platform/upstream/llvm.git
8 years agoAvoid use of -fsanitize=vptr in MSVC env, it is not yet supported
Reid Kleckner [Mon, 29 Feb 2016 20:38:35 +0000 (20:38 +0000)]
Avoid use of -fsanitize=vptr in MSVC env, it is not yet supported

llvm-svn: 262251

8 years agoEnable LoopLoadElimination by default
Adam Nemet [Mon, 29 Feb 2016 20:35:11 +0000 (20:35 +0000)]
Enable LoopLoadElimination by default

Summary:
I re-benchmarked this and results are similar to original results in
D13259:

On ARM64:
  SingleSource/Benchmarks/Polybench/linear-algebra/solvers/dynprog -59.27%
  SingleSource/Benchmarks/Polybench/stencils/adi                   -19.78%

On x86:
  SingleSource/Benchmarks/Polybench/linear-algebra/solvers/dynprog  -27.14%

And of course the original ~20% gain on SPECint_2006/456.hmmer with Loop
Distribution.

In terms of compile time, there is ~5% increase on both
SingleSource/Benchmarks/Misc/oourafft and
SingleSource/Benchmarks/Linkpack/linkpack-pc.  These are both very tiny
loop-intensive programs where SCEV computations dominates compile time.

The reason that time spent in SCEV increases has to do with the design
of the old pass manager.  If a transform pass does not preserve an
analysis we *invalidate* the analysis even if there was *no*
modification made by the transform pass.

This means that currently we don't take advantage of LLE and LV sharing
the same analysis (LAA) and unfortunately we recompute LAA *and* SCEV
for LLE.

(There should be a way to work around this limitation in the case of
SCEV and LAA since both compute things on demand and internally cache
their result.  Thus we could pretend that transform passes preserve
these analyses and manually invalidate them upon actual modification.
On the other hand the new pass manager is supposed to solve so I am not
sure if this is worthwhile.)

Reviewers: hfinkel, dberlin

Subscribers: dberlin, reames, mssimpso, aemerson, joker.eph, llvm-commits

Differential Revision: http://reviews.llvm.org/D16300

llvm-svn: 262250

8 years agoFixup MIPS testcase after r262247 and make it a little more robust.
Adrian Prantl [Mon, 29 Feb 2016 20:25:10 +0000 (20:25 +0000)]
Fixup MIPS testcase after r262247 and make it a little more robust.

llvm-svn: 262249

8 years ago[AArch64] Fix isLegalAddImmediate() to return true for valid negative values.
Geoff Berry [Mon, 29 Feb 2016 19:53:22 +0000 (19:53 +0000)]
[AArch64] Fix isLegalAddImmediate() to return true for valid negative values.

Reviewers: t.p.northover, jmolloy

Subscribers: mcrosier, aemerson, llvm-commits, rengolin

Differential Revision: http://reviews.llvm.org/D17463

llvm-svn: 262248

8 years agoFix PR26585 by improving the promotion of DBG_VALUEs to DW_AT_locations.
Adrian Prantl [Mon, 29 Feb 2016 19:49:46 +0000 (19:49 +0000)]
Fix PR26585 by improving the promotion of DBG_VALUEs to DW_AT_locations.
When a variable is described by a single DBG_VALUE instruction we can
often use a more efficient inline DW_AT_location instead of using a
location list.

This commit makes the heuristic that decides when to apply this
optimization stricter by also verifying that the DBG_VALUE is live at the
entry of the function (instead of just checking that it is valid until
the end of the function).

<rdar://problem/24611008>

llvm-svn: 262247

8 years agoFix Clang-tidy modernize-use-nullptr warnings in source/Plugins/Language; other minor...
Eugene Zelenko [Mon, 29 Feb 2016 19:41:30 +0000 (19:41 +0000)]
Fix Clang-tidy modernize-use-nullptr warnings in source/Plugins/Language; other minor fixes.

llvm-svn: 262246

8 years agoRename embedded bitcode section in MachO
Steven Wu [Mon, 29 Feb 2016 19:40:10 +0000 (19:40 +0000)]
Rename embedded bitcode section in MachO

Summary:
Rename the section embeds bitcode from ".llvmbc,.llvmbc" to "__LLVM,__bitcode".
The new name matches MachO section naming convention.

Reviewers: rafael, pcc

Subscribers: davide, llvm-commits, joker.eph

Differential Revision: http://reviews.llvm.org/D17388

llvm-svn: 262245

8 years ago[X86] Move the ATOMIC_LOAD_OP ISel from DAGToDAG to ISelLowering. NFCI.
Ahmed Bougacha [Mon, 29 Feb 2016 19:28:07 +0000 (19:28 +0000)]
[X86] Move the ATOMIC_LOAD_OP ISel from DAGToDAG to ISelLowering. NFCI.

This is long-standing dirtiness, as acknowledged by r77582:

    The current trick is to select it into a merge_values with
    the first definition being an implicit_def. The proper solution is
    to add new ISD opcodes for the no-output variant.

