incorrectly check only the
existed flag instead of first checking the error code.

Instead, pass in a boolean flag to say if the previous (non-)existence should be
an error or not.

Callers of the of the old simple versions are not affected. They still ignore
the previous (non-)existence as they did before.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201979 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse simpler form of llvm::sys::fs::remove.
Rafael Espindola [Sun, 23 Feb 2014 13:37:37 +0000 (13:37 +0000)]
Use simpler form of llvm::sys::fs::remove.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201975 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoLocalStackSlotAllocation: Turn one-iteration loop into if.
Benjamin Kramer [Sun, 23 Feb 2014 13:34:21 +0000 (13:34 +0000)]
LocalStackSlotAllocation: Turn one-iteration loop into if.

No functionality change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201974 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTableGen.cmake: Functionalize and reformat.
NAKAMURA Takumi [Sun, 23 Feb 2014 12:54:15 +0000 (12:54 +0000)]
TableGen.cmake: Functionalize and reformat.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201972 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSimplify linking to system libraries
NAKAMURA Takumi [Sun, 23 Feb 2014 06:27:04 +0000 (06:27 +0000)]
Simplify linking to system libraries

The LLVMSupport library implementation consolidates all dependencies on
system libraries.  Move the logic gathering system libraries out of
'cmake/modules/LLVM-Config.cmake' and into 'lib/Support/CMakeLists.txt'.
Use the target_link_libraries() command there to tell CMake about the
link dependencies of the LLVMSupport implementation.  CMake will
automatically propagate this to all targets that link LLVMSupport
directly or indirectly.

We still need to build knowledge of system library dependencies into
'llvm-config'.  Store the list of libraries needed in a property on
LLVMSupport and teach 'tools/llvm-config/CMakeLists.txt' to retrieve it
from there.

Drop all calls to 'link_system_libs' and 'get_system_libs' from our
CMake code.  Replace their implementations with a warning that explains
the calls are no longer necessary.  Also drop from 'LLVMConfig.cmake'
the HAVE_* and related variables that were published there only to allow
'get_system_libs' to run outside our build process.

Contributed by Brad King.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201969 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoARM IAS: support .short and .hword
Saleem Abdulrasool [Sun, 23 Feb 2014 06:22:09 +0000 (06:22 +0000)]
ARM IAS: support .short and .hword

This adds support for the .short and its alias .hword for adding literal values
into the object file.  This is similar to the .word directive, however, rather
than inserting a value of 4 bytes, adds a 2-byte value.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201968 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[python-bindings] Remove some cruft that snuck in.
Michael Gottesman [Sun, 23 Feb 2014 04:43:26 +0000 (04:43 +0000)]
[python-bindings] Remove some cruft that snuck in.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201966 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMake test more resilient against scheduling decisions.
Benjamin Kramer [Sat, 22 Feb 2014 20:14:02 +0000 (20:14 +0000)]
Make test more resilient against scheduling decisions.

Should bring the atom buildbots back to life.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201951 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFix typo
Manman Ren [Sat, 22 Feb 2014 19:31:28 +0000 (19:31 +0000)]
Fix typo

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201944 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMC: Support COFF string tables larger than 10MB
Nico Rieck [Sat, 22 Feb 2014 16:12:20 +0000 (16:12 +0000)]
MC: Support COFF string tables larger than 10MB

Offsets past the range of single-slash encoding are encoded as base64,
padded to 6 characters, and prefixed with two slashes. This encoding is
undocumented but used by MSVC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201940 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDon't inline get[S|U]LEB128Size() until they are proved to be hot.
Logan Chien [Sat, 22 Feb 2014 15:39:39 +0000 (15:39 +0000)]
Don't inline get[S|U]LEB128Size() until they are proved to be hot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201939 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoMove get[S|U]LEB128Size() to LEB128.h.
Logan Chien [Sat, 22 Feb 2014 14:00:39 +0000 (14:00 +0000)]
Move get[S|U]LEB128Size() to LEB128.h.

This commit moves getSLEB128Size() and getULEB128Size() from
MCAsmInfo to LEB128.h and removes some copy-and-paste code.

