platform/upstream/llvm.git
8 years agoPM: Reorder the functions used for SinkingPass. NFC
Justin Bogner [Fri, 22 Apr 2016 19:54:04 +0000 (19:54 +0000)]
PM: Reorder the functions used for SinkingPass. NFC

This will make the port to the new PM easier to follow.

llvm-svn: 267198

8 years ago[DeadStoreElimination] Shorten beginning of memset overwritten by later stores
Jun Bum Lim [Fri, 22 Apr 2016 19:51:29 +0000 (19:51 +0000)]
[DeadStoreElimination] Shorten beginning of memset overwritten by later stores

Summary: This change will shorten memset if the beginning of memset is overwritten by later stores.

Reviewers: hfinkel, eeckstein, dberlin, mcrosier

Subscribers: mgrang, mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D18906

llvm-svn: 267197

8 years agoPM: Port DCE to the new pass manager
Justin Bogner [Fri, 22 Apr 2016 19:40:41 +0000 (19:40 +0000)]
PM: Port DCE to the new pass manager

Also add a very basic test, since apparently there aren't any tests
for DCE whatsoever to add the new pass version to.

llvm-svn: 267196

8 years agoInline SymbolTable::compareCommons and add comments. NFC.
Rui Ueyama [Fri, 22 Apr 2016 19:34:59 +0000 (19:34 +0000)]
Inline SymbolTable::compareCommons and add comments. NFC.

llvm-svn: 267195

8 years ago Revert "GC entries of SHF_MERGE sections."
Rafael Espindola [Fri, 22 Apr 2016 19:31:35 +0000 (19:31 +0000)]
Revert "GC entries of SHF_MERGE sections."

    This reverts commit r267164.

    Revert "Trying to fix the windows build."

    This reverts commit r267168.

Debugging a bootstrap problem.

llvm-svn: 267194

8 years agoMachineScheduler: Move code to initialize a Candidate out of tryCandidate(); NFC
Matthias Braun [Fri, 22 Apr 2016 19:10:15 +0000 (19:10 +0000)]
MachineScheduler: Move code to initialize a Candidate out of tryCandidate(); NFC

llvm-svn: 267191

8 years ago[LoopUtils] Extend findStringMetadataForLoop to return the value for metadata
Adam Nemet [Fri, 22 Apr 2016 19:10:05 +0000 (19:10 +0000)]
[LoopUtils] Extend findStringMetadataForLoop to return the value for metadata

E.g. for:

  !1 = {"llvm.distribute", i32 1}

it now returns the MDOperand for 1.

I will use this in LoopDistribution to check the value of the metadata.

Note that the change is backward-compatible with its current use in
LoopVersioningLICM.  An Optional implicitly converts to a bool depending
whether it contains a value or not.

llvm-svn: 267190

8 years agoMachineScheduler: Limit the size of the ready list.
Matthias Braun [Fri, 22 Apr 2016 19:09:17 +0000 (19:09 +0000)]
MachineScheduler: Limit the size of the ready list.

Avoid quadratic complexity in unusually large basic blocks by limiting
the size of the ready lists.

Differential Revision: http://reviews.llvm.org/D19349

llvm-svn: 267189

8 years agoELF: Scan for undefined symbols in shlibs and symbols in dynamic lists before LTO.
Peter Collingbourne [Fri, 22 Apr 2016 18:47:52 +0000 (18:47 +0000)]
ELF: Scan for undefined symbols in shlibs and symbols in dynamic lists before LTO.

This ensures that these features work correctly with bitcode files.

llvm-svn: 267188

8 years ago[EarlyCSE/CVP] Add stats for CVPs and make sure to account for any Changes.
Chad Rosier [Fri, 22 Apr 2016 18:47:21 +0000 (18:47 +0000)]
[EarlyCSE/CVP] Add stats for CVPs and make sure to account for any Changes.

llvm-svn: 267187

8 years agoFix a bug involving deferred decl emission and PCH
Reid Kleckner [Fri, 22 Apr 2016 18:46:33 +0000 (18:46 +0000)]
Fix a bug involving deferred decl emission and PCH

For various reasons, involving dllexport and class linkage compuations,
we have to wait until after the semicolon after a class declaration to
emit inline methods. These are "deferred" decls. Before this change,
finishing the tag decl would trigger us to deserialize some PCH so that
we could make a "pretty" IR-level type. Deserializing the PCH triggered
calls to HandleTopLevelDecl, which, when done, checked the deferred decl
list, and emitted some dllexported decls that weren't ready.

Avoid this re-entrancy. Deferred decls should not get emitted when a tag
is finished, they should only be emitted after a real top level decl in
the main file.

llvm-svn: 267186

8 years ago[AArch64] When creating MRS instruction, make sure the destination register is
Quentin Colombet [Fri, 22 Apr 2016 18:46:17 +0000 (18:46 +0000)]
[AArch64] When creating MRS instruction, make sure the destination register is
declared as a definition.

This fixes the machine verifier error for CodeGen/AArch64/nzcv-save.ll.

llvm-svn: 267185

8 years agoELF: Implement --export-dynamic-symbol.
Peter Collingbourne [Fri, 22 Apr 2016 18:44:06 +0000 (18:44 +0000)]
ELF: Implement --export-dynamic-symbol.

llvm-svn: 267184

8 years agoELF: Move Visibility, IsUsedInRegularObj and MustBeInDynSym flags to Symbol.
Peter Collingbourne [Fri, 22 Apr 2016 18:42:48 +0000 (18:42 +0000)]
ELF: Move Visibility, IsUsedInRegularObj and MustBeInDynSym flags to Symbol.