Doing this before selection will let us combine away some constructs.

Differential Revision: http://reviews.llvm.org/D17659

llvm-svn: 262244

8 years ago[Hexagon] Setting sign mismatch flag on expression instead of using bit tricks.
Colin LeMahieu [Mon, 29 Feb 2016 19:17:56 +0000 (19:17 +0000)]
[Hexagon] Setting sign mismatch flag on expression instead of using bit tricks.

llvm-svn: 262243

8 years agoMinor code cleanup. NFC
Rong Xu [Mon, 29 Feb 2016 19:16:04 +0000 (19:16 +0000)]
Minor code cleanup. NFC

llvm-svn: 262242

8 years ago[WinEH] Make setjmp work correctly with EH
David Majnemer [Mon, 29 Feb 2016 19:16:03 +0000 (19:16 +0000)]
[WinEH] Make setjmp work correctly with EH

32-bit X86 EH on Windows utilizes a stack of registration nodes
allocated and deallocated on entry/exit.  A registration node contains a
bunch of EH personality specific information like which try-state we are
currently in.

Because a setjmp target allows control flow from arbitrary program
points, there is no way to ensure that the try-state we are in is
correctly updated once we transfer control.

MSVC compatible compilers, like MSVC and ICC, utilize runtime helpers to
reinitialize the try-state when a longjmp occurs.  This is implemented
by adding additional arguments to _setjmp3: the desired try-state and
a helper routine to update the try-state.

Differential Revision: http://reviews.llvm.org/D17721

llvm-svn: 262241

8 years agoMove discriminator assignment to the right place.
Dehao Chen [Mon, 29 Feb 2016 18:59:48 +0000 (18:59 +0000)]
Move discriminator assignment to the right place.

Summary: Now discriminator is assigned per-function instead of per-module.

Reviewers: davidxl, dnovillo

Subscribers: dblaikie, llvm-commits

Differential Revision: http://reviews.llvm.org/D17664

llvm-svn: 262240

8 years ago[PGO] clang cc1 option change to enable IR level instrumentation
Rong Xu [Mon, 29 Feb 2016 18:54:59 +0000 (18:54 +0000)]
[PGO] clang cc1 option change to enable IR level instrumentation

This patch expands cc1 option -fprofile-instrument= with a new value: -fprofile-instrument=llvm
which enables IR level PGO instrumentation.

Reviewers: davidxl, silvas

Differential Revision: http://reviews.llvm.org/D17622

llvm-svn: 262239

8 years ago[Hexagon] Using MustExtend flag on expression instead of passing around bools.
Colin LeMahieu [Mon, 29 Feb 2016 18:39:51 +0000 (18:39 +0000)]
[Hexagon] Using MustExtend flag on expression instead of passing around bools.

llvm-svn: 262238

8 years agoFix LLD tests that used CHECK-NEXT-NOT. FileCheck does not support
Paul Robinson [Mon, 29 Feb 2016 18:05:21 +0000 (18:05 +0000)]
Fix LLD tests that used CHECK-NEXT-NOT. FileCheck does not support
combined suffixes.

Differential Revision: http://reviews.llvm.org/D17665

llvm-svn: 262237

8 years agofix typo in comment
Adrian Prantl [Mon, 29 Feb 2016 17:06:46 +0000 (17:06 +0000)]
fix typo in comment

llvm-svn: 262236

8 years ago[ScopDetection] Fix use-after-free.
Michael Kruse [Mon, 29 Feb 2016 16:54:18 +0000 (16:54 +0000)]
[ScopDetection] Fix use-after-free.

removeCachedResults deletes the DetectionContext from
DetectionContextMap such that any it cannot be used anymore.
Unfortunately invalid<ReportUnprofitable> and RejectLogs.insert still do
use it. Because the memory is part of a map and not returned to to the
OS immediatly, such that the observable effect was only a memory leak
due to reference counters not decreased when the second call to
removeCachedResults does not remove the DetectionContext because because
it already has been removed.

Fix by not removing the DetectionContext prematurely. The second call to
removeCachedResults will handle it anyway.

llvm-svn: 262235

8 years agoRevert "Implement new interfaces for code-formatting when applying replacements."
Manuel Klimek [Mon, 29 Feb 2016 16:44:16 +0000 (16:44 +0000)]
Revert "Implement new interfaces for code-formatting when applying replacements."

This reverts commit r262232.

llvm-svn: 262234

8 years agoFix for PR26180
Nemanja Ivanovic [Mon, 29 Feb 2016 16:42:27 +0000 (16:42 +0000)]
Fix for PR26180

Corresponds to Phabricator review:
http://reviews.llvm.org/D16592

This fix includes both an update to how we handle the "generic" CPU on LE
systems as well as Anton's fix for the Fast Isel issue.

llvm-svn: 262233

8 years agoImplement new interfaces for code-formatting when applying replacements.
Manuel Klimek [Mon, 29 Feb 2016 16:27:41 +0000 (16:27 +0000)]
Implement new interfaces for code-formatting when applying replacements.

