platform/upstream/llvm.git
8 years agoExtract out a SelectionDAGBuilder::LowerAsStatepoint; NFC
Sanjoy Das [Wed, 16 Mar 2016 23:08:00 +0000 (23:08 +0000)]
Extract out a SelectionDAGBuilder::LowerAsStatepoint; NFC

Summary:
This is a step towards implementing "direct" lowering of calls and
invokes with deopt operand bundles into STATEPOINT nodes (as opposed to
having them mandatorily pass through RewriteStatepointsForGC, which is
the case today).

This change extracts out a `SelectionDAGBuilder::LowerAsStatepoint`
helper function that is able to lower a "statepoint like thing", and
uses it to lower `gc.statepoint` calls.  This is an NFC now, but in a
later change we will use `LowerAsStatepoint` to directly lower calls and
invokes with operand bundles without going through an intermediate
`gc.statepoint` IR representation.

FYI: I expect `SelectionDAGBuilder::StatepointInfo` will evolve as I add
support for lowering non gc.statepoints, right now it is fairly tightly
coupled with an IR level `gc.statepoint`.

Reviewers: reames, pgavlin, JosephTremoulet

Subscribers: sanjoy, mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D18106

llvm-svn: 263671

8 years agoDon't pass SymbolBody to the relax* functions.
Rafael Espindola [Wed, 16 Mar 2016 23:01:56 +0000 (23:01 +0000)]
Don't pass SymbolBody to the relax* functions.

We already pass the symbol value.

llvm-svn: 263670

8 years agoDon't create a got entry if we optimize it out.
Rafael Espindola [Wed, 16 Mar 2016 22:43:36 +0000 (22:43 +0000)]
Don't create a got entry if we optimize it out.

llvm-svn: 263669

8 years agouse FileCheck for tighter checking
Sanjay Patel [Wed, 16 Mar 2016 22:34:57 +0000 (22:34 +0000)]
use FileCheck for tighter checking

I'm testing out a script that auto-generates the check lines.
It's 98% copied from utils/update_llc_test_checks.py.
If others think this is useful, please let me know.

llvm-svn: 263668

8 years agouse FileCheck for tighter checking
Sanjay Patel [Wed, 16 Mar 2016 22:29:07 +0000 (22:29 +0000)]
use FileCheck for tighter checking

I'm testing out a script that auto-generates the check lines.
It's 98% copied from utils/update_llc_test_checks.py.
If others think this is useful, please let me know.

llvm-svn: 263667

8 years agoVariable name cleanup /NFC
Xinliang David Li [Wed, 16 Mar 2016 22:13:41 +0000 (22:13 +0000)]
Variable name cleanup /NFC

llvm-svn: 263666

8 years agoTweak some atomics functions in preparation for larger changes; NFC.
James Y Knight [Wed, 16 Mar 2016 22:12:04 +0000 (22:12 +0000)]
Tweak some atomics functions in preparation for larger changes; NFC.

- Rename getATOMIC to getSYNC, as llvm will soon be able to emit both
  '__sync' libcalls and '__atomic' libcalls, and this function is for
  the '__sync' ones.

- getInsertFencesForAtomic() has been replaced with
  shouldInsertFencesForAtomic(Instruction), so that the decision can be
  made per-instruction. This functionality will be used soon.

- emitLeadingFence/emitTrailingFence are no longer called if
  shouldInsertFencesForAtomic returns false, and thus don't need to
  check the condition themselves.

llvm-svn: 263665

8 years ago[ELF] SHF_MERGE section with 0 entsize is not fatal
Ed Maste [Wed, 16 Mar 2016 21:12:53 +0000 (21:12 +0000)]
[ELF] SHF_MERGE section with 0 entsize is not fatal

For now just treat such sections as non-mergeable.

Resubmit r263660 with test fix.