These are properties of a symbol name, rather than a particular instance
of a symbol in an object file. We can simplify the code by collecting these
properties in Symbol.

The MustBeInDynSym flag has been renamed ExportDynamic, as its semantics
have been changed to be the same as those of --dynamic-list and
--export-dynamic-symbol, which do not cause hidden symbols to be exported.

Differential Revision: http://reviews.llvm.org/D19400

llvm-svn: 267183

8 years ago[LoopVersioningLICM] Add test coverage for llvm.loop.licm_versioning.disable
Adam Nemet [Fri, 22 Apr 2016 18:34:50 +0000 (18:34 +0000)]
[LoopVersioningLICM] Add test coverage for llvm.loop.licm_versioning.disable

In the next change, I am generalizing the function
findStringMetadataForLoop and I want to make sure I don't break this.
Looks like there was no coverage for this so far.

llvm-svn: 267182

8 years ago[LTO] Create Undefined Bitcode symbol when we drop a comdat member.
Davide Italiano [Fri, 22 Apr 2016 18:26:33 +0000 (18:26 +0000)]
[LTO] Create Undefined Bitcode symbol when we drop a comdat member.

Differential Revision:  http://reviews.llvm.org/D19389

llvm-svn: 267181

8 years ago[AArch64][AdvSIMDScalar] Update the kill flags correctly.
Quentin Colombet [Fri, 22 Apr 2016 18:09:14 +0000 (18:09 +0000)]
[AArch64][AdvSIMDScalar] Update the kill flags correctly.

We used to simply set the kill flags to true when transforming a scalar
instruction to a vector one.
SrcScalar1 = copy SrcVector1
... = opScalar SrcScalar1
=>
SrcScalar1 = copy SrcVector1
... = opVector SrcVector1<kill>

This is obviously wrong. The proper update consists in:
1. Propagate the kill status from the copy to the new opVector
2. Reset the kill status on the copy, since the live-range of
   SrcVector1 got extended.

This fixes some of the machine verifier errors for AArch64 with make check.

llvm-svn: 267180

8 years agotest: split test into two runs
Saleem Abdulrasool [Fri, 22 Apr 2016 18:06:51 +0000 (18:06 +0000)]
test: split test into two runs

Rather than checking both stdout and stderr simultaneously, split it into two
tests.  This apparently breaks on Windows where MSVCRT does not buffer output
correctly.  NFC.

Thanks to chapuni for bringing the issue to my attention!

llvm-svn: 267179

8 years ago[Hexagon] Use common Pat classes for selecting code for intrinsics
Krzysztof Parzyszek [Fri, 22 Apr 2016 18:05:55 +0000 (18:05 +0000)]
[Hexagon] Use common Pat classes for selecting code for intrinsics

llvm-svn: 267178

8 years ago[SimplifyCFG] Add final missing implications to isImpliedTrueByMatchingCmp.
Chad Rosier [Fri, 22 Apr 2016 17:57:34 +0000 (17:57 +0000)]
[SimplifyCFG] Add final missing implications to isImpliedTrueByMatchingCmp.

Summary: eq imply [u|s]ge and [u|s]le are true.

Remove redundant logic by implementing isImpliedFalseByMatchingCmp(Pred1, Pred2)
as isImpliedTrueByMatchingCmp(Pred1, getInversePredicate(Pred2)).

llvm-svn: 267177

8 years agoPoint link to CODE_OWNERS.txt to klaus instead of viewvc.
Michael Kuperstein [Fri, 22 Apr 2016 17:41:12 +0000 (17:41 +0000)]
Point link to CODE_OWNERS.txt to klaus instead of viewvc.

llvm-svn: 267176

8 years agoHave isKnownNotFullPoison be smarter around control flow
Sanjoy Das [Fri, 22 Apr 2016 17:41:06 +0000 (17:41 +0000)]
Have isKnownNotFullPoison be smarter around control flow

Summary:
(... while still not using a PostDomTree)

The way we use isKnownNotFullPoison from SCEV today, the new CFG walking
logic will not trigger for any realistic cases -- it will kick in only
for situations where we could have merged the contiguous basic blocks
anyway[0], since the poison generating instruction dominates all of its
non-PHI uses (which are the only uses we consider right now).

However, having this change in place will allow a later bugfix to break
fewer llvm-lit tests.

[0]: i.e. cases where block A branches to block B and B is A's only
successor and A is B's only predecessor.

Reviewers: broune, bjarke.roune

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D19212

llvm-svn: 267175

8 years ago[Hexagon] Properly close live range in HexagonBlockRanges ---add testcase
Krzysztof Parzyszek [Fri, 22 Apr 2016 17:30:13 +0000 (17:30 +0000)]
[Hexagon] Properly close live range in HexagonBlockRanges ---add testcase

llvm-svn: 267174

8 years ago[Hexagon] Properly close live range in HexagonBlockRanges
Krzysztof Parzyszek [Fri, 22 Apr 2016 17:27:22 +0000 (17:27 +0000)]
[Hexagon] Properly close live range in HexagonBlockRanges

llvm-svn: 267173

8 years agoPM: Remove some redundant name() methods
Justin Bogner [Fri, 22 Apr 2016 17:25:43 +0000 (17:25 +0000)]
PM: Remove some redundant name() methods

These passes all get names from PassInfoMixin already, we don't need
to override them.

llvm-svn: 267172

8 years ago[SimplifyCFG] Add missing implications to isImpliedTrueByMatchingCmp.
Chad Rosier [Fri, 22 Apr 2016 17:14:12 +0000 (17:14 +0000)]
[SimplifyCFG] Add missing implications to isImpliedTrueByMatchingCmp.

