Simon Pilgrim [Sat, 27 Feb 2016 12:33:08 +0000 (12:33 +0000)]
[X86][AVX] Added AVX1 target shuffle combine tests
llvm-svn: 262132
Simon Pilgrim [Sat, 27 Feb 2016 11:49:16 +0000 (11:49 +0000)]
Strip trailing whitespace. NFCI.
llvm-svn: 262131
Chandler Carruth [Sat, 27 Feb 2016 11:07:16 +0000 (11:07 +0000)]
[PM] Loosen the regex for the proxy template name even further to cope
with 'class' keywords in the template arguments and other silliness.
llvm-svn: 262130
Chandler Carruth [Sat, 27 Feb 2016 10:48:14 +0000 (10:48 +0000)]
[PM] Use a boring regex instead of explicitly naming the analysis
manager as some compilers print the typedef name and others print the
"canonical" name of the underlying class template.
This isn't really an important artifact of the test anyways so it seems
fine to just loosen the test assertions here.
llvm-svn: 262129
Chandler Carruth [Sat, 27 Feb 2016 10:45:35 +0000 (10:45 +0000)]
[PM] Provide explicit instantiation declarations and definitions for the
PassManager and AnalysisManager template specializations as well.
llvm-svn: 262128
Chandler Carruth [Sat, 27 Feb 2016 10:38:10 +0000 (10:38 +0000)]
[PM] Provide two templates for the two directionalities of analysis
manager proxies and use those rather than repeating their definition
four times.
There are real differences between the two directions: outer AMs are
const and don't need to have invalidation tracked. But every proxy in
a particular direction is identical except for the analysis manager type
and the IR unit they proxy into. This makes them prime candidates for
nice templates.
I've started introducing explicit template instantiation declarations
and definitions as well because we really shouldn't be emitting all this
everywhere. I'm going to go back and add the same for the other
templates like this in a follow-up patch.
I've left the analysis manager as an opaque type rather than using two
IR units and requiring it to be an AnalysisManager template
specialization. I think its important that users retain the ability to
provide their own custom analysis management layer and provided it has
the appropriate API everything should Just Work.
llvm-svn: 262127
Matt Arsenault [Sat, 27 Feb 2016 09:54:43 +0000 (09:54 +0000)]
AMDGPU: Add builtins for recently added intrinsics
llvm-svn: 262126
Matt Arsenault [Sat, 27 Feb 2016 09:38:05 +0000 (09:38 +0000)]
DAGCombiner: Relax sqrt NaN folding check
This is OK for +0 since compares to +/-0 give the same result.
llvm-svn: 262125
Matt Arsenault [Sat, 27 Feb 2016 09:06:26 +0000 (09:06 +0000)]
AMDGPU: Fix broken/confusing predefined macro
amdgcn should not be defining __R600__
llvm-svn: 262124
Matt Arsenault [Sat, 27 Feb 2016 09:06:22 +0000 (09:06 +0000)]
AMDGPU: Fix inconsistent register name for flat_scratch
llvm-svn: 262123
Matt Arsenault [Sat, 27 Feb 2016 09:06:18 +0000 (09:06 +0000)]
Add __builtin_canonicalize
llvm-svn: 262122
Matt Arsenault [Sat, 27 Feb 2016 08:53:55 +0000 (08:53 +0000)]
AMDGPU: Split vi-insts subtarget feature
This will be more useful for marking builtins acceptable for which
subtargets.
llvm-svn: 262121
Matt Arsenault [Sat, 27 Feb 2016 08:53:52 +0000 (08:53 +0000)]
AMDGPU: Add s_sleep intrinsic
llvm-svn: 262120
Matt Arsenault [Sat, 27 Feb 2016 08:53:46 +0000 (08:53 +0000)]
AMDGPU: Implement readcyclecounter
This matches the behavior of the HSAIL clock instruction.
s_realmemtime is used if the subtarget supports it, and falls
back to s_memtime if not.