Differential Revision:     http://reviews.llvm.org/D18225

llvm-svn: 263664

8 years ago[SelectionDAG] Extract out populateCallLoweringInfo; NFC
Sanjoy Das [Wed, 16 Mar 2016 20:49:31 +0000 (20:49 +0000)]
[SelectionDAG] Extract out populateCallLoweringInfo; NFC

SelectionDAGBuilder::populateCallLoweringInfo is now used instead of
SelectionDAGBuilder::lowerCallOperands.  The populateCallLoweringInfo
interface is more composable in face of design changes like
http://reviews.llvm.org/D18106

llvm-svn: 263663

8 years ago[ProfileData] Make a utility method public, NFC
Vedant Kumar [Wed, 16 Mar 2016 20:49:26 +0000 (20:49 +0000)]
[ProfileData] Make a utility method public, NFC

The swift frontend needs to be able to look up PGO function name
variables based on the original raw function name. That's because it's
not possible to create PGO function name variables while emitting swift
IR. Instead, we have to create the name variables while lowering swift
IR to llvm IR, at which point we fix up all calls to the increment
intrinsic to point to the right name variable.

llvm-svn: 263662

8 years agoRevert r263660 as one of the tests is broken.
Ed Maste [Wed, 16 Mar 2016 20:42:27 +0000 (20:42 +0000)]
Revert r263660 as one of the tests is broken.

llvm-svn: 263661

8 years ago[ELF] SHF_MERGE section with 0 entsize is not fatal
Ed Maste [Wed, 16 Mar 2016 20:32:32 +0000 (20:32 +0000)]
[ELF] SHF_MERGE section with 0 entsize is not fatal

For now just treat such sections as non-mergeable.

Differential Revision: http://reviews.llvm.org/D18222

llvm-svn: 263660

8 years agoAdd __unconstref for future use
Eric Fiselier [Wed, 16 Mar 2016 20:32:07 +0000 (20:32 +0000)]
Add __unconstref for future use

llvm-svn: 263659

8 years agoAMDGPU: Prevent uniform loops from becoming infinite
Nicolai Haehnle [Wed, 16 Mar 2016 20:14:33 +0000 (20:14 +0000)]
AMDGPU: Prevent uniform loops from becoming infinite

Summary:
Uniform loops where the branch leaving the loop is predicated on VCCNZ
must be skipped if EXEC = 0, otherwise they will be infinite.

Reviewers: tstellarAMD, arsenm

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D18137

llvm-svn: 263658

8 years ago[Hexagon] Adding missing break in switch statement. Extra operands would have been...
Colin LeMahieu [Wed, 16 Mar 2016 20:00:38 +0000 (20:00 +0000)]
[Hexagon] Adding missing break in switch statement.  Extra operands would have been appended to the end.

llvm-svn: 263657

8 years ago[SLP] Make DataLayout a member variable.
Chad Rosier [Wed, 16 Mar 2016 19:48:42 +0000 (19:48 +0000)]
[SLP] Make DataLayout a member variable.

llvm-svn: 263656

8 years agoRevert "[LSR] Create fewer redundant instructions."
Geoff Berry [Wed, 16 Mar 2016 19:21:47 +0000 (19:21 +0000)]
Revert "[LSR] Create fewer redundant instructions."

This reverts commit r263644.  Investigating bootstrap failures.

llvm-svn: 263655

8 years ago[OPENMP] Support for codegen of private clause of target, host side
Carlo Bertolli [Wed, 16 Mar 2016 19:04:22 +0000 (19:04 +0000)]
[OPENMP] Support for codegen of private clause of target, host side

This patch adds support for codegen of private clause of target and a regression test for host code generation, when the host is used as target device. I believe that code generation for nvptx backend would not require anything additional or different to what is done for the host.

http://reviews.llvm.org/D18105

llvm-svn: 263654

8 years agoSplit relaxTls into one per relaxation type.
Rafael Espindola [Wed, 16 Mar 2016 19:03:58 +0000 (19:03 +0000)]
Split relaxTls into one per relaxation type.