Summary: [u|s]gt and [u|s]lt imply [u|s]ge and [u|s]le are true, respectively.
I've simplified the existing tests and added additional tests to cover the new
cases mentioned above.  I've also added tests for all the cases where the
first compare doesn't imply anything about the second compare.

llvm-svn: 267171

8 years ago[SimplifyCFG] Simplify code review by temporarily removing this test file.
Chad Rosier [Fri, 22 Apr 2016 17:14:08 +0000 (17:14 +0000)]
[SimplifyCFG] Simplify code review by temporarily removing this test file.

A followup commit will replace these tests with simplified and more inclusive
tests.  The diff is unreadable if this were to be done in a single commit.

llvm-svn: 267170

8 years agounwind: unify some more macros
Saleem Abdulrasool [Fri, 22 Apr 2016 17:11:05 +0000 (17:11 +0000)]
unwind: unify some more macros

The macros were defined identically across both cases.  Unify the definitions to
have a single definition for _LIBUWNIND_{HIDDEN,EXPORT} and _LIBUNWIND_LOG.
NFC.

llvm-svn: 267169

8 years agoTrying to fix the windows build.
Rafael Espindola [Fri, 22 Apr 2016 17:10:28 +0000 (17:10 +0000)]
Trying to fix the windows build.

llvm-svn: 267168

8 years ago[AMDGPU] Insert nop pass: take care of outstanding feedback
Konstantin Zhuravlyov [Fri, 22 Apr 2016 17:04:51 +0000 (17:04 +0000)]
[AMDGPU] Insert nop pass: take care of outstanding feedback

- Switch few loops to range-based for loops
- Fix nop insertion at the end of BB
- Fix formatting
- Check for endpgm

Differential Revision: http://reviews.llvm.org/D19380

llvm-svn: 267167

8 years ago[mips][microMIPS] Revert commit r266861.
Zoran Jovanovic [Fri, 22 Apr 2016 16:53:15 +0000 (16:53 +0000)]
[mips][microMIPS] Revert commit r266861.
Commit r266861 was the reason for failing tests in LLVM test suite.

llvm-svn: 267166

8 years ago[Hexagon] Teach mux expansion how to deal with undef predicates
Krzysztof Parzyszek [Fri, 22 Apr 2016 16:47:01 +0000 (16:47 +0000)]
[Hexagon] Teach mux expansion how to deal with undef predicates

llvm-svn: 267165

8 years agoGC entries of SHF_MERGE sections.
Rafael Espindola [Fri, 22 Apr 2016 16:46:08 +0000 (16:46 +0000)]
GC entries of SHF_MERGE sections.

It is a fairly direct extension of the gc algorithm. For merge sections
instead of remembering just a live bit, we remember which offsets were
used.

This reduces the .rodata sections in chromium from 9648861 to 9477472
bytes.

llvm-svn: 267164

8 years agoThis reverts commit r267154 and r267161.
Rafael Espindola [Fri, 22 Apr 2016 16:39:59 +0000 (16:39 +0000)]
This reverts commit r267154 and r267161.

It turns out that this will read data from the section to properly
handle Elf_Rel implicit addends.

Sorry for the noise.

Original messages:

Try to fix Windows lld build.

Move getRelocTarget to ObjectFile.
It doesn't use anything from the InputSection.

llvm-svn: 267163

8 years ago[Hexagon] Add definitions for trap/pause instructions
Krzysztof Parzyszek [Fri, 22 Apr 2016 16:25:00 +0000 (16:25 +0000)]
[Hexagon] Add definitions for trap/pause instructions

Also add tests for other instructions from HexagonSystemInst.td.

llvm-svn: 267162

8 years agoTry to fix Windows lld build.
Nico Weber [Fri, 22 Apr 2016 15:23:02 +0000 (15:23 +0000)]
Try to fix Windows lld build.

llvm\tools\lld\ELF\MarkLive.cpp(49): error C2872: 'ObjectFile': ambiguous symbol
llvm\tools\lld\elf\InputFiles.h(100): note: could be 'lld::elf::ObjectFile'
llvm\include\llvm/Object/IRObjectFile.h(26): note: or       'llvm::object::ObjectFile'
llvm\tools\lld\ELF\MarkLive.cpp(133): note: see reference to function template instantiation
    'void forEachSuccessor<ELFT>(lld::elf::InputSection<ELFT> *,
                                 std::function<void (lld::elf::InputSectionBase<ELFT> *)>)'
    being compiled with
                [ ELFT=llvm::object::ELF32LE ]
llvm\tools\lld\ELF\MarkLive.cpp(136): note: see reference to function template instantiation
    'void lld::elf::markLive<llvm::object::ELF32LE>(lld::elf::SymbolTable<llvm::object::ELF32LE> *)
    being compiled

llvm-svn: 267161

8 years agoPostRAHazardRecocgnizer: Fix unused-private-field warning
Tom Stellard [Fri, 22 Apr 2016 15:11:08 +0000 (15:11 +0000)]
PostRAHazardRecocgnizer: Fix unused-private-field warning

llvm-svn: 267160

8 years ago[Hexagon] Add definitions for circular and bit-reverse loads/stores
Krzysztof Parzyszek [Fri, 22 Apr 2016 14:58:46 +0000 (14:58 +0000)]
[Hexagon] Add definitions for circular and bit-reverse loads/stores

llvm-svn: 267159

8 years agoRevert "[LSAN] Fix test swapcontext.cc on MIPS"
Renato Golin [Fri, 22 Apr 2016 14:56:30 +0000 (14:56 +0000)]
Revert "[LSAN] Fix test swapcontext.cc on MIPS"

This reverts commit r266716, as it breaks the self-hosting on Thumb2 buildbot.

llvm-svn: 267158

8 years ago[MemorySSA] Fix bug in CachingMemorySSAWalker::invalidateInfo
Geoff Berry [Fri, 22 Apr 2016 14:44:10 +0000 (14:44 +0000)]
[MemorySSA] Fix bug in CachingMemorySSAWalker::invalidateInfo

Summary:
CachingMemorySSAWalker::invalidateInfo was using IsCall to determine
which cache map needed to be cleared of entries referring to the invalidated
MemoryAccess, but there could also be entries referring to it in the
other cache map (value entries, not key entries).  This change just
clears both tables to be conservatively correct.