Also introduces new intrinsics for each of s_memtime / s_memrealtime.
llvm-svn: 262119
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 07:00:35 +0000 (07:00 +0000)]
CodeGen: Avoid implicit conversion in MachineInstrBuilder, NFC
Avoid another implicit conversion from MachineInstrBundleIterator to
MachineInstr*, this time in MachineInstrBuilder.h (this is in pursuit of
PR26753).
llvm-svn: 262118
Tobias Grosser [Sat, 27 Feb 2016 06:59:30 +0000 (06:59 +0000)]
ScopInfo: Drop some debug statements
This debug output distracts from the -debug-only=polly-scops output. As it is
rather verbose and only really needed for debugging the domain construction
I drop this output. The domain construction is meanwhile stable enough to
not require regular debugging.
llvm-svn: 262117
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 06:51:00 +0000 (06:51 +0000)]
CodeGen: Remove implicit iterator to pointer conversions, NFC
Remove a couple of implicit conversions from MachineInstrBundleIterator
to MachineInstr*.
llvm-svn: 262116
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 06:40:41 +0000 (06:40 +0000)]
CodeGen: Take MachineInstr& in SlotIndexes and LiveIntervals, NFC
Take MachineInstr by reference instead of by pointer in SlotIndexes and
the SlotIndex wrappers in LiveIntervals. The MachineInstrs here are
never null, so this cleans up the API a bit. It also incidentally
removes a few implicit conversions from MachineInstrBundleIterator to
MachineInstr* (see PR26753).
At a couple of call sites it was convenient to convert to a range-based
for loop over MachineBasicBlock::instr_begin/instr_end, so I added
MachineBasicBlock::instrs.
llvm-svn: 262115
Tobias Grosser [Sat, 27 Feb 2016 06:24:58 +0000 (06:24 +0000)]
LoopGenerators: Expose some parts of the parallel loop generator
Some of this functionality is useful beyond the generation of a normal OpenMP
loop.
llvm-svn: 262114
Tobias Grosser [Sat, 27 Feb 2016 06:04:40 +0000 (06:04 +0000)]
ScopInfo: Add function to invalidate ScopArrayInfo object
In case the underlying basepointer of a ScopArrayInfo object is moved to another
module while the scop is still processed is it necessary to free dependent
ScopArrayInfo objects as they might otherwise be looked accidentally when a
new llvm basepointer value is reassigned the very same memory location as the
llvm value that has been moved earlier.
This function is not yet used in Polly itself, but is useful for external users.
llvm-svn: 262113
Sean Silva [Sat, 27 Feb 2016 06:01:26 +0000 (06:01 +0000)]
[instrprof] Use __{start,stop}_SECNAME on PS4 too.
Summary:
The PS4 linker seems to handle this fine.
Hi David, it seems that indeed most ELF linkers support
__{start,stop}_SECNAME, as our proprietary linker does as well.
This follows the pattern of r250679 w.r.t. the testing.
Maggie, Phillip, Paul: I've tested this with the PS4 SDK 3.5 toolchain
prerelease and it seems to work fine.
Reviewers: davidxl
Subscribers: probinson, phillip.power, MaggieYi
Differential Revision: http://reviews.llvm.org/D17672
llvm-svn: 262112
Mike Aizatsky [Sat, 27 Feb 2016 05:50:40 +0000 (05:50 +0000)]
[sancov] properly initializing pass.
llvm-svn: 262111
Kostya Serebryany [Sat, 27 Feb 2016 05:45:12 +0000 (05:45 +0000)]
[libFuzzer] don't emit callbacks to sanitizer run-time in -fsanitize-coverage=trace-pc mode; update libFuzzer doc for previous commit
llvm-svn: 262110
Philip Reames [Sat, 27 Feb 2016 05:18:30 +0000 (05:18 +0000)]
Suppress an uncovered switch warning [NFC]
llvm-svn: 262109
Chandler Carruth [Sat, 27 Feb 2016 04:34:07 +0000 (04:34 +0000)]
[LICM] Teach LICM how to handle cases where the alias set tracker was
merged into a loop that was subsequently unrolled (or otherwise nuked).
In this case it can't merge in the ASTs for any remaining nested loops,
it needs to re-add their instructions dircetly.
The fix is very isolated, but I've pulled the code for merging blocks
into the AST into a single place in the process. The only behavior
change is in the case which would have crashed before.
This fixes a crash reported by Mikael Holmen on the list after r261316
restored much of the loop pass pipelining and allowed us to actually do
this kind of nested transformation sequenc. I've taken that test case
and further reduced it into the somewhat twisty maze of loops in the
included test case. This does in fact trigger the bug even in this
reduced form.
llvm-svn: 262108
Kostya Serebryany [Sat, 27 Feb 2016 03:14:23 +0000 (03:14 +0000)]
[libFuzzer] fixing the bot
llvm-svn: 262106
Richard Smith [Sat, 27 Feb 2016 02:36:43 +0000 (02:36 +0000)]
Fix bug in using shadow decl checking: a using shadow decl should not conflict
with a prior UsingDecl -- those should not even really be found by the lookup
here, except that we use the same lookup results for two different checks, and
the other check needs them.