This reduces code duplication in each target.

llvm-svn: 263653

8 years agoAdd an optional string argument to DeprecatedAttr for Fix-It.
Manman Ren [Wed, 16 Mar 2016 18:50:49 +0000 (18:50 +0000)]
Add an optional string argument to DeprecatedAttr for Fix-It.

We only add this to __attribute__((deprecated)).

Differential Revision: http://reviews.llvm.org/D17865

llvm-svn: 263652

8 years ago[CMake] Fix Windows build problem for CMake versions < 3.3
Jonathan Peyton [Wed, 16 Mar 2016 18:44:18 +0000 (18:44 +0000)]
[CMake] Fix Windows build problem for CMake versions < 3.3

Building libomp using CMake versions < 3.3 caused a link time error.  These
errors occurred because when assembling z_Windows_NT-586_asm.asm, the
definitions: OMPT_SUPPORT, _M_AMD64|_M_IA32 weren't defined on the command line.
To fix the problem, the COMPILE_FLAGS property for the assembly file is appended
to instead of the COMPILE_DEFINITIONS property being set.  For whatever reason, the
COMPILE_DEFINITIONS property doesn't pick up the definitions for assembly files
for the older CMake versions.

llvm-svn: 263651

8 years agoRemoved trailing whitespace
Simon Pilgrim [Wed, 16 Mar 2016 18:37:44 +0000 (18:37 +0000)]
Removed trailing whitespace

llvm-svn: 263650

8 years ago[X86] Reduced alignment of widened vector load/stores to better match PR26953 cases
Simon Pilgrim [Wed, 16 Mar 2016 18:32:44 +0000 (18:32 +0000)]
[X86] Reduced alignment of widened vector load/stores to better match PR26953 cases

llvm-svn: 263649

8 years agoadd checks for 'unpredictable' metadata preservation
Sanjay Patel [Wed, 16 Mar 2016 18:15:34 +0000 (18:15 +0000)]
add checks for 'unpredictable' metadata preservation

llvm-svn: 263648

8 years agoAdd attributes for preserve_mostcc/preserve_allcc calling conventions to the C/C...
Roman Levenstein [Wed, 16 Mar 2016 18:00:46 +0000 (18:00 +0000)]
Add attributes for preserve_mostcc/preserve_allcc calling conventions to the C/C++ front-end

Till now, preserve_mostcc/preserve_allcc calling convention attributes were only
available at the LLVM IR level. This patch adds attributes for
preserve_mostcc/preserve_allcc calling conventions to the C/C++ front-end.

The code was mostly written by Juergen Ributzka.
I just added support for the AArch64 target and tests.

Differential Revision: http://reviews.llvm.org/D18025

llvm-svn: 263647

8 years agofix function names; NFC
Sanjay Patel [Wed, 16 Mar 2016 18:00:09 +0000 (18:00 +0000)]
fix function names; NFC

llvm-svn: 263646

8 years ago[msan] Add a comment with a bug link.
Evgeniy Stepanov [Wed, 16 Mar 2016 17:39:17 +0000 (17:39 +0000)]
[msan] Add a comment with a bug link.

llvm-svn: 263645

8 years ago[LSR] Create fewer redundant instructions.
Geoff Berry [Wed, 16 Mar 2016 17:29:49 +0000 (17:29 +0000)]
[LSR] Create fewer redundant instructions.

Summary:
Fix LSRInstance::HoistInsertPosition() to check the original insert
position block first for a canonical insertion point that is dominated
by all inputs.  This leads to SCEV being able to reuse more instructions
since it currently tracks the instructions it creates for reuse by
keeping a table of <Value, insert point> pairs.