Also add a verifyRemoved() function, called when expensive
checks (i.e. XDEBUG) are enabled to verify that the invalidated
MemoryAccess object is not referenced in any of the caches.

Reviewers: dberlin, george.burgess.iv

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D19388

llvm-svn: 267157

8 years agoCodeGen: Add a stand-alone hazard recognizer pass
Tom Stellard [Fri, 22 Apr 2016 14:43:50 +0000 (14:43 +0000)]
CodeGen: Add a stand-alone hazard recognizer pass

Summary:
This new pass allows targets to use the hazard recognizer without having
to also run one of the schedulers.  This is useful when compiling with
optimizations disabled for targets that still need noop hazards
to be handled correctly.

Reviewers: hfinkel, atrick

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D18594

llvm-svn: 267156

8 years ago[clang-tidy] fix link in Release Notes
Haojian Wu [Fri, 22 Apr 2016 14:43:39 +0000 (14:43 +0000)]
[clang-tidy] fix link in Release Notes

Summary: This is intended to fix https://llvm.org/bugs/show_bug.cgi?id=27426

Patch by Kirill Bobyrev!

Reviewers: alexfh, LegalizeAdulthood, hokein

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D19406

llvm-svn: 267155

8 years agoMove getRelocTarget to ObjectFile.
Rafael Espindola [Fri, 22 Apr 2016 14:17:14 +0000 (14:17 +0000)]
Move getRelocTarget to ObjectFile.

It doesn't use anything from the InputSection.

llvm-svn: 267154

8 years ago[EarlyCSE] Don't add the overflow flags to the hash
David Majnemer [Fri, 22 Apr 2016 14:12:50 +0000 (14:12 +0000)]
[EarlyCSE] Don't add the overflow flags to the hash

We take the intersection of overflow flags while CSE'ing.
This permits us to consider two instructions with different overflow
behavior to be replaceable.

llvm-svn: 267153

8 years agoEmit code16 in assembly in 16-bit mode
Nirav Dave [Fri, 22 Apr 2016 13:36:11 +0000 (13:36 +0000)]
Emit code16 in assembly in 16-bit mode

Summary:
When generating assembly using -m16 we must explicitly mark it as
16-bit. Emit .code16 at beginning of file. Fixes wrong results when
using -fno-integrated-as.

Reviewers: dwmw2

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D19392

llvm-svn: 267152

8 years ago[mips] Fix select patterns for MIPS64
Simon Dardis [Fri, 22 Apr 2016 13:19:22 +0000 (13:19 +0000)]
[mips] Fix select patterns for MIPS64

When targetting MIPS64R6 some of the patterns for select were guarded by a
broken predicate. The predicate was supposed to test if a constant value
could fit in a 16 bit zero-extended field. Instead the value was tested to
fit in a 16 bit sign-extended field. For negative constants of native word
width this resulted in wrong code generation.

Reviewers: vkalintiris, dsanders

Differential Review: http://reviews.llvm.org/D19378

llvm-svn: 267151

8 years agoDocument source_filename in LangRef.
Teresa Johnson [Fri, 22 Apr 2016 13:09:17 +0000 (13:09 +0000)]
Document source_filename in LangRef.

Summary: Add documentation for new LLVM IR source_filename identifier.

Reviewers: joker.eph, majnemer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D18857

llvm-svn: 267150

8 years ago[mips] Fix a small typo that would leave BLTZC out of getAnalyzableBrOpc().'
Vasileios Kalintiris [Fri, 22 Apr 2016 13:05:51 +0000 (13:05 +0000)]
[mips] Fix a small typo that would leave BLTZC out of getAnalyzableBrOpc().'

llvm-svn: 267149

8 years agoRevert r267049, r26706[16789], r267071 - Refactor raw pdb dumper into library
Daniel Sanders [Fri, 22 Apr 2016 12:04:42 +0000 (12:04 +0000)]
Revert r267049, r26706[16789], r267071 - Refactor raw pdb dumper into library

r267049 broke multiple buildbots (e.g. clang-cmake-mips, and clang-x86_64-linux-selfhost-modules) which the follow-ups have not yet resolved and this is preventing subsequent committers from being notified about additional failures on the affected buildbots.

llvm-svn: 267148

8 years agoSimplify. NFC.
Rafael Espindola [Fri, 22 Apr 2016 11:46:28 +0000 (11:46 +0000)]
Simplify. NFC.

llvm-svn: 267147

8 years agoBail for complex execution contexts of invariant loads
Johannes Doerfert [Fri, 22 Apr 2016 11:41:14 +0000 (11:41 +0000)]
Bail for complex execution contexts of invariant loads

llvm-svn: 267146

8 years ago[ELF] - Implemented linkerscript ALIGN command
George Rimar [Fri, 22 Apr 2016 11:40:53 +0000 (11:40 +0000)]
[ELF] - Implemented linkerscript ALIGN command

ALIGN(exp)
Return the location counter (.) aligned to the next exp boundary. (https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/4/html/Using_ld_the_GNU_Linker/expressions.html)

Patch implements this command.
This fixes PR27406.