This happens to work in *almost all* cases, because either the lookup results
list the UsingDecl first (and the NonTag result gets replaced by something
else) or because the problematic declaration is a function (which causes us to
use different logic to detect conflicts). This can also be triggered from a
state only reachable through modules (where the name lookup results can contain
multiple UsingDecls in the same scope).
llvm-svn: 262105
Mike Aizatsky [Sat, 27 Feb 2016 02:21:44 +0000 (02:21 +0000)]
[sancov] print_coverage_points command.
Differential Revision: http://reviews.llvm.org/D17670
llvm-svn: 262104
Mike Aizatsky [Sat, 27 Feb 2016 02:10:27 +0000 (02:10 +0000)]
[sancov] do not instrument nodes that are full pre-dominators
Summary:
Without tree pruning clang has 2,667,552 points.
Wiht only dominators pruning: 1,515,586.
With both dominators & predominators pruning: 1,340,534.
Differential Revision: http://reviews.llvm.org/D17671
llvm-svn: 262103
Kostya Serebryany [Sat, 27 Feb 2016 01:50:16 +0000 (01:50 +0000)]
[libFuzzer] speedup path coverage handling
llvm-svn: 262102
Hongbin Zheng [Sat, 27 Feb 2016 01:50:01 +0000 (01:50 +0000)]
Update the fine-grain dependences analysis test case.
llvm-svn: 262101
Hongbin Zheng [Sat, 27 Feb 2016 01:49:58 +0000 (01:49 +0000)]
Enable llvm's isa/cast/dyn_cast on MemAccInst.
Differential Revision: http://reviews.llvm.org/D17250
llvm-svn: 262100
Junmo Park [Sat, 27 Feb 2016 01:10:43 +0000 (01:10 +0000)]
Minor code cleanup. NFC.
llvm-svn: 262096
Reid Kleckner [Sat, 27 Feb 2016 00:53:54 +0000 (00:53 +0000)]
[InstCombine] Be more conservative about removing stackrestore
We ended up removing a save/restore pair around an inalloca call,
leading to a miscompile in Chromium.
llvm-svn: 262095
Samuel Antao [Sat, 27 Feb 2016 00:01:58 +0000 (00:01 +0000)]
[OpenMP] Fix parsing of delete map clause modifier in C++ mode.
Summary: The map modifier 'delete' is parser in c++ mode as a delete keyword, which requires special handling in the map clause parsing.
Reviewers: hfinkel, carlo.bertolli, arpith-jacob, kkwli0, ABataev
Subscribers: cfe-commits, fraggamuffin, caomhin
Differential Revision: http://reviews.llvm.org/D17629
llvm-svn: 262094
Paul Robinson [Fri, 26 Feb 2016 23:44:10 +0000 (23:44 +0000)]
Revert r262092, caught LLD tests
llvm-svn: 262093
Paul Robinson [Fri, 26 Feb 2016 23:34:02 +0000 (23:34 +0000)]
[FileCheck] Abort if -NOT is combined with another suffix.
Combinations of suffixes that look useful actually are ignored;
complaining about them will avoid mistakes.
Differential Revision: http://reviews.llvm.org/D17587
llvm-svn: 262092
Cong Hou [Fri, 26 Feb 2016 23:25:30 +0000 (23:25 +0000)]
Fix a bug in isVectorReductionOp() in SelectionDAGBuilder.cpp that may cause assertion failure on AArch64.
llvm-svn: 262091
Greg Clayton [Fri, 26 Feb 2016 23:20:08 +0000 (23:20 +0000)]
Make LLDB safer to use with respect to the global destructor chain.
llvm-svn: 262090
Ahmed Bougacha [Fri, 26 Feb 2016 23:00:03 +0000 (23:00 +0000)]
[X86] Move an encoding test from CodeGen to MC. NFC.
llvm-svn: 262089
Ahmed Bougacha [Fri, 26 Feb 2016 23:00:00 +0000 (23:00 +0000)]
[X86] Delete old redundant test. NFC.