Reviewers: atrick

Subscribers: mcrosier, mzolotukhin, llvm-commits

Differential Revision: http://reviews.llvm.org/D18001

llvm-svn: 263644

8 years ago[tsan] Detect uses of uninitialized, destroyed and invalid mutexes
Kuba Brecka [Wed, 16 Mar 2016 15:39:20 +0000 (15:39 +0000)]
[tsan] Detect uses of uninitialized, destroyed and invalid mutexes

This patch adds a new TSan report type, ReportTypeMutexInvalidAccess, which is triggered when pthread_mutex_lock or pthread_mutex_unlock returns EINVAL (this means the mutex is invalid, uninitialized or already destroyed).

Differential Revision: http://reviews.llvm.org/D18132

llvm-svn: 263641

8 years ago[X86] Regenerated + extended widened vector conversion tests
Simon Pilgrim [Wed, 16 Mar 2016 15:33:43 +0000 (15:33 +0000)]
[X86] Regenerated + extended widened vector conversion tests
- Ensure we test X86 + X64
- sitopfp / uitofp requires testing for SSE2 and SSE42 as well (part of the fix for PR26953)

llvm-svn: 263640

8 years agoFix destructor definition of invalid classes
Olivier Goffart [Wed, 16 Mar 2016 14:36:11 +0000 (14:36 +0000)]
Fix destructor definition of invalid classes

The declaration of the destructor of an invalid class was not properly marked
as noexcept. As a result, the definition of the same destructor, which was
properly implicitly marked as noexcept, would not match the definition.
This would cause the definition CXXDestructorDecl to be matked as invalid
and omited from the AST.

Differential Revision: http://reviews.llvm.org/D17988

llvm-svn: 263639

8 years agoAdd visualizers for more clang types. Create more C++-like visualizations for existin...
Mike Spertus [Wed, 16 Mar 2016 14:27:50 +0000 (14:27 +0000)]
Add visualizers for more clang types. Create more C++-like visualizations for existing Clang types

Created visualizer for PointerType, LValueReferenceType, RValueReferenceType, and TemplateParmType.
In addition, cleaned up the display of existing types to be more C++-like. For example, instead of

  SubstTemplateTypeParmType: {Identifier (("T"))} => Record (25), {Identifier (("A"))}

it now displays more readably as

  SubstTemplateTypeParmType: {typename T <= struct A}

The <expand> sections still can be used for all the gory details if necessary.

llvm-svn: 263638

8 years agoFix null pointer "dereference" in DomainSocket
Pavel Labath [Wed, 16 Mar 2016 14:03:20 +0000 (14:03 +0000)]
Fix null pointer "dereference" in DomainSocket

offsetof is the official way to get the offset of a field in a structure.

llvm-svn: 263637

8 years agoRevert r263617, "Reapply: [VFS] Add support for handling path traversals"
NAKAMURA Takumi [Wed, 16 Mar 2016 12:15:29 +0000 (12:15 +0000)]
Revert r263617, "Reapply: [VFS] Add support for handling path traversals"

It broke standalone clang build.

llvm-svn: 263636

8 years agoDirectly get the canonical Type instead of going around through a CanQualType tempora...
Yaron Keren [Wed, 16 Mar 2016 12:14:43 +0000 (12:14 +0000)]
Directly get the canonical Type instead of going around through a CanQualType temporary, NFC.

llvm-svn: 263635

8 years ago[modules] Fix adding a templated friend functions to a namespace from another module.
Vassil Vassilev [Wed, 16 Mar 2016 11:17:04 +0000 (11:17 +0000)]
[modules] Fix adding a templated friend functions to a namespace from another module.

When clang adds argument dependent lookup candidates, it can perform template
instantiation. For example, it can instantiate a templated friend function and
register it in the enclosing namespace's lookup table.

Fixes https://llvm.org/bugs/show_bug.cgi?id=24954

Reviewed by Richard Smith.

llvm-svn: 263634

8 years agoSwitch from unittest2.expectedFailure to our own decorator on TestSTL
Pavel Labath [Wed, 16 Mar 2016 10:39:33 +0000 (10:39 +0000)]
Switch from unittest2.expectedFailure to our own decorator on TestSTL

the main reason is that our decorator contains extra fluff to "expect" crashes (which seem to
happen occasionaly on the android buildbot).

llvm-svn: 263633

8 years agoAdd more ARM Cortex-R8 regression tests to Clang.
Pablo Barrio [Wed, 16 Mar 2016 10:21:04 +0000 (10:21 +0000)]
Add more ARM Cortex-R8 regression tests to Clang.