Patch by Eric Liu.

llvm-svn: 262232

8 years ago[mips] Range check uimm20 and fixed a bug this revealed.
Daniel Sanders [Mon, 29 Feb 2016 16:06:38 +0000 (16:06 +0000)]
[mips] Range check uimm20 and fixed a bug this revealed.

Summary:
The bug was that dextu's operand 3 would print 0-31 instead of 32-63 when
printing assembly. This came up when replacing
MipsInstPrinter::printUnsignedImm() with a version that could handle arbitrary
bit widths.

MipsAsmPrinter::printUnsignedImm*() don't seem to be used so they have been
removed.

Reviewers: vkalintiris

Subscribers: dsanders, llvm-commits

Differential Revision: http://reviews.llvm.org/D15521

llvm-svn: 262231

8 years ago[mips] Do not use SLL for ANY_EXTEND nodes as the high bits are undefined.
Vasileios Kalintiris [Mon, 29 Feb 2016 15:58:12 +0000 (15:58 +0000)]
[mips] Do not use SLL for ANY_EXTEND nodes as the high bits are undefined.

Reviewers: dsanders

Subscribers: dsanders, llvm-commits

Differential Revision: http://reviews.llvm.org/D15420

llvm-svn: 262230

8 years ago[mips] Make isel select the correct DEXT variant up front.
Daniel Sanders [Mon, 29 Feb 2016 15:26:54 +0000 (15:26 +0000)]
[mips] Make isel select the correct DEXT variant up front.

Summary:
Previously, it would always select DEXT and substitute any invalid matches
for DEXTU/DEXTM during MipsMCCodeEmitter::encodeInstruction(). This works
but causes problems when adding range checked immediates to IAS.

Now isel selects the correct variant up front.

Reviewers: vkalintiris

Subscribers: dsanders, llvm-commits

Differential Revision: http://reviews.llvm.org/D16810

llvm-svn: 262229

8 years agoAMDGPU: Add missing Volcanic Islands targets
Tom Stellard [Mon, 29 Feb 2016 15:08:56 +0000 (15:08 +0000)]
AMDGPU: Add missing Volcanic Islands targets

Reviewers: arsenm

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D17645

llvm-svn: 262228

8 years agoReapply "Add update_test.py script."
Michael Kruse [Mon, 29 Feb 2016 14:58:13 +0000 (14:58 +0000)]
Reapply "Add update_test.py script."

Originally committed in r261899 and reverted in r262202 due to failing
in out-of-LLVM tree builds.

Replace the use of LLVM_TOOLS_BINARY_DIR by LLVM_TOOLS_DIR which exists
in both, in-tree and out-of-tree builds.

Original commit message:
The script updates a lit test case that uses FileCheck using the actual
output of the 'RUN:'-lines program. Useful when updating test cases due
to expected output changes and diff'ing expected and actual output.

llvm-svn: 262227

8 years agolto: don't fetch members for weak undef.
Rafael Espindola [Mon, 29 Feb 2016 14:29:48 +0000 (14:29 +0000)]
lto: don't fetch members for weak undef.

llvm-svn: 262225

8 years agoRevert a part of "Add/Improve complex, vector, aggregate types handling for SysV...
Pavel Labath [Mon, 29 Feb 2016 14:26:45 +0000 (14:26 +0000)]
Revert a part of "Add/Improve complex, vector, aggregate types handling for SysV ARM (hard/soft) ABI."

This partially reverts commit r262218.

The commit added additional checks to a test case. The test case is too big so it's not feasible
to XFAIL it completely. Suggest to implement the checks as a separate test case, which can then
be XFAILed more surgically.

llvm-svn: 262223

8 years agoIRObject: Mark extern_weak as weak.
Rafael Espindola [Mon, 29 Feb 2016 14:26:06 +0000 (14:26 +0000)]
IRObject: Mark extern_weak as weak.

llvm-svn: 262222

8 years agoRemove bogus assert.
Rafael Espindola [Mon, 29 Feb 2016 13:46:39 +0000 (13:46 +0000)]
Remove bogus assert.

llvm-svn: 262219

8 years agoAdd/Improve complex, vector, aggregate types handling for SysV ARM (hard/soft) ABI.
Omair Javaid [Mon, 29 Feb 2016 13:39:20 +0000 (13:39 +0000)]
Add/Improve complex, vector, aggregate types handling for SysV ARM (hard/soft) ABI.

For details see:

Differential revision: http://reviews.llvm.org/D17708

llvm-svn: 262218

8 years agoclang-format: Don't format unrelated nested blocks.
Daniel Jasper [Mon, 29 Feb 2016 12:26:20 +0000 (12:26 +0000)]
clang-format: Don't format unrelated nested blocks.