Differential revision: http://reviews.llvm.org/D19364

llvm-svn: 267145

8 years agoAMDGPU/SI: Add test missed in rL266865
Nikolay Haustov [Fri, 22 Apr 2016 11:39:43 +0000 (11:39 +0000)]
AMDGPU/SI: Add test missed in rL266865

llvm-svn: 267144

8 years agoEarly exit for addInvariantLoads
Johannes Doerfert [Fri, 22 Apr 2016 11:38:44 +0000 (11:38 +0000)]
Early exit for addInvariantLoads

llvm-svn: 267143

8 years agoBail for complex alias checks
Johannes Doerfert [Fri, 22 Apr 2016 11:38:19 +0000 (11:38 +0000)]
Bail for complex alias checks

llvm-svn: 267142

8 years agoRepair doxygen comment [NFC]
Johannes Doerfert [Fri, 22 Apr 2016 11:36:13 +0000 (11:36 +0000)]
Repair doxygen comment [NFC]

llvm-svn: 267141

8 years agoRecommitted r267132 "[ELF] - implemented ternary operator for linkerscript expressions"
George Rimar [Fri, 22 Apr 2016 11:28:54 +0000 (11:28 +0000)]
Recommitted r267132 "[ELF] - implemented ternary operator for linkerscript expressions"
With fix: removed redundant Dot parameter.

Original commit message:
[ELF] - implemented ternary operator for linkerscript expressions

Patch implements ternary operator for linkerscript expressions.
Like:

SECTIONS {
 . = 0x1 ? 0x2 : 0x3;
...
}

Differential revision: http://reviews.llvm.org/D19332

llvm-svn: 267140

8 years ago[InstCombine] Preserve fast math flags when combining PHIs
Silviu Baranga [Fri, 22 Apr 2016 11:21:36 +0000 (11:21 +0000)]
[InstCombine] Preserve fast math flags when combining PHIs

Summary:
When optimizing PHIs which have inputs floating point binary
operators, we preserve all IR flags except the fast math
flags.

This change removes the logic which tracked some of the IR flags
(no wrap, exact) and replaces it by doing an and on the IR flags of
all inputs to the PHI - which will also handle the fast math
flags.

Reviewers: majnemer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D19370

llvm-svn: 267139

8 years ago[mips][microMIPS] Implement SLT, SLTI, SLTIU, SLTU microMIPS32r6 instructions
Hrvoje Varga [Fri, 22 Apr 2016 11:18:40 +0000 (11:18 +0000)]
[mips][microMIPS] Implement SLT, SLTI, SLTIU, SLTU microMIPS32r6 instructions
Differential Revision: http://reviews.llvm.org/D19354

llvm-svn: 267137

8 years agoRevert "[sanitizer] Allow the sanitizer allocator to use a non-fixed address range...
Renato Golin [Fri, 22 Apr 2016 11:18:14 +0000 (11:18 +0000)]
Revert "[sanitizer] Allow the sanitizer allocator to use a non-fixed address range. An allocator with a non-fixed address range will be attack-resistan. NFC for the sanitizers at this point."

This reverts commit r267094, because it broke a lot of MSAN tests in AArch64.
Being NFC and all, this needs some deeper investigation before it goes in again.

llvm-svn: 267136

8 years ago[clang][AVX512][Builtin] adding intrinsics for vf{n}madd{ss|sd} and vf{n}sub{ss|sd...
Michael Zuckerman [Fri, 22 Apr 2016 10:56:24 +0000 (10:56 +0000)]
[clang][AVX512][Builtin] adding intrinsics for vf{n}madd{ss|sd} and vf{n}sub{ss|sd} instruction set

Differential Revision: http://reviews.llvm.org/D19320

llvm-svn: 267135

8 years agoRevert r267132 as it broke buildbot.
George Rimar [Fri, 22 Apr 2016 10:51:34 +0000 (10:51 +0000)]
Revert r267132 as it broke buildbot.

llvm-svn: 267134

8 years agoRenumber ThreadSanitizer-provided thread IDs to match LLDB thread numbers.
Kuba Brecka [Fri, 22 Apr 2016 10:40:14 +0000 (10:40 +0000)]
Renumber ThreadSanitizer-provided thread IDs to match LLDB thread numbers.

llvm-svn: 267133

8 years ago[ELF] - implemented ternary operator for linkerscript expressions
George Rimar [Fri, 22 Apr 2016 10:35:34 +0000 (10:35 +0000)]
[ELF] - implemented ternary operator for linkerscript expressions

Patch implements ternary operator for linkerscript expressions.
Like:

SECTIONS {
 . = 0x1 ? 0x2 : 0x3;
...
}

Differential revision: http://reviews.llvm.org/D19332

llvm-svn: 267132

8 years agoFix some non-standard parts of our test suite. Reported by STL
Eric Fiselier [Fri, 22 Apr 2016 10:33:56 +0000 (10:33 +0000)]
Fix some non-standard parts of our test suite. Reported by STL

llvm-svn: 267131

8 years ago[mips][microMIPS] Add R_MICROMIPS_PC18_S3 relocation
Zoran Jovanovic [Fri, 22 Apr 2016 10:15:12 +0000 (10:15 +0000)]
[mips][microMIPS] Add R_MICROMIPS_PC18_S3 relocation

Differential Revision: http://reviews.llvm.org/D15026

llvm-svn: 267130

8 years ago[Clang][AVX512][BUILTIN] Adding scalar intrinsics for rsqrt14 ,rcp14, getexp and...
Michael Zuckerman [Fri, 22 Apr 2016 10:06:10 +0000 (10:06 +0000)]
[Clang][AVX512][BUILTIN] Adding scalar intrinsics for rsqrt14 ,rcp14, getexp and getmant instruction set