Summary:
This patch adds Clang tests for Cortex-R8 related to FP capabilities and
hardware integer divide.

Reviewers: rengolin, bsmith

Subscribers: aemerson, cfe-commits, rengolin

Differential Revision: http://reviews.llvm.org/D18193

llvm-svn: 263632

8 years agoMark an LLGS test as flaky
Pavel Labath [Wed, 16 Mar 2016 09:58:34 +0000 (09:58 +0000)]
Mark an LLGS test as flaky

cause: Async output arrival over pty
llvm-svn: 263631

8 years agoAvoid using LookupResult's implicit copy ctor and assignment operator to avoid warnings
Marina Yatsina [Wed, 16 Mar 2016 09:56:58 +0000 (09:56 +0000)]
Avoid using LookupResult's implicit copy ctor and assignment operator to avoid warnings

The purpose of this patch is to keep the same functionality without using LookupResult's implicit copy ctor and assignment operator, because they cause warnings when -Wdeprecated is passed.
This patch is meant to help the following review: http://reviews.llvm.org/D18123.
The functionality is covered by the tests in my original commit (255890)
The test case in this patch was added to test a bug caught in the review of the first version of this fix.

Differential Revision: http://reviews.llvm.org/D18175

llvm-svn: 263630

8 years ago[test] Persist packets between expect_gdbremote_sequence invocations
Pavel Labath [Wed, 16 Mar 2016 09:44:49 +0000 (09:44 +0000)]
[test] Persist packets between expect_gdbremote_sequence invocations

Summary:
Some tests (Hc_then_Csignal_signals_correct_thread, at least) were sending a "continue" packet in
one expect_gdbremote_sequence invocation, and "expecting" the stop-reply in another call. This
posed a problem, because the were packets were not persisted between the two invocations, and if
the stub was exceptionally fast to respond, the packet would be received in the first invocation
(where it would be ignored) and then the second invocation would fail because it could not find
the packet.

Since doing matching in two invocations seems like a reasonable use of the packet pump, instead
of fixing the test, I make sure the packet_pump supports this usage by making the list of
captured packets persistent.

Reviewers: tfiala

Subscribers: lldb-commits

Differential Revision: http://reviews.llvm.org/D18140

llvm-svn: 263629

8 years agoFix thread/process ID reading from linux core files
Pavel Labath [Wed, 16 Mar 2016 09:19:57 +0000 (09:19 +0000)]
Fix thread/process ID reading from linux core files

Summary:
This also adds a basic smoke test for linux core file reading. I'm checking in the core files as
well, so that the tests can run on all platforms. With some tricks I was able to produce
reasonably-sized core files (~40K).

This fixes the first part of pr26322.

Reviewers: zturner

Subscribers: lldb-commits

Differential Revision: http://reviews.llvm.org/D18176

llvm-svn: 263628

8 years agoAMDGPU: Verify instructions in non-debug builds as well
Michel Danzer [Wed, 16 Mar 2016 09:10:42 +0000 (09:10 +0000)]
AMDGPU: Verify instructions in non-debug builds as well

And emit an error if it fails.

This prevents illegal instructions from getting sent to the GPU, which
would potentially result in a hang.

This is a candidate for the stable branch(es).

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
llvm-svn: 263627

8 years agoAMDGPU/SI: Clean up indentation in SIInstrInfo::getDefaultRsrcDataFormat
Michel Danzer [Wed, 16 Mar 2016 09:10:35 +0000 (09:10 +0000)]
AMDGPU/SI: Clean up indentation in SIInstrInfo::getDefaultRsrcDataFormat

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 263626

8 years agoMove OperatingSystem plugins to SystemInitializerFull
Pavel Labath [Wed, 16 Mar 2016 08:48:56 +0000 (08:48 +0000)]
Move OperatingSystem plugins to SystemInitializerFull

Summary: These are not needed in lldb-server. Removing them shrinks the server size by about 1.5%.