With this change:

  SomeFunction(
      [] {
int i;
 return i;  // Format this line.
      },
      [] {
 return 2;  // Don't "fix" this.
      });

llvm-svn: 262216

8 years ago[InstSimplify] Restore fsub 0.0, (fsub 0.0, X) ==> X optzn
Benjamin Kramer [Mon, 29 Feb 2016 12:18:25 +0000 (12:18 +0000)]
[InstSimplify] Restore fsub 0.0, (fsub 0.0, X) ==> X optzn

I accidentally removed this in r262212 but there was no test coverage to
detect it.

llvm-svn: 262215

8 years agoFix compiler warnings in the java code
Pavel Labath [Mon, 29 Feb 2016 11:44:15 +0000 (11:44 +0000)]
Fix compiler warnings in the java code

llvm-svn: 262214

8 years ago[mips] Make symbols an acceptable branch target when expanding compare-to-immediate...
Daniel Sanders [Mon, 29 Feb 2016 11:24:49 +0000 (11:24 +0000)]
[mips] Make symbols an acceptable branch target when expanding compare-to-immediate-and-branch macros.

Reviewers: vkalintiris

Subscribers: llvm-commits, vkalintiris, dim, seanbruno, dsanders

Differential Revision: http://reviews.llvm.org/D15369

llvm-svn: 262213

8 years ago[InstSimplify] fsub 0.0, (fsub -0.0, X) ==> X is only safe if signed zeros are ignored.
Benjamin Kramer [Mon, 29 Feb 2016 11:12:23 +0000 (11:12 +0000)]
[InstSimplify] fsub 0.0, (fsub -0.0, X) ==> X is only safe if signed zeros are ignored.

Only allow fsub -0.0, (fsub -0.0, X) ==> X without nsz. PR26746.

llvm-svn: 262212

8 years ago[test-release.sh] Add lldb to list of projects (disabled by default)
Daniel Sanders [Mon, 29 Feb 2016 11:04:39 +0000 (11:04 +0000)]
[test-release.sh] Add lldb to list of projects (disabled by default)

Reviewers: hans

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D17070

llvm-svn: 262211

8 years ago[asan] Fix UB in test/asan/TestCases/Linux/recvfrom.cc testcase.
Maxim Ostapenko [Mon, 29 Feb 2016 08:56:26 +0000 (08:56 +0000)]
[asan] Fix UB in test/asan/TestCases/Linux/recvfrom.cc testcase.

llvm-svn: 262210

8 years agoUpdate the link of issue
Mohit K. Bhakkad [Mon, 29 Feb 2016 08:22:55 +0000 (08:22 +0000)]
Update the link of issue

llvm-svn: 262209

8 years ago[index] Print and test module import references.
Argyrios Kyrtzidis [Mon, 29 Feb 2016 07:56:07 +0000 (07:56 +0000)]
[index] Print and test module import references.

llvm-svn: 262208

8 years ago[index] Add a caller relation for a call reference.
Argyrios Kyrtzidis [Mon, 29 Feb 2016 07:56:00 +0000 (07:56 +0000)]
[index] Add a caller relation for a call reference.

llvm-svn: 262207

8 years ago[AST/RecursiveASTVisitor] Correction so that dataTraverseStmtPost will get called...
Argyrios Kyrtzidis [Mon, 29 Feb 2016 07:55:55 +0000 (07:55 +0000)]
[AST/RecursiveASTVisitor] Correction so that dataTraverseStmtPost will get called after the statement has been visited.

Fixes the indexing client of this.

llvm-svn: 262206

8 years ago[index] Use ',' to separate symbol roles when printing.
Argyrios Kyrtzidis [Mon, 29 Feb 2016 07:55:51 +0000 (07:55 +0000)]
[index] Use ',' to separate symbol roles when printing.

llvm-svn: 262205

8 years ago[asan] Re-enable test/asan/TestCases/Linux/recvfrom.cc testcase.
Maxim Ostapenko [Mon, 29 Feb 2016 07:47:35 +0000 (07:47 +0000)]
[asan] Re-enable test/asan/TestCases/Linux/recvfrom.cc testcase.

This testcase failed on sanitizer-x86_64-linux buildbot in large parallel build due to race on
port 1234 between AddressSanitizer-i386-linux and AddressSanitizer-x86_64-linux instances of recvfrom.cc.
This patch tries to resolve the issue by relying on kernel to choose available port instead of hardcoding
its number in testcase.

Differential Revision: http://reviews.llvm.org/D17639

llvm-svn: 262204

8 years agoScopInfo: Remove indentation in hoistInvariantLoads
Tobias Grosser [Mon, 29 Feb 2016 07:29:42 +0000 (07:29 +0000)]
ScopInfo: Remove indentation in hoistInvariantLoads

We move verifyInvariantLoads out of this function to allow for an early return
without the need for code duplication. A similar transformation was suggested
by Johannes Doerfert in post commit review of r262033.

llvm-svn: 262203

8 years agoRevert "Add update_test.py script."
Tobias Grosser [Mon, 29 Feb 2016 07:12:10 +0000 (07:12 +0000)]
Revert "Add update_test.py script."