Besides, this commit also adds some unit tests for the LEB128
functions.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201937 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert "Revert "Use relative rpath so that the installation and build dirs are reloca...
Rafael Espindola [Sat, 22 Feb 2014 13:29:31 +0000 (13:29 +0000)]
Revert "Revert "Use relative rpath so that the installation and build dirs are relocatable.""

This reverts commit r201934.

Polly has been fixed to work with this.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201936 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRevert "Use relative rpath so that the installation and build dirs are relocatable."
Rafael Espindola [Sat, 22 Feb 2014 12:36:28 +0000 (12:36 +0000)]
Revert "Use relative rpath so that the installation and build dirs are relocatable."

This reverts commit r201921.

This should bring the polly bots back. I will try to build it locally to
understand how cmake was setting the rpath of LLVMPolly.so.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201934 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd const to some member functions of SuccIterator.
Logan Chien [Sat, 22 Feb 2014 09:06:55 +0000 (09:06 +0000)]
Add const to some member functions of SuccIterator.

The operator+() and operator-() do not change the member
variables of SuccIterator.  This CL will qualify the *this*
pointer with const.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201933 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoPrune debug message in AddLLVM.cmake.
NAKAMURA Takumi [Sat, 22 Feb 2014 07:31:10 +0000 (07:31 +0000)]
Prune debug message in AddLLVM.cmake.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201932 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm/test/CodeGen/X86/shift-pcmp.ll: Tweak to appease FileCheck. "CHECK-LABEL" doesn...
NAKAMURA Takumi [Sat, 22 Feb 2014 07:27:04 +0000 (07:27 +0000)]
llvm/test/CodeGen/X86/shift-pcmp.ll: Tweak to appease FileCheck. "CHECK-LABEL" doesn't identify labels magically and CHECK-LABEL behaves free from other contexts.

For targeting pecoff, ".def foo" appears before ".short 32".

          .def    foo;
  ...
  .LCPI0_0:
          .short  32
  foo:

CHECK-LABEL seeks not from ".short 32" but from the top of the input.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201931 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoInclude <cctype> for isdigit().
Logan Chien [Sat, 22 Feb 2014 06:34:10 +0000 (06:34 +0000)]
Include <cctype> for isdigit().

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201930 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[python-bindings] Added OpCode like support for all enumerations with unittests.
Michael Gottesman [Sat, 22 Feb 2014 05:39:35 +0000 (05:39 +0000)]
[python-bindings] Added OpCode like support for all enumerations with unittests.

Also fixed some trailing whitespace issues.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201929 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CodeGenPrepare] Fix the check of the legality of an instruction.
Quentin Colombet [Sat, 22 Feb 2014 01:06:41 +0000 (01:06 +0000)]
[CodeGenPrepare] Fix the check of the legality of an instruction.
The API expects an ISD opcode, not an IR opcode.
Fixes a regression for R600.

Related to <rdar://problem/15519855>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201923 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoUse relative rpath so that the installation and build dirs are relocatable.
Rafael Espindola [Sat, 22 Feb 2014 00:54:14 +0000 (00:54 +0000)]
Use relative rpath so that the installation and build dirs are relocatable.

This works by asking cmake to use the "install rpath", but setting that rpath
to be relative.

Thanks a lot to Brad King for the help with CMake!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201921 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoSwitch the LLVM error reporting on the new diagnostic system.
Quentin Colombet [Sat, 22 Feb 2014 00:34:11 +0000 (00:34 +0000)]
Switch the LLVM error reporting on the new diagnostic system.
The LLVM diagnostic are now wired-up in clang (since r200931),
thus the user experience will not be impacted by this change
anymore.

Related to <rdar://problem/15886697>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201915 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CodeGenPrepare] Move CodeGenPrepare into lib/CodeGen.
Quentin Colombet [Sat, 22 Feb 2014 00:07:45 +0000 (00:07 +0000)]
[CodeGenPrepare] Move CodeGenPrepare into lib/CodeGen.

CodeGenPrepare uses extensively TargetLowering which is part of libLLVMCodeGen.
This is a layer violation which would introduce eventually a dependence on
CodeGen in ScalarOpts.

Move CodeGenPrepare into libLLVMCodeGen to avoid that.