llvm-svn: 262088
Ahmed Bougacha [Fri, 26 Feb 2016 22:59:57 +0000 (22:59 +0000)]
[X86] Fix a stale comment. NFC.
llvm-svn: 262087
Ahmed Bougacha [Fri, 26 Feb 2016 22:59:41 +0000 (22:59 +0000)]
[X86] Remove the unused SDTX86atomicBinary. NFC.
llvm-svn: 262086
Philip Reames [Fri, 26 Feb 2016 22:53:59 +0000 (22:53 +0000)]
[LVI] Extend select handling to catch min/max/clamp idioms
Most of this is fairly straight forward. Add handling for min/max via existing matcher utility and ConstantRange routines. Add handling for clamp by exploiting condition constraints on inputs.
Note that I'm only handling two constant ranges at this point. It would be reasonable to consider treating overdefined as a full range if the instruction is typed as an integer, but that should be a separate change.
Differential Revision: http://reviews.llvm.org/D17184
llvm-svn: 262085
Kostya Serebryany [Fri, 26 Feb 2016 22:42:23 +0000 (22:42 +0000)]
[libFuzzer] add -print_final_stats=1 flag
llvm-svn: 262084
Simon Pilgrim [Fri, 26 Feb 2016 22:28:50 +0000 (22:28 +0000)]
Strip trailing whitespace. NFCI.
llvm-svn: 262083
Eugene Zelenko [Fri, 26 Feb 2016 22:26:21 +0000 (22:26 +0000)]
Fix Clang-tidy modernize-use-nullptr warnings in remaining files in source/Plugins/ABI; other minor fixes.
llvm-svn: 262082
Chaoren Lin [Fri, 26 Feb 2016 22:12:35 +0000 (22:12 +0000)]
Register value is not necessarily scalar.
Reviewers: aidan.dodds, mamai
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D17658
llvm-svn: 262081
Philip Reames [Fri, 26 Feb 2016 22:08:18 +0000 (22:08 +0000)]
[ConstantRange] Add umin/smin operators
This was split off from http://reviews.llvm.org/D17184.
Reviewed by: Sanjoy
llvm-svn: 262080
Kit Barton [Fri, 26 Feb 2016 21:59:44 +0000 (21:59 +0000)]
[PPC] Legalize FNEG on PPC when possible
Currently we always expand ISD::FNEG. For v4f32 and v2f64 vector types VSX has
native support for this opcode
Phabricator: http://reviews.llvm.org/D17647
llvm-svn: 262079
Simon Pilgrim [Fri, 26 Feb 2016 21:56:27 +0000 (21:56 +0000)]
Fix spelling. NFCI.
llvm-svn: 262078
Sanjay Patel [Fri, 26 Feb 2016 21:51:44 +0000 (21:51 +0000)]
[x86, InstCombine] transform x86 AVX2 masked stores to LLVM intrinsics
Replicate everything for integers...because x86.
Continuation of:
http://reviews.llvm.org/rL262064
llvm-svn: 262077
Rafael Espindola [Fri, 26 Feb 2016 21:49:38 +0000 (21:49 +0000)]
Report duplicated symbols in bitcode.
llvm-svn: 262076
Kostya Serebryany [Fri, 26 Feb 2016 21:33:56 +0000 (21:33 +0000)]
[libFuzzer] initial implementation of path coverage based on -fsanitize-coverage=trace-pc. This does not scale well yet, but already cracks FullCoverageSetTest in seconds
llvm-svn: 262073
Rafael Espindola [Fri, 26 Feb 2016 21:31:34 +0000 (21:31 +0000)]
Make SymbolBodies private.
This makes BitcodeFile a bit closer to ObjectFile.
If anyone prefers I can go the other way and delete getSymbols from
ObjectFile.
llvm-svn: 262072
Chris Bieneman [Fri, 26 Feb 2016 21:23:59 +0000 (21:23 +0000)]
[CMake] Only configure Native target in stage 1, configure all in other stages
This patch causes the 3-stage build pipeline to only build a host compiler in the first stage, and to build all targets for subsequent stages. The host target is determined via the Native target specifier added in r262070.
llvm-svn: 262071
Chris Bieneman [Fri, 26 Feb 2016 21:21:40 +0000 (21:21 +0000)]
[CMake] Allow LLVM_TARGETS_TO_BUILD to accept "Native"
This allows a user to specify "Native" as a target when configuring LLVM. Native will resolve to the LLVM_NATIVE_ARCH, which is the target that supports code generation for the host.