This reverts commit r261899. Even though I am not yet 100% certain, this is
commit is the only one that has some relation to the recent cmake failures
in Polly.

llvm-svn: 262202

8 years ago[X86] Disabling avx512f should also disable avx512vbmi and avx512ifma. Enabling avx51...
Craig Topper [Mon, 29 Feb 2016 06:51:38 +0000 (06:51 +0000)]
[X86] Disabling avx512f should also disable avx512vbmi and avx512ifma. Enabling avx512vbmi or avx512ifma should enable avx512f. Add command line switches and header defines for avx512ifma and avx512vbmi.

llvm-svn: 262201

8 years ago[X86] Enabling xsave should not enable AVX. I seem to have done this, but I don't...
Craig Topper [Mon, 29 Feb 2016 06:51:34 +0000 (06:51 +0000)]
[X86] Enabling xsave should not enable AVX. I seem to have done this, but I don't know why.

llvm-svn: 262200

8 years ago[OPENMP 4.5] Initial support for data members in 'reduction' clauses.
Alexey Bataev [Mon, 29 Feb 2016 05:54:20 +0000 (05:54 +0000)]
[OPENMP 4.5] Initial support for data members in 'reduction' clauses.

OpenMP 4.5 allows to privatize non-static data members of current class
in non-static member functions. Patch adds initial parsing/semantic
analysis for data members support in 'reduction' clauses.

llvm-svn: 262199

8 years ago[clang-cl] /EHc should not effect functions with explicit exception specifications
David Majnemer [Mon, 29 Feb 2016 01:40:36 +0000 (01:40 +0000)]
[clang-cl] /EHc should not effect functions with explicit exception specifications

Functions with an explicit exception specification have their behavior
dictated by the specification.  The additional /EHc behavior only comes
into play if no exception specification is given.

llvm-svn: 262198

8 years ago[clang-cl] /EHc should not have an effect on /EHa
David Majnemer [Mon, 29 Feb 2016 01:40:30 +0000 (01:40 +0000)]
[clang-cl] /EHc should not have an effect on /EHa

This matches behavior with MSVC.

llvm-svn: 262197

8 years ago[PM] Wire up optimization levels and default pipeline construction APIs
Chandler Carruth [Sun, 28 Feb 2016 22:16:03 +0000 (22:16 +0000)]
[PM] Wire up optimization levels and default pipeline construction APIs
in the PassBuilder.

These are really just stubs for now, but they give a nice API surface
that Clang or other tools can start learning about and enabling for
experimentation.

I've also wired up parsing various synthetic module pass names to
generate these set pipelines. This allows the pipelines to be combined
with other passes and have their order controlled, with clear separation
between the *kind* of canned pipeline, and the *level* of optimization
to be used within that canned pipeline.

The most interesting part of this patch is almost certainly the spec for
the different optimization levels. I don't think we can ever have hard
and fast rules that would make it easy to determine whether a particular
optimization makes sense at a particular level -- it will always be in
large part a judgement call. But hopefully this will outline the
expected rationale that should be used, and the direction that the
pipelines should be taken. Much of this was based on a long llvm-dev
discussion I started years ago to try and crystalize the intent behind
these pipelines, and now, at long long last I'm returning to the task of
actually writing it down somewhere that we can cite and try to be
consistent with.

Differential Revision: http://reviews.llvm.org/D12826

llvm-svn: 262196

8 years agoRemove dead code for ELF.
Rui Ueyama [Sun, 28 Feb 2016 21:59:02 +0000 (21:59 +0000)]
Remove dead code for ELF.

llvm-svn: 262195

8 years agoRemove dead code for ELF.
Rui Ueyama [Sun, 28 Feb 2016 21:22:44 +0000 (21:22 +0000)]
Remove dead code for ELF.

The preload feature was buggy that we had disabled it even for ELF.

llvm-svn: 262194

8 years agoRemove remaining code for COFF.
Rui Ueyama [Sun, 28 Feb 2016 21:22:40 +0000 (21:22 +0000)]
Remove remaining code for COFF.

llvm-svn: 262193

8 years agoRemove DriverTest.h.
Rui Ueyama [Sun, 28 Feb 2016 20:56:34 +0000 (20:56 +0000)]
Remove DriverTest.h.

Because the class is used only by one instance, we do not have to
use template there in DriverTest.h. Everything can be moved to
DarwinLdDriverTest.cpp.

llvm-svn: 262192

8 years agoMake the entry point function calls consistent. NFC.
Rui Ueyama [Sun, 28 Feb 2016 19:54:51 +0000 (19:54 +0000)]
Make the entry point function calls consistent. NFC.

llvm-svn: 262191

8 years agoMove functionality of UniversalDriver to the entry point file.
Rui Ueyama [Sun, 28 Feb 2016 19:50:14 +0000 (19:50 +0000)]
Move functionality of UniversalDriver to the entry point file.