Follow-up of <rdar://problem/15519855>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201912 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[DAGCombiner] PCMP* sets its result to all ones or zeros so we can AND with the
Quentin Colombet [Fri, 21 Feb 2014 23:42:41 +0000 (23:42 +0000)]
[DAGCombiner] PCMP* sets its result to all ones or zeros so we can AND with the
shifted mask rather than masking and shifting separately.

The patch adds this transformation to the DAGCombiner:

  (shl (and (setcc:i8v16 ...) N01C) N1C) -> (and (setcc:i8v16 ...) N01C<<N1C)

<rdar://problem/16054492>

Patch by Adam Nemet <anemet@apple.com>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201906 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[Stackmaps] Move the target-independent frame index elimination for stackmaps and...
Juergen Ributzka [Fri, 21 Feb 2014 23:29:32 +0000 (23:29 +0000)]
[Stackmaps] Move the target-independent frame index elimination for stackmaps and patchpoints into target-specific code.

The lowering of the frame index for stackmaps and patchpoints requires some
target-specific magic and should therefore be handled in the target-specific
eliminateFrameIndex method.

This is related to <rdar://problem/16106219>

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201904 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoShankar kindly pointed out that I wasn't following the coding convention properly...
Aaron Ballman [Fri, 21 Feb 2014 20:46:48 +0000 (20:46 +0000)]
Shankar kindly pointed out that I wasn't following the coding convention properly, so moving raw_ostream.h above system_error.h.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201885 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoFixing the MSVC build by including a file.
Aaron Ballman [Fri, 21 Feb 2014 20:42:18 +0000 (20:42 +0000)]
Fixing the MSVC build by including a file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201884 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agollvm-objdump/COFF: LoadConfiguration does not exist in object file.
Rui Ueyama [Fri, 21 Feb 2014 20:27:15 +0000 (20:27 +0000)]
llvm-objdump/COFF: LoadConfiguration does not exist in object file.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201883 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd missing include.
Rafael Espindola [Fri, 21 Feb 2014 20:21:55 +0000 (20:21 +0000)]
Add missing include.

Should fix the bots.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201882 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoAdd a SymbolicFile interface between Binary and ObjectFile.
Rafael Espindola [Fri, 21 Feb 2014 20:10:59 +0000 (20:10 +0000)]
Add a SymbolicFile interface between Binary and ObjectFile.

This interface allows IRObjectFile to be implemented without having dummy
methods for all section and segment related methods.

Both llvm-ar and llvm-nm are changed to use it. Unfortunately the mangler is
still not plugged in since it requires some refactoring to make a Module hold
a DataLayout.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201881 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoDebugInfo: Remove the empty macinfo section.
David Blaikie [Fri, 21 Feb 2014 19:13:09 +0000 (19:13 +0000)]
DebugInfo: Remove the empty macinfo section.

We were just emitting a label for this section for no real reason - this
caused us to emit the section even though we never put anything in it.

Not bothering with a test (though not adamantly anti-test) because it
seems somewhat arbitrary to test for the absence of this section anymore
than the absence of any other section.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201876 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRename a few more DataLayout variables from TD to DL.
Rafael Espindola [Fri, 21 Feb 2014 18:34:28 +0000 (18:34 +0000)]
Rename a few more DataLayout variables from TD to DL.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201870 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoadd -da-delinearize runs and checks to MIV testcases
Sebastian Pop [Fri, 21 Feb 2014 18:15:18 +0000 (18:15 +0000)]
add -da-delinearize runs and checks to MIV testcases

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201869 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agofix a corner case in delinearization
Sebastian Pop [Fri, 21 Feb 2014 18:15:15 +0000 (18:15 +0000)]
fix a corner case in delinearization

handle special cases Step==1, Step==-1, GCD==1, and GCD==-1

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201868 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agonormalize the last delinearized dimension
Sebastian Pop [Fri, 21 Feb 2014 18:15:11 +0000 (18:15 +0000)]
normalize the last delinearized dimension

in the dependence test, we used to discard some information that the
delinearization provides: the size of the innermost dimension of an array,
i.e., the size of scalars stored in the array, and the remainder of the
delinearization that provides the offset from which the array reads start,
i.e., the base address of the array.