llvm-svn: 262070
Paul Robinson [Fri, 26 Feb 2016 21:18:34 +0000 (21:18 +0000)]
Reapply r262054 with triple fix.
llvm-svn: 262069
Kit Barton [Fri, 26 Feb 2016 21:11:55 +0000 (21:11 +0000)]
Power9] Implement new vsx instructions: compare and conversion
This change implements the following vsx instructions:
Quad/Double-Precision Compare:
xscmpoqp xscmpuqp
xscmpexpdp xscmpexpqp
xscmpeqdp xscmpgedp xscmpgtdp xscmpnedp
xvcmpnedp(.) xvcmpnesp(.)
Quad-Precision Floating-Point Conversion
xscvqpdp(o) xscvdpqp
xscvqpsdz xscvqpswz xscvqpudz xscvqpuwz xscvsdqp xscvudqp
xscvdphp xscvhpdp xvcvhpsp xvcvsphp
xsrqpi xsrqpix xsrqpxp
28 instructions
Phabricator: http://reviews.llvm.org/D16709
llvm-svn: 262068
Chris Bieneman [Fri, 26 Feb 2016 21:07:04 +0000 (21:07 +0000)]
[CMake] Add the gold plugin before clang
This is needed to connect dependencies between the LLVMgold plugin and the clang stage-2 builds due to limitations in ExternalProject_Add.
Patch by Mike Edwards
Differential Revision: http://reviews.llvm.org/D17655
llvm-svn: 262067
Chris Bieneman [Fri, 26 Feb 2016 21:07:02 +0000 (21:07 +0000)]
[CMake] Assigning the LTO component to lto.h
This makes it so lto.h is installed when you run the install-LTO target.
llvm-svn: 262066
Chris Bieneman [Fri, 26 Feb 2016 21:04:41 +0000 (21:04 +0000)]
[CMake] Add BINUTILS_INCDIR to the default passthrough list for multi-stage builds
This is needed to build the gold plugin in multi-stage builds.
Patch by Mike Edwards
Differential Revision: http://reviews.llvm.org/D17655
llvm-svn: 262065
Sanjay Patel [Fri, 26 Feb 2016 21:04:14 +0000 (21:04 +0000)]
[x86, InstCombine] transform x86 AVX masked stores to LLVM intrinsics
The intended effect of this patch in conjunction with:
http://reviews.llvm.org/rL259392
http://reviews.llvm.org/rL260145
is that customers using the AVX intrinsics in C will benefit from combines when
the store mask is constant:
void mstore_zero_mask(float *f, __m128 v) {
_mm_maskstore_ps(f, _mm_set1_epi32(0), v);
}
void mstore_fake_ones_mask(float *f, __m128 v) {
_mm_maskstore_ps(f, _mm_set1_epi32(1), v);
}
void mstore_ones_mask(float *f, __m128 v) {
_mm_maskstore_ps(f, _mm_set1_epi32(0x80000000), v);
}
void mstore_one_set_elt_mask(float *f, __m128 v) {
_mm_maskstore_ps(f, _mm_set_epi32(0x80000000, 0, 0, 0), v);
}
...so none of the above will actually generate a masked store for optimized code.
Differential Revision: http://reviews.llvm.org/D17485
llvm-svn: 262064
Alexey Samsonov [Fri, 26 Feb 2016 20:59:40 +0000 (20:59 +0000)]
[CMake] Fix a typo in add_compiler_rt_library.
llvm-svn: 262063
Sanjay Patel [Fri, 26 Feb 2016 20:59:05 +0000 (20:59 +0000)]
[x86] refactor to eliminate duplicated code; NFCI
llvm-svn: 262062
Amaury Sechet [Fri, 26 Feb 2016 20:49:07 +0000 (20:49 +0000)]
Fix warning in DwarfCFIException. NFC
llvm-svn: 262061
Paul Robinson [Fri, 26 Feb 2016 20:41:07 +0000 (20:41 +0000)]
Revert r262054 on one file that fails sometimes.
llvm-svn: 262060
Amaury Sechet [Fri, 26 Feb 2016 20:30:37 +0000 (20:30 +0000)]
Extract the method to begin and end a fragment in AsmPrinterHandler in their own method. NFC
Summary: This is extracted from D17555
Reviewers: davidxl, reames, sanjoy, MatzeB, pete
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D17580
llvm-svn: 262058
Quentin Colombet [Fri, 26 Feb 2016 20:05:02 +0000 (20:05 +0000)]
[GlobalISel] Fix a ranlib warning about empty TOC.