UniversalDriver was used as a dispatcher to each platform-specific driver.
It had its own Options.td file. It was not just too much to parse only a
few options (we only want to parse -core, -flavor or argv[0]),
but also interpreted arguments too early. For example, if you invoke lld as
"lld -flavor gnu ... -help", then you'd get the UniversalDriver's help
message instead of GnuDriver's. This patch eliminates the use of Options
from the dispatcher.

http://reviews.llvm.org/D17686

llvm-svn: 262190

8 years ago[modules] Prefer more complete array types.
Vassil Vassilev [Sun, 28 Feb 2016 19:08:24 +0000 (19:08 +0000)]
[modules] Prefer more complete array types.

If we import a module that has a complete array type and one that has an
incomplete array type, the declaration found by name lookup might be the one with
the incomplete type, possibly resulting in rejects-valid.

Now, the name lookup prefers decls with a complete array types. Also,
diagnose cases when the redecl chain has array bound, different from the merge
candidate.

Reviewed by Richard Smith.

llvm-svn: 262189

8 years ago[PM] Appease mingw32's auto-import DLL build with minimal tweaks, with fix for clang.
NAKAMURA Takumi [Sun, 28 Feb 2016 17:17:00 +0000 (17:17 +0000)]
[PM] Appease mingw32's auto-import DLL build with minimal tweaks, with fix for clang.

char AnalysisBase::ID should be declared as extern and defined in one module.

llvm-svn: 262188

8 years ago[mips] Remove unused function declarations from MipsRegisterInfo.h. NFC.
Vasileios Kalintiris [Sun, 28 Feb 2016 16:55:28 +0000 (16:55 +0000)]
[mips] Remove unused function declarations from MipsRegisterInfo.h. NFC.

llvm-svn: 262187

8 years agoRevert r262185, "[PM] Appease mingw32's auto-import DLL build with minimal tweaks."
NAKAMURA Takumi [Sun, 28 Feb 2016 16:54:06 +0000 (16:54 +0000)]
Revert r262185, "[PM] Appease mingw32's auto-import DLL build with minimal tweaks."

I'll rework soon.

llvm-svn: 262186

8 years ago[PM] Appease mingw32's auto-import DLL build with minimal tweaks.
NAKAMURA Takumi [Sun, 28 Feb 2016 16:38:46 +0000 (16:38 +0000)]
[PM] Appease mingw32's auto-import DLL build with minimal tweaks.

char AnalysisBase::ID should be declared as extern and defined in one module.

llvm-svn: 262185

8 years agoDelete more COFF and ELF bits from the old linker.
Rafael Espindola [Sun, 28 Feb 2016 16:27:08 +0000 (16:27 +0000)]
Delete more COFF and ELF bits from the old linker.

llvm-svn: 262184

8 years agoDelete dead arch types.
Rafael Espindola [Sun, 28 Feb 2016 16:11:53 +0000 (16:11 +0000)]
Delete dead arch types.

llvm-svn: 262183

8 years agoDelete a bit more ELF and COFF from the old linker.
Rafael Espindola [Sun, 28 Feb 2016 16:08:21 +0000 (16:08 +0000)]
Delete a bit more ELF and COFF from the old linker.

llvm-svn: 262182

8 years agoDelete more ELF bits from the old linker.
Rafael Espindola [Sun, 28 Feb 2016 16:03:37 +0000 (16:03 +0000)]
Delete more ELF bits from the old linker.

llvm-svn: 262181

8 years agoWebAssembly: fix test
JF Bastien [Sun, 28 Feb 2016 15:44:54 +0000 (15:44 +0000)]
WebAssembly: fix test

Operand order seems to have changed, the new one is nicer.

llvm-svn: 262180

8 years agoWebAssembly: fix build
JF Bastien [Sun, 28 Feb 2016 15:33:53 +0000 (15:33 +0000)]
WebAssembly: fix build

More API churn, experimental target got sad.

llvm-svn: 262179

8 years agoAdd ARM EHABI-related constants to unwind.h.
Logan Chien [Sun, 28 Feb 2016 15:01:42 +0000 (15:01 +0000)]
Add ARM EHABI-related constants to unwind.h.

Adds a number of constants, defined in the ARM EHABI spec, to the Clang
lib/Headers/unwind.h header. This is prerequisite for landing
http://reviews.llvm.org/D15781, as previously discussed there.

Patch by Timon Van Overveldt.

llvm-svn: 262178

8 years ago[CLANG] [AVX512] [BUILTIN] Adding PSLL{V|W|Wi}{128|256|512} builtin
Michael Zuckerman [Sun, 28 Feb 2016 07:39:34 +0000 (07:39 +0000)]
[CLANG] [AVX512] [BUILTIN] Adding PSLL{V|W|Wi}{128|256|512} builtin

Differential Revision: http://reviews.llvm.org/D17685

llvm-svn: 262177

8 years ago[AVX512][PSLLW ][PSLLV] Change imm8 to int
Michael Zuckerman [Sun, 28 Feb 2016 07:32:10 +0000 (07:32 +0000)]
[AVX512][PSLLW ][PSLLV] Change imm8 to int

Differential Revision: http://reviews.llvm.org/D17684

llvm-svn: 262176

8 years agoELF: Remove OutSection class and use a map instead.
Rui Ueyama [Sun, 28 Feb 2016 05:09:11 +0000 (05:09 +0000)]
ELF: Remove OutSection class and use a map instead.