Differential Revision: http://reviews.llvm.org/D19326

llvm-svn: 267129

8 years agoFix -Wunused-variable in non-asserts build.
Eric Liu [Fri, 22 Apr 2016 09:50:31 +0000 (09:50 +0000)]
Fix -Wunused-variable in non-asserts build.

llvm-svn: 267128

8 years agoRevert r267098 - [MachineCombiner] Support for floating-point FMA on ARM64
Daniel Sanders [Fri, 22 Apr 2016 09:37:26 +0000 (09:37 +0000)]
Revert r267098 - [MachineCombiner] Support for floating-point FMA on ARM64

It introduced buildbot failures on clang-cmake-mips, clang-ppc64le-linux, among others.

llvm-svn: 267127

8 years ago[ASAN] Use struct instead of array in sancov.py
Sagar Thakur [Fri, 22 Apr 2016 09:20:22 +0000 (09:20 +0000)]
[ASAN] Use struct instead of array in sancov.py

Summary: When using 32-bit python with 64-bit asan the pc array in sancov.py cannot fit in 64-bit pc's because the type-code 'L' for
arrays in python corresponds to the C type long which is only of 4 bytes. Because of this some of the coverage tool tests fail on
mips. To fix these test possible solutions are to use 64-bit python or use struct.unpack with the 'Q' type-code. We have used
struct.unpack with 'Q' type code since it is not appropriate to have a 64-bit python on all hosts.

Reviewed by kcc, aizatsky

Differential: http://reviews.llvm.org/D18817
llvm-svn: 267126

8 years ago[OPENMP] Fix for PR27463: Privatizing struct fields with array type
Alexey Bataev [Fri, 22 Apr 2016 09:05:03 +0000 (09:05 +0000)]
[OPENMP] Fix for PR27463: Privatizing struct fields with array type
causes code generation failure.

The codegen part of firstprivate clause for member decls used type of
original variable without skipping reference type from
OMPCapturedExprDecl. Patch fixes this problem.

llvm-svn: 267125

8 years agoUpdate comment in lldb-enumerations.h
Pavel Labath [Fri, 22 Apr 2016 08:41:07 +0000 (08:41 +0000)]
Update comment in lldb-enumerations.h

llvm-svn: 267124

8 years ago[X86]: Changing cost for “TRUNCATE v16i32 to v16i8” in SSE4.1 mode.
Ashutosh Nema [Fri, 22 Apr 2016 08:34:05 +0000 (08:34 +0000)]
[X86]: Changing cost for “TRUNCATE v16i32 to v16i8” in SSE4.1 mode.

Summary:
rL256194 transforms truncations between vectors of integers into PACKUS/PACKSS
operations during DAG combine. This generates better code for truncate, so cost
of truncate needs to be changed but looks like it got changed only in SSE2 table
Whereas this change is also applicable for SSE4.1, so the cost of truncate needs
to be changed for that as well. Cost of “TRUNCATE v16i32 to v16i8” & “TRUNCATE
v16i16 to v16i8” should be same in SSE4.1 & SSE2 table. Removing their cost from
SSE4.1, so it will fall back to SSE2.

Reviewers: Simon Pilgrim
llvm-svn: 267123

8 years ago[Sparc] This provides support for itineraries on Sparc.
Chris Dewhurst [Fri, 22 Apr 2016 08:17:17 +0000 (08:17 +0000)]
[Sparc] This provides support for itineraries on Sparc.

Specifically, itineraries for LEON processors has been added, along with several LEON processor Subtargets. Although currently all these targets are pretty much identical, support for features that will differ among these processors will be added in the very near future.

The different Instruction Itinerary Classes (IICs) added are sufficient to differentiate between the instruction timings used by LEON and, quite probably, by generic Sparc processors too, but the focus of the exercise has been for LEON processors, as the requirement of my project. If the IICs are not sufficient for other Sparc processor types and you want to add a new itinerary for one of those, it should be relatively trivial to adapt this.

As none of the LEON processors has Quad Floats, or is a Version 9 processor, none of those instructions have itinerary classes defined and revert to the default "NoItinerary" instruction itinerary.

Phabricator Review: http://reviews.llvm.org/D19359

llvm-svn: 267121

8 years agoThe following code would not work before this patch, due to the inability to take...
Chris Dewhurst [Fri, 22 Apr 2016 08:13:47 +0000 (08:13 +0000)]
The following code would not work before this patch, due to the inability to take the address of a global object:

void func1() {

...
}

int main(int argc, char** argv) {

void (*pFunc)();
pFunc = &func1
pFunc();
...
}

Phabricator review: http://reviews.llvm.org/D19368

llvm-svn: 267120

8 years agoAdd missing include of <algorithm>
Marshall Clow [Fri, 22 Apr 2016 07:39:05 +0000 (07:39 +0000)]
Add missing include of <algorithm>

llvm-svn: 267119

8 years ago[index] Add SymbolSubKinds for ObjC IB annotations.
Argyrios Kyrtzidis [Fri, 22 Apr 2016 07:21:16 +0000 (07:21 +0000)]
[index] Add SymbolSubKinds for ObjC IB annotations.

llvm-svn: 267118

8 years ago[index] Add a SymbolSubKind for an ObjC unit test.
Argyrios Kyrtzidis [Fri, 22 Apr 2016 07:21:10 +0000 (07:21 +0000)]
[index] Add a SymbolSubKind for an ObjC unit test.

llvm-svn: 267117

8 years ago[index] Change SymbolCXXTemplateKind to a 'SymbolSubKinds' bitset.
Argyrios Kyrtzidis [Fri, 22 Apr 2016 07:21:04 +0000 (07:21 +0000)]
[index] Change SymbolCXXTemplateKind to a 'SymbolSubKinds' bitset.