Reviewers: zturner

Subscribers: lldb-commits

Differential Revision: http://reviews.llvm.org/D18188

llvm-svn: 263625

8 years agoAVX512BW: Fix SRA v64i8 lowering. Use PCMPGTM (cmp result in k register) for 512bit...
Igor Breger [Wed, 16 Mar 2016 08:48:26 +0000 (08:48 +0000)]
AVX512BW: Fix SRA v64i8 lowering. Use PCMPGTM (cmp result in k register) for 512bit vector because PCMPGT supported only for 128/256bit.

Differential Revision: http://reviews.llvm.org/D18204

llvm-svn: 263624

8 years agoRevert r263622, "clang/lib/Frontend/ModuleDependencyCollector.cpp: Use clang/Config...
NAKAMURA Takumi [Wed, 16 Mar 2016 08:44:37 +0000 (08:44 +0000)]
Revert r263622, "clang/lib/Frontend/ModuleDependencyCollector.cpp: Use clang/Config/config.h instead of llvm's."

I'll commit better fix(es) later. "llvm/Config/config.h" shouldn't be available here.

llvm-svn: 263623

8 years agoclang/lib/Frontend/ModuleDependencyCollector.cpp: Use clang/Config/config.h instead...
NAKAMURA Takumi [Wed, 16 Mar 2016 08:28:55 +0000 (08:28 +0000)]
clang/lib/Frontend/ModuleDependencyCollector.cpp: Use clang/Config/config.h instead of llvm's.

llvm-svn: 263622

8 years ago[ASAN] Add support for mips/mips64 android
Mohit K. Bhakkad [Wed, 16 Mar 2016 08:23:10 +0000 (08:23 +0000)]
[ASAN] Add support for mips/mips64 android

Patch by Duane Sand

Reviewers: samsonov

Subscribers: duanesand, jaydeep, sagar, llvm-commits, filcab.

Differential Revision: http://reviews.llvm.org/D17883

llvm-svn: 263621

8 years ago[Bitcode] Add compatibility test for the 3.8 release
Vedant Kumar [Wed, 16 Mar 2016 05:43:03 +0000 (05:43 +0000)]
[Bitcode] Add compatibility test for the 3.8 release

Fork off compatibility.ll for the 3.8 release. The *.bc file in this
commit was produced using a Release build of the release_38 branch.

llvm-svn: 263620

8 years ago[JumpThreading] See through Cast Instructions
Haicheng Wu [Wed, 16 Mar 2016 04:52:52 +0000 (04:52 +0000)]
[JumpThreading] See through Cast Instructions

To capture more jump-thread opportunity.

llvm-svn: 263618

8 years agoReapply: [VFS] Add support for handling path traversals
Bruno Cardoso Lopes [Wed, 16 Mar 2016 04:39:38 +0000 (04:39 +0000)]
Reapply: [VFS] Add support for handling path traversals

This is originally r261551, reverted because of windows bots failing on
unittests. Change the current behavior to do not handle path traversals
on windows.

Handle ".", ".." and "./" with trailing slashes while collecting files
to be dumped into the vfs overlay directory.

Include the support for symlinks into components. Given the path:

/install-dir/bin/../lib/clang/3.8.0/include/altivec.h, if "bin"
component is a symlink, it's not safe to use `path::remove_dots` here,
and `realpath` is used to get the right answer. Since `realpath`
is expensive, we only do it at collecting time (which only happens
during the crash reproducer) and cache the base directory for fast lookups.

Overall, this makes the input to the VFS YAML file to be canonicalized
to never contain traversal components.