It is easier to handle section filler data separately rather than
merging with section names.

llvm-svn: 262175

8 years agoSimplify. NFC.
Rui Ueyama [Sun, 28 Feb 2016 04:48:54 +0000 (04:48 +0000)]
Simplify. NFC.

llvm-svn: 262174

8 years agoRemove dead header file.
Rui Ueyama [Sun, 28 Feb 2016 04:44:10 +0000 (04:44 +0000)]
Remove dead header file.

llvm-svn: 262173

8 years agoRemove obsolte TODOs. They are for the old gnu driver.
Rui Ueyama [Sun, 28 Feb 2016 04:07:16 +0000 (04:07 +0000)]
Remove obsolte TODOs. They are for the old gnu driver.

llvm-svn: 262172

8 years agoRemove lld/Core/range.h.
Rui Ueyama [Sun, 28 Feb 2016 04:03:01 +0000 (04:03 +0000)]
Remove lld/Core/range.h.

IIUC, range was an experiment to see how N3350 would work in LLD.
It turned out it didn't get traction, and it is basically duplicate
of iterator_range in ADT. We have only two occurrences of range,
and all of them are easily rewritten without it.

http://reviews.llvm.org/D17687

llvm-svn: 262171

8 years agoRemove an empty TODO file.
Rui Ueyama [Sun, 28 Feb 2016 03:34:34 +0000 (03:34 +0000)]
Remove an empty TODO file.

llvm-svn: 262170

8 years agoRemove unused #include.
Rui Ueyama [Sun, 28 Feb 2016 03:27:09 +0000 (03:27 +0000)]
Remove unused #include.

llvm-svn: 262169

8 years agoELF: Add --help option.
Rui Ueyama [Sun, 28 Feb 2016 03:18:09 +0000 (03:18 +0000)]
ELF: Add --help option.

llvm-svn: 262168

8 years agoELF: Add --version option.
Rui Ueyama [Sun, 28 Feb 2016 03:18:07 +0000 (03:18 +0000)]
ELF: Add --version option.

llvm-svn: 262167

8 years agoRemove dead code. This error type is for the old linker script reader.
Rui Ueyama [Sun, 28 Feb 2016 02:51:02 +0000 (02:51 +0000)]
Remove dead code. This error type is for the old linker script reader.

llvm-svn: 262166

8 years agoRemove dead header files.
Rui Ueyama [Sun, 28 Feb 2016 02:47:52 +0000 (02:47 +0000)]
Remove dead header files.

llvm-svn: 262165

8 years agoRemove a few bits of elf from the old linker.
Rafael Espindola [Sun, 28 Feb 2016 02:22:24 +0000 (02:22 +0000)]
Remove a few bits of elf from the old linker.

llvm-svn: 262164

8 years agoDelete the old script parser.
Rafael Espindola [Sun, 28 Feb 2016 02:14:48 +0000 (02:14 +0000)]
Delete the old script parser.

It was ELF specific.

llvm-svn: 262163

8 years agoTest cleanups
Xinliang David Li [Sun, 28 Feb 2016 00:54:28 +0000 (00:54 +0000)]
Test cleanups

llvm-svn: 262162

8 years ago[PGO] add a test for available_externally functions
Xinliang David Li [Sun, 28 Feb 2016 00:45:13 +0000 (00:45 +0000)]
[PGO] add a test for available_externally functions

llvm-svn: 262161

8 years agoRemove dead code.
Rui Ueyama [Sun, 28 Feb 2016 00:37:52 +0000 (00:37 +0000)]
Remove dead code.

llvm-svn: 262160

8 years agoRename elf2 to elf.
Rafael Espindola [Sun, 28 Feb 2016 00:25:54 +0000 (00:25 +0000)]
Rename elf2 to elf.

llvm-svn: 262159

8 years agoRemove the old ELF linker.
Rafael Espindola [Sun, 28 Feb 2016 00:10:58 +0000 (00:10 +0000)]
Remove the old ELF linker.

I think it is clear by now that the new linker is viable.

llvm-svn: 262158

8 years ago[PGO] Remove redundant counter copies for avail_extern functions.
Xinliang David Li [Sat, 27 Feb 2016 23:11:30 +0000 (23:11 +0000)]
[PGO] Remove redundant counter copies for avail_extern functions.

Differential Revision: http://reviews.llvm.org/D17654

llvm-svn: 262157

8 years agoRevert "Fix bug with register values byte order in expression evaluation."
Todd Fiala [Sat, 27 Feb 2016 22:48:50 +0000 (22:48 +0000)]
Revert "Fix bug with register values byte order in expression evaluation."

This reverts commit r262041, which caused asserts starting yesterday on the OS X testbot.