This provides a more general and flexible way to annotate special symbols.

llvm-svn: 267116

8 years agoRevert "Initial implementation of optimization bisect support."
Vedant Kumar [Fri, 22 Apr 2016 06:51:37 +0000 (06:51 +0000)]
Revert "Initial implementation of optimization bisect support."

This reverts commit r267022, due to an ASan failure:

  http://lab.llvm.org:8080/green/job/clang-stage2-cmake-RgSan_check/1549

llvm-svn: 267115

8 years ago[mips][microMIPS] Implement DVP, EVP and JALRC.HB instructions
Zlatko Buljan [Fri, 22 Apr 2016 06:44:34 +0000 (06:44 +0000)]
[mips][microMIPS] Implement DVP, EVP and JALRC.HB instructions
Differential Revision: http://reviews.llvm.org/D18687

llvm-svn: 267114

8 years ago[GVN] Respect fast-math-flags on fcmps
David Majnemer [Fri, 22 Apr 2016 06:37:51 +0000 (06:37 +0000)]
[GVN] Respect fast-math-flags on fcmps

We assumed that flags were only present on binary operators.  This is
not true, they may also be present on calls and fcmps.

llvm-svn: 267113

8 years agoFix some spelling mistakes
David Majnemer [Fri, 22 Apr 2016 06:37:48 +0000 (06:37 +0000)]
Fix some spelling mistakes

llvm-svn: 267112

8 years ago[EarlyCSE] Take the intersection of flags on instructions
David Majnemer [Fri, 22 Apr 2016 06:37:45 +0000 (06:37 +0000)]
[EarlyCSE] Take the intersection of flags on instructions

EarlyCSE had inconsistent behavior with regards to flag'd instructions:
- In some cases, it would pessimize if the available instruction had
  different flags by not performing CSE.
- In other cases, it would miscompile if it replaced an instruction
  which had no flags with an instruction which has flags.

Fix this by being more consistent with our flag handling by utilizing
andIRFlags.

llvm-svn: 267111

8 years ago[SCEV] Extract out a `isSCEVExprNeverPoison` helper; NFCI
Sanjoy Das [Fri, 22 Apr 2016 05:38:54 +0000 (05:38 +0000)]
[SCEV] Extract out a `isSCEVExprNeverPoison` helper; NFCI

Summary:
Also adds a small comment blurb on control flow + no-wrap flags, since
that question came up a few days back on llvm-dev.

Reviewers: bjarke.roune, broune

Subscribers: sanjoy, mcrosier, llvm-commits, mzolotukhin

Differential Revision: http://reviews.llvm.org/D19209

llvm-svn: 267110

8 years ago[SystemZ] Mark CTTZ_ZERO_UNDEF/CTLZ_ZERO_UNDEF as Expand instead of Custom since...
Craig Topper [Fri, 22 Apr 2016 05:29:58 +0000 (05:29 +0000)]
[SystemZ] Mark CTTZ_ZERO_UNDEF/CTLZ_ZERO_UNDEF as Expand instead of Custom since the custom logic just did what Expand does when CTTZ/CTLZ are Legal. NFC

llvm-svn: 267109

8 years ago[Lanai] Set CTLZ_ZERO_UNDEF/CTTZ_ZERO_UNDEF to Expand instead of Legal so they will...
Craig Topper [Fri, 22 Apr 2016 05:13:01 +0000 (05:13 +0000)]
[Lanai] Set CTLZ_ZERO_UNDEF/CTTZ_ZERO_UNDEF to Expand instead of Legal so they will be converted to CTLZ/CTTZ by LegalizeDAG. Remove extra instructions that only existed to to contain patterns that match the zero_undef operations. NFC

llvm-svn: 267108

8 years ago[Lanai] Remove unused methods declarations. NFC
Craig Topper [Fri, 22 Apr 2016 05:12:57 +0000 (05:12 +0000)]
[Lanai] Remove unused methods declarations. NFC

llvm-svn: 267107

8 years agoClean the API for CollectAsmUndefinedRefs, taking a Triple and a String InlineAsm...
Mehdi Amini [Fri, 22 Apr 2016 04:58:12 +0000 (04:58 +0000)]
Clean the API for CollectAsmUndefinedRefs, taking a Triple and a String InlineAsm instead of a Module (NFC)

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267106

8 years agoIRObjectFile, clang-format fixup for r267104
Mehdi Amini [Fri, 22 Apr 2016 04:49:46 +0000 (04:49 +0000)]
IRObjectFile, clang-format fixup for r267104

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267105

8 years agoUse std::move on the enum to insert it into the pair to please MSVC
Mehdi Amini [Fri, 22 Apr 2016 04:45:57 +0000 (04:45 +0000)]
Use std::move on the enum to insert it into the pair to please MSVC

(I have no idea why is it needed)
Fixup for r267103

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267104

8 years agoRefactor IRObjectFile, extract a static CollectAsmUndefinedRefs() method to parse...
Mehdi Amini [Fri, 22 Apr 2016 04:28:05 +0000 (04:28 +0000)]
Refactor IRObjectFile, extract a static CollectAsmUndefinedRefs() method to parse inline assembly (NFC)

I plan to call this from ThinLTOCodeGenerator.

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 267103

8 years agoAMDGPU/SI: add llvm.amdgcn.ps.live intrinsic
Nicolai Haehnle [Fri, 22 Apr 2016 04:04:08 +0000 (04:04 +0000)]
AMDGPU/SI: add llvm.amdgcn.ps.live intrinsic

Summary:
This intrinsic returns true if the current thread belongs to a live pixel
and false if it belongs to a pixel that we are executing only for derivative
computation. It will be used by Mesa to implement gl_HelperInvocation.