Fixes PR26733
llvm-svn: 262057
Reid Kleckner [Fri, 26 Feb 2016 19:51:02 +0000 (19:51 +0000)]
[dllexport] Sort out emission order of delayed exported classes
Relands r260194 with a fix. If we have a template that transitions from
an extern template to an explicitly instantiated dllexport template, we
would add that class to the delayed exported class list without flushing
it.
For explicit instantiations, we can just flush the list of delayed
classes immediately. We don't have to worry about the bug fixed in
r260194 in this case because explicit instantiations can only occur at
file and namespace scope.
Fixes PR26490.
llvm-svn: 262056
Greg Clayton [Fri, 26 Feb 2016 19:41:49 +0000 (19:41 +0000)]
Make sure to #include <atomic> for the std::atomic<bool> that was recently added.
llvm-svn: 262055
Paul Robinson [Fri, 26 Feb 2016 19:40:34 +0000 (19:40 +0000)]
Fix tests that used CHECK-NEXT-NOT and CHECK-DAG-NOT.
FileCheck actually doesn't support combo suffixes.
Differential Revision: http://reviews.llvm.org/D17588
llvm-svn: 262054
Greg Clayton [Fri, 26 Feb 2016 19:38:18 +0000 (19:38 +0000)]
Make sure the Target, Process and Thread GetGlobalProperties() static methods are thread safe.
<rdar://problem/
22595283>
llvm-svn: 262053
Paul Robinson [Fri, 26 Feb 2016 19:34:01 +0000 (19:34 +0000)]
Fix Clang tests that used CHECK-NEXT-NOT and CHECK-DAG-NOT.
FileCheck actually doesn't support combo suffixes.
Differential Revision: http://reviews.llvm.org/D17589
llvm-svn: 262052
Jim Ingham [Fri, 26 Feb 2016 19:33:11 +0000 (19:33 +0000)]
SymbolFileDWARFDebugMap::FindTypes didn't obey the max_matches flag,
but kept looking through .o files even after it had found as many
matches as were requested.
llvm-svn: 262051
Duncan P. N. Exon Smith [Fri, 26 Feb 2016 19:27:00 +0000 (19:27 +0000)]
SemaCXX: Support templates in availability attributes
If the availability context is `FunctionTemplateDecl`, we should look
through it to the `FunctionDecl`. This prevents a diagnostic in the
following case:
class C __attribute__((unavailable));
template <class T> void foo(C&) __attribute__((unavailable));
This adds tests for availability in templates in many other cases, but
that was the only case that failed before this patch.
I added a feature `__has_feature(attribute_availability_in_templates)`
so users can test for this.
rdar://problem/
24561029
llvm-svn: 262050
Nirav Dave [Fri, 26 Feb 2016 18:55:22 +0000 (18:55 +0000)]
Fix Sparc 32bit Lowering to rebundle up v2i32 values.
Summary: Fix LowerCall to rebundle v2i32 values after lowering and add testcase
Reviewers: jyknight
Subscribers: llvm-commits, jyknight
Differential Revision: http://reviews.llvm.org/D17615
llvm-svn: 262048
Sanjay Patel [Fri, 26 Feb 2016 18:42:50 +0000 (18:42 +0000)]
[x86, AVX] fold 'isPositive' 256-bit vector integer operations (PR26701)
This extends the fold introduced with:
http://reviews.llvm.org/rL262036
llvm-svn: 262047
Dmitry Vyukov [Fri, 26 Feb 2016 18:26:48 +0000 (18:26 +0000)]
tsan: revert r262037
Broke aarch64 and darwin bots.
llvm-svn: 262046
Reid Kleckner [Fri, 26 Feb 2016 18:08:59 +0000 (18:08 +0000)]
[IR] Optimize bitfield layout of Value for MSVC
This should save a pointer of padding from all MSVC Value subclasses.