See details in:
https://llvm.org/bugs/show_bug.cgi?id=26758

llvm-svn: 262156

8 years agoELF: Remove relSize function from ICF.cpp. NFC.
Rui Ueyama [Sat, 27 Feb 2016 20:29:45 +0000 (20:29 +0000)]
ELF: Remove relSize function from ICF.cpp. NFC.

llvm-svn: 262155

8 years agoCodeGen: Remove an iterator => pointer conversion, NFC
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 20:27:44 +0000 (20:27 +0000)]
CodeGen: Remove an iterator => pointer conversion, NFC

Part of PR26753.

llvm-svn: 262154

8 years agoAMDGPU: More bits of frame index are known to be zero
Matt Arsenault [Sat, 27 Feb 2016 20:26:57 +0000 (20:26 +0000)]
AMDGPU: More bits of frame index are known to be zero

The maximum private allocation for the whole GPU is 4G,
so the maximum possible index for a single workitem is the
maximum size divided by the smallest granularity for a dispatch.

This increases the number of known zero high bits, which
enables more offset folding. The maximum private size per
workitem with this is 128M but may be smaller still.

llvm-svn: 262153

8 years agoCodeGen: Use MachineInstr& in InlineSpiller::rematerializeFor()
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 20:23:14 +0000 (20:23 +0000)]
CodeGen: Use MachineInstr& in InlineSpiller::rematerializeFor()

InlineSpiller::rematerializeFor() never uses its parameter as an
iterator, so take it by reference instead.  This removes an implicit
conversion from MachineBasicBlock::iterator to MachineInstr*.

llvm-svn: 262152

8 years agoCodeGen: Update LiveIntervalAnalysis API to use MachineInstr&, NFC
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 20:14:29 +0000 (20:14 +0000)]
CodeGen: Update LiveIntervalAnalysis API to use MachineInstr&, NFC

These parameters aren't expected to be null, so take them by reference.

llvm-svn: 262151

8 years agoCodeGen: Change MachineInstr to use MachineInstr&, NFC
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 20:01:33 +0000 (20:01 +0000)]
CodeGen: Change MachineInstr to use MachineInstr&, NFC

Change MachineInstr API to prefer MachineInstr& over MachineInstr*
whenever the parameter is expected to be non-null.  Slowly inching
toward being able to fix PR26753.

llvm-svn: 262149

8 years agoDAGCombiner: Don't unnecessarily swap operands in ReassociateOps
Matt Arsenault [Sat, 27 Feb 2016 19:57:45 +0000 (19:57 +0000)]
DAGCombiner: Don't unnecessarily swap operands in ReassociateOps

In the case where op = add, y = base_ptr, and x = offset, this
transform:

(op y, (op x, c1)) -> (op (op x, y), c1)

breaks the canonical form of add by putting the base pointer in the
second operand and the offset in the first.

This fix is important for the R600 target, because for some address
spaces the base pointer and the offset are stored in separate register
classes. The old pattern caused the ISel code for matching addressing
modes to put the base pointer and offset in the wrong register classes,
which required no-trivial code transformations to fix.

llvm-svn: 262148

8 years ago[UBSan] Fix isDerivedFromAtOffset on iOS ARM64
Filipe Cabecinhas [Sat, 27 Feb 2016 19:57:44 +0000 (19:57 +0000)]
[UBSan] Fix isDerivedFromAtOffset on iOS ARM64

Summary:
iOS on ARM64 doesn't unique RTTI.
Ref: clang's iOS64CXXABI::shouldRTTIBeUnique()

Due to this, pointer-equality will not necessarily work in this
architecture, across dylib boundaries.

dynamic_cast<>() will (as expected) still work, since Apple ships with
one prepared for this, but we can't rely on the type names being
pointer-equal.

I've limited the expensive strcmp check to the specific architecture
which needs it.

Example which triggers this bug:

lib.h:
  struct X {
    virtual ~X() {}
  };
  X *libCall();

lib.mm:
  X *libCall() {
    return new X;
  }

prog.mm:
  int main() {
    X *px = libCall();
    delete px;
  }

Expected output: Nothing
Actual output:
<unknown>: runtime error: member call on address 0x00017001ef50 which does not point to an object of type 'X'
0x00017001ef50: note: object is of type 'X'
 00 00 00 00  60 00 0f 00 01 00 00 00  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  00 00 00 00
              ^~~~~~~~~~~~~~~~~~~~~~~
              vptr for ‘X’

Reviewers: kubabrecka, samsonov, eugenis, rsmith

Subscribers: aemerson, llvm-commits, rengolin

Differential Revision: http://reviews.llvm.org/D11502

llvm-svn: 262147

8 years agoCodeGen: Use MachineInstr& in HashMachineInstr, NFC
Duncan P. N. Exon Smith [Sat, 27 Feb 2016 19:48:01 +0000 (19:48 +0000)]
CodeGen: Use MachineInstr& in HashMachineInstr, NFC

Also update HashEndOfMBB to take MachineBasicBlock&.

llvm-svn: 262146