Note that for pixels that are killed during the shader, this implementation
also returns true, but it doesn't matter because those pixels are always
disabled in the EXEC mask.

This unearthed a corner case in the instruction verifier, which complained
about a v_cndmask 0, 1, exec, exec<imp-use> instruction. That's stupid but
correct code, so make the verifier accept it as such.

Reviewers: arsenm, tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D19191

llvm-svn: 267102

8 years ago[OPENMP] Fix for LCV in simd directives in explicit clauses.
Alexey Bataev [Fri, 22 Apr 2016 03:56:56 +0000 (03:56 +0000)]
[OPENMP] Fix for LCV in simd directives in explicit clauses.

If loop control variable for simd-based directives is explicitly marked
as linear/lastprivate in clauses, codegen for such construct would
crash. Patch fixes this problem.

llvm-svn: 267101

8 years ago[AVX512] Teach lowering to use vplzcntd/q to implement 128/256-bit CTTZ_ZERO_UNDEF...
Craig Topper [Fri, 22 Apr 2016 03:22:38 +0000 (03:22 +0000)]
[AVX512] Teach lowering to use vplzcntd/q to implement 128/256-bit CTTZ_ZERO_UNDEF even without VLX support. We can just extend to 512-bits and extract like we do for CTLZ.

llvm-svn: 267100

8 years agoValueMapper/Enumerator: Clean up code in post-order traversals, NFC
Duncan P. N. Exon Smith [Fri, 22 Apr 2016 02:33:06 +0000 (02:33 +0000)]
ValueMapper/Enumerator: Clean up code in post-order traversals, NFC

Re-layer the functions in the new (i.e., newly correct) post-order
traversals in ValueEnumerator (r266947) and ValueMapper (r266949).
Instead of adding a node to the worklist in a helper function and
returning a flag to say what happened, return the node itself.  This
makes the code way cleaner: the worklist is local to the main function,
there is no flag for an early loop exit (since we can cleanly bury the
loop), and it's perfectly clear when pointers into the worklist might be
invalidated.

I'm fixing both algorithms in the same commit to avoid repeating the
commit message; if you take the time to understand one the other should
be easy.  The diff itself isn't entirely obvious since the traversals
have some noise (i.e., things to do), but here's the high-level change:

    auto helper = [&WL](T *Op) {     auto helper = [](T **&I, T **E) {
                                 =>    while (I != E) {
      if (shouldVisit(Op)) {             T *Op = *I++;
        WL.push(Op, Op->begin());        if (shouldVisit(Op)) {
        return true;                       return Op;
      }                                }
      return false;                    return nullptr;
    };                               };
                                 =>
    WL.push(S, S->begin());          WL.push(S, S->begin());
    while (!empty()) {               while (!empty()) {
      auto *N = WL.top().N;            auto *N = WL.top().N;
      auto *&I = WL.top().I;           auto *&I = WL.top().I;
      bool DidChange = false;
      while (I != N->end())
        if (helper(*I++)) {      =>    if (T *Op = helper(I, N->end()) {
          DidChange = true;              WL.push(Op, Op->begin());
          break;                         continue;
        }                              }
      if (DidChange)
        continue;

      POT.push(WL.pop());        =>    POT.push(WL.pop());
    }                                }

Thanks to Mehdi for helping me find a better way to layer this.

llvm-svn: 267099

8 years ago[MachineCombiner] Support for floating-point FMA on ARM64
Gerolf Hoflehner [Fri, 22 Apr 2016 02:15:19 +0000 (02:15 +0000)]
[MachineCombiner] Support for floating-point FMA on ARM64

Evaluates fmul+fadd -> fmadd combines and similar code sequences in the
machine combiner. It adds support for float and double similar to the existing
integer implementation. The key features are:

- DAGCombiner checks whether it should combine greedily or let the machine
combiner do the evaluation. This is only supported on ARM64.
- It gives preference to throughput over latency: the heuristic used is
to combine always in loops. The targets decides whether the machine
combiner should optimize for throughput or latency.
- Supports for fmadd, f(n)msub, fmla, fmls patterns
- On by default at O3 ffast-math

llvm-svn: 267098

8 years ago[ThinLTO] Remove unused/incomplete lazy summary reading support (NFC)
Teresa Johnson [Fri, 22 Apr 2016 01:52:00 +0000 (01:52 +0000)]
[ThinLTO] Remove unused/incomplete lazy summary reading support (NFC)

This removes the interfaces added (and not yet complete) to support
lazy reading of summaries. This support is not expected to be needed
since we are moving to a model where the full index is only being
traversed in the thin link step, instead of the back ends.

(The second part of this that I plan to do next is remove the
GlobalValueInfo from the ModuleSummaryIndex - it was mostly needed to
support lazy parsing of summaries. The index can instead reference the
summary structures directly.)

llvm-svn: 267097

8 years agoUntabify.
NAKAMURA Takumi [Fri, 22 Apr 2016 01:33:50 +0000 (01:33 +0000)]
Untabify.

llvm-svn: 267096

8 years agoTry to fix UNRESOLVED: LLVM :: CodeGen/AArch64/arm64-regress-opt-cmp.s on bots.
Nico Weber [Fri, 22 Apr 2016 01:08:56 +0000 (01:08 +0000)]
Try to fix UNRESOLVED: LLVM :: CodeGen/AArch64/arm64-regress-opt-cmp.s on bots.

This test used to write a .s file until r266971 fixed that.  But on most bots,
the .s file still exists.  Add an rm statement to clean up the bots.  In a few
days, this statement can go away again.

llvm-svn: 267095