Recall that MSVC will not pack the following bitfields together:
unsigned Bits : 29;
unsigned Flag1 : 1;
unsigned Flag2 : 1;
unsigned Flag3 : 1;
Add a static_assert because LLVM developers always trip over this
behavior. This regressed in June.
llvm-svn: 262045
Sanjay Patel [Fri, 26 Feb 2016 18:07:58 +0000 (18:07 +0000)]
[x86, AVX] add 256-bit tests
llvm-svn: 262044
Aidan Dodds [Fri, 26 Feb 2016 18:03:06 +0000 (18:03 +0000)]
remove unused local string in IRForTarget.cpp
Committed on behalf of: ldrumm <luke.drummond@codeplay.com>
Differential revision: http://reviews.llvm.org/D16412
llvm-svn: 262043
Sunil Srivastava [Fri, 26 Feb 2016 18:01:12 +0000 (18:01 +0000)]
Minor tweak to match the overall style.
llvm-svn: 262042
Aidan Dodds [Fri, 26 Feb 2016 17:40:50 +0000 (17:40 +0000)]
Fix bug with register values byte order in expression evaluation.
The evaluation of expressions containing register values was broken for targets for which endianness differs from host.
Committed on behalf of: mamai <marianne.mailhot.sarrasin@gmail.com>
Differential revision: http://reviews.llvm.org/D17167
llvm-svn: 262041
Greg Clayton [Fri, 26 Feb 2016 17:36:44 +0000 (17:36 +0000)]
The IOHandlerProcessSTDIO is the _only_ IOHandler that gets pushed and popped from functions that are run due to something that is NOT input from the user. All other IOHandler objects result from input from the user. An issue rose up where if a command caused the process to resume and stop and process state changed, where state changed Event objects were broadcast, it would cause the IOHandlerProcessSTDIO to have its IOHandlerProcessSTDIO::Cancel() function called. This used to always write a byte to the control pipe (IOHandlerProcessSTDIO::m_pipe) even if the IOHandlerProcessSTDIO::Run() was never called. What would happen is:
(lldb) command_that_steps_process_thousands_of_times
As the "command_that_steps_process_thousands_of_times" could be a python command that resumed the process thousands of times and in doing so the IOHandlerProcessSTDIO would get pushed when the process resumed, and popped when it stoppped, causing the call to IOHandlerProcessSTDIO::Cancel(). Since the IOHandler thread is currently in IOHandlerEditline::Run() for the command interpreter handling the "command_that_steps_process_thousands_of_times" command, IOHandlerProcessSTDIO::Run() would never get called, even though the IOHandlerProcessSTDIO is on the top of the stack. This caused the command pipe to keep getting 1 bytes written each time the IOHandlerProcessSTDIO::Cancel() was called and eventually we will deadlock since the write buffer is full.
The fix here is to make sure we are in IOHandlerProcessSTDIO::Run() before we write anything to the command pipe, and just call SetIsDone(true) if we are not.
<rdar://problem/
22361364>
llvm-svn: 262040
Hongbin Zheng [Fri, 26 Feb 2016 17:05:24 +0000 (17:05 +0000)]
Introduce fine-grain dependence analysis by tagging access functions and schedules tree with either the id of memory access or memory references.
Differential Revision: http://reviews.llvm.org/D17381
llvm-svn: 262039
Renato Golin [Fri, 26 Feb 2016 17:01:45 +0000 (17:01 +0000)]
[CMAKE] Update build on recent Haiku
This patch updates cmake build scripts to build on Haiku. It adds Haiku x86_64 to config.guess.
Please consider reviewing.
Pathc by Jérôme Duval.
llvm-svn: 262038
Dmitry Vyukov [Fri, 26 Feb 2016 16:57:14 +0000 (16:57 +0000)]
tsan: split thread into logical and physical state
Currently ThreadState holds both logical state (required for race-detection algorithm, user-visible)
and physical state (various caches, most notably malloc cache). Move physical state in a new
Process entity. Besides just being the right thing from abstraction point of view, this solves several
problems:
1. Cache everything on P level in Go. Currently we cache on a mix of goroutine and OS thread levels.
This unnecessary increases memory consumption.
2. Properly handle free operations in Go. Frees are issue by GC which don't have goroutine context.
As the result we could not do anything more than just clearing shadow. For example, we leaked
sync objects and heap block descriptors.
3. This will allow to get rid of libc malloc in Go (now we have Processor context for internal allocator cache).
This in turn will allow to get rid of dependency on libc entirely.
4. Potentially we can make Processor per-CPU in C++ mode instead of per-thread, which will
reduce resource consumption.