To avoid losing this data in the rest of the data dependence analysis, the fix
is to multiply the access function in the last delinearized dimension by its
size, effectively making the size of the last dimension to always be in bytes,
and then add the remainder of delinearization to the last subscript,
effectively making the last subscript start at the base address of the array.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201867 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agofail delinearization when the size of subscripts differs
Sebastian Pop [Fri, 21 Feb 2014 18:15:07 +0000 (18:15 +0000)]
fail delinearization when the size of subscripts differs

Because the delinearization is not a global analysis pass, it will compute the
delinearization independently of knowledge about the way the delinearization
happened for other data accesses to the same array: the dependence analysis will
only trigger the delinearization on a tuple of access functions, and thus
delinearization may compute different subscripts sizes for a same array.  When
that happens the safest is to discard the delinearized information.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201866 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoRemove dead declaration
Sylvestre Ledru [Fri, 21 Feb 2014 17:58:12 +0000 (17:58 +0000)]
Remove dead declaration

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201863 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] llvm-c/lto.h should be installed if libLTO.a is built.
NAKAMURA Takumi [Fri, 21 Feb 2014 14:57:09 +0000 (14:57 +0000)]
[CMake] llvm-c/lto.h should be installed if libLTO.a is built.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201859 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] libLTO: Use (SHARED|STATIC) to build both shared LTO and LTO_static.
NAKAMURA Takumi [Fri, 21 Feb 2014 14:17:43 +0000 (14:17 +0000)]
[CMake] libLTO: Use (SHARED|STATIC) to build both shared LTO and LTO_static.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201856 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] llvm_add_library: Add dependencies also to objlibs as workaround of CMake...
NAKAMURA Takumi [Fri, 21 Feb 2014 14:17:29 +0000 (14:17 +0000)]
[CMake] llvm_add_library: Add dependencies also to objlibs as workaround of CMake issue 14747.

http://www.cmake.org/Bug/view.php?id=14747

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201855 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[CMake] Introduce "llvm_add_library(SHARED STATIC)" to build both shared lib and...
NAKAMURA Takumi [Fri, 21 Feb 2014 14:17:17 +0000 (14:17 +0000)]
[CMake] Introduce "llvm_add_library(SHARED STATIC)" to build both shared lib and static lib simulataneously.

  llvm_add_library(foo SHARED STATIC
    DEPENDS <dependent targets...>
    LINK_LIBS <required libraries...>
    )

It generates both foo (foo.so) and foo_static(foo.a) and both of them depend on DEPENDS and LINK_LIBS.
Then, also obj.foo is generated. obj.foo depends on DEPENDS, but doesn't depend on LINK_LIBS.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201854 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach LLVM-Config to use logical target names (2/2)
NAKAMURA Takumi [Fri, 21 Feb 2014 14:17:07 +0000 (14:17 +0000)]
Teach LLVM-Config to use logical target names (2/2)

The module still needs to collect the list of all available libraries
in order to satisfy the 'all' component.  Provide this in the package
configuration file, 'LLVMConfig.cmake', as a LLVM_AVAILABLE_LIBS
variable.  (A variable is scoped better than a global property.)
Since this won't be set for our own build, fall back to looking up the
LLVM_LIBS property to get the value when it is not set.

Contributed by Brad King.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201853 91177308-0d34-0410-b5e6-96231b3b80d8

10 years agoTeach LLVM-Config to use logical target names (1/2)
NAKAMURA Takumi [Fri, 21 Feb 2014 14:16:52 +0000 (14:16 +0000)]
Teach LLVM-Config to use logical target names (1/2)

LLVM library names are now available as logical CMake targets both
to our own build and to application CMake code.  Replace use of
'list(FIND)' with a simple 'if(TARGET)' to determine whether a
library is available.

Contributed by Brad King.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201852 91177308-0d34-0410-b5e6-96231b3b80d8

10 years ago[AArch64] Add register constraints to avoid generating STLXR and STXR with unpredicta...
Kevin Qin [Fri, 21 Feb 2014 07:45:48 +0000 (07:45 +0000)]
[AArch64] Add register constraints to avoid generating STLXR and STXR with unpredictable behavior.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201841 91177308-0d34-0410-b5e6-96231b3b80d8