The distinction between Thread and Processor is currently used only by Go, C++ creates Processor per OS thread,
which is equivalent to the current scheme.
llvm-svn: 262037
Sanjay Patel [Fri, 26 Feb 2016 16:56:03 +0000 (16:56 +0000)]
[x86, SSE] fold 'isPositive' vector integer operations (PR26701)
This is one of the cases shown in:
https://llvm.org/bugs/show_bug.cgi?id=26701
Shift and negate is what InstCombine appears to prefer, so I've started with that pattern.
Note that the 'pcmpeq' instructions are always generating the negative one for the actual
'pcmpgt' comparison in each case (side note: why isn't there an alias mnemonic for that?).
Differential Revision: http://reviews.llvm.org/D17630
llvm-svn: 262036
Reid Kleckner [Fri, 26 Feb 2016 16:53:19 +0000 (16:53 +0000)]
[WinEH] Fix funclet return block clobber mask placement
MBB slot index intervals are half open, not closed. getMBBEndIndex()
returns the slot index of the start of the next block in layout order.
Placing a register mask there is incorrect if the successor of the
funclet return is not laid out after the return. Clang generates IR for
catch bodies before generating the following normal code, so we never
noticed this issue until the D frontend authors filed a bug about it.
Instead, we can put the clobber mask on the last instruction of the
funclet return block. We still aren't using a register mask operand on
the CATCHRET instruction because it would cause PEI to spill all CSRs,
including XMM regs, in the prologue.
Fixes PR26679.
llvm-svn: 262035
Rui Ueyama [Fri, 26 Feb 2016 16:49:54 +0000 (16:49 +0000)]
Fix broken buildbots.
llvm-svn: 262034
Tobias Grosser [Fri, 26 Feb 2016 16:43:35 +0000 (16:43 +0000)]
ScopDetect/Info: Add option to disable invariant load hoisting
This is helpful for test case reduction and other experiments.
llvm-svn: 262033
Michael Kruse [Fri, 26 Feb 2016 16:40:35 +0000 (16:40 +0000)]
ScopDetection: Fix mix-up of isLoad and isStore.
This was accidentally introduced in r258947.
Thanks to Hongbin Zheng for finding this.
Found-by: etherzhhb
llvm-svn: 262032
Rui Ueyama [Fri, 26 Feb 2016 16:38:39 +0000 (16:38 +0000)]
Simplify. NFC.
Regarding the comment, it is out of context because it describes
what it does not do there. It got too long because it was originally
two different comments that were simply merged together.
The semantics is described in fixAbsoluteSymbols, so we don't need it.
llvm-svn: 262031
Saleem Abdulrasool [Fri, 26 Feb 2016 16:34:01 +0000 (16:34 +0000)]
Basic: fix __USER_LABEL_PREFIX__ on Cygwin
Adjust the user label prefix for cygwin x86_64.
Resolves PR26744.
llvm-svn: 262030
Michael Kruse [Fri, 26 Feb 2016 16:08:24 +0000 (16:08 +0000)]
Reduce indention. NFC.
The functions buildAccessMultiDimFixed and buildAccessMultiDimParam were
refactored from buildMemoryAccess. In their own functions, the control
flow can be shortcut and simplified using returns.
Suggested-by: etherzhhb
llvm-svn: 262029
Tamas Berghammer [Fri, 26 Feb 2016 15:47:35 +0000 (15:47 +0000)]
Add new java plugin files to the xcode project
llvm-svn: 262028
Rui Ueyama [Fri, 26 Feb 2016 15:42:06 +0000 (15:42 +0000)]
Simplify. NFC.
llvm-svn: 262027
Rui Ueyama [Fri, 26 Feb 2016 15:39:26 +0000 (15:39 +0000)]
Add comment on AMDGPU that the difference has no obvious reason.
llvm-svn: 262026
Andy Gibbs [Fri, 26 Feb 2016 15:35:16 +0000 (15:35 +0000)]
Reduce false positives in printf/scanf format checker
Summary:
The printf/scanf format checker is a little over-zealous in handling the conditional operator. This patch reduces work by not checking code-paths that are never used and reduces false positives regarding uncovered arguments, for example in the code fragment:
printf(minimal ? "%i\n" : "%i: %s\n", code, msg);
Reviewers: rtrieu
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D15636
llvm-svn: 262025