Differential Revision: http://reviews.llvm.org/D17104

rdar://problem/24499339

llvm-svn: 263617

8 years agoAdd clang thread safety annotations to mutex and lock_guard. Patch by jamesr@google...
Eric Fiselier [Wed, 16 Mar 2016 02:30:06 +0000 (02:30 +0000)]
Add clang thread safety annotations to mutex and lock_guard. Patch by jamesr@google.com.

This adds clang thread safety annotations to std::mutex and
std::lock_guard so code using these types can use these types directly
instead of having to wrap the types to provide annotations. These checks
when enabled by -Wthread-safety provide simple but useful static
checking to detect potential race conditions.
See http://clang.llvm.org/docs/ThreadSafetyAnalysis.html for details.

This patch was reviewed in http://reviews.llvm.org/D14731.

llvm-svn: 263611

8 years ago[Support] Update Error unit test to remove implementation specific behaviour.
Lang Hames [Wed, 16 Mar 2016 01:20:54 +0000 (01:20 +0000)]
[Support] Update Error unit test to remove implementation specific behaviour.

llvm-svn: 263610

8 years ago[Support] Add the 'Error' class for structured error handling.
Lang Hames [Wed, 16 Mar 2016 01:02:46 +0000 (01:02 +0000)]
[Support] Add the 'Error' class for structured error handling.

This patch introduces the Error classs for lightweight, structured,
recoverable error handling. It includes utilities for creating, manipulating
and handling errors. The scheme is similar to exceptions, in that errors are
described with user-defined types. Unlike exceptions however, errors are
represented as ordinary return types in the API (similar to the way
std::error_code is used).

For usage notes see the LLVM programmer's manual, and the Error.h header.
Usage examples can be found in unittests/Support/ErrorTest.cpp.

Many thanks to David Blaikie, Mehdi Amini, Kevin Enderby and others on the
llvm-dev and llvm-commits lists for lots of discussion and review.

llvm-svn: 263609

8 years ago[X86] Regenerated widen load tests
Simon Pilgrim [Wed, 16 Mar 2016 00:41:21 +0000 (00:41 +0000)]
[X86] Regenerated widen load tests

llvm-svn: 263608

8 years agoConvert some ObjC msgSends to runtime calls.
Pete Cooper [Wed, 16 Mar 2016 00:33:21 +0000 (00:33 +0000)]
Convert some ObjC msgSends to runtime calls.

It is faster to directly call the ObjC runtime for methods such as retain/release instead of sending a message to those functions.

This patch adds support for converting messages to retain/release/alloc/autorelease to their equivalent runtime calls.

Tests included for the positive case of applying this transformation, negative tests that we ensure we only convert "alloc" to objc_alloc, not "alloc2", and also a driver test to ensure we enable this only for supported runtime versions.

Reviewed by John McCall.

Differential Revision: http://reviews.llvm.org/D14737

llvm-svn: 263607

8 years ago[X86][SSE41] Additional tests for extracting zeroable shuffle elements
Simon Pilgrim [Wed, 16 Mar 2016 00:13:36 +0000 (00:13 +0000)]
[X86][SSE41] Additional tests for extracting zeroable shuffle elements

We can currently only match zeroable vector elements of the same size as the shuffle type - these tests demonstrate the problem and a solution will be shortly added in an updated D14261

llvm-svn: 263606

8 years agoRevert "[JumpThreading] Simplify Instructions first in ComputeValueKnownInPredecessors()"
Haicheng Wu [Tue, 15 Mar 2016 23:38:47 +0000 (23:38 +0000)]
Revert "[JumpThreading] Simplify Instructions first in ComputeValueKnownInPredecessors()"

Not sure it handles undef properly.

llvm-svn: 263605

8 years agoOn some platforms, the compiler is allowed to assume that BOOL == bool. On others...
Enrico Granata [Tue, 15 Mar 2016 23:38:04 +0000 (23:38 +0000)]
On some platforms, the compiler is allowed to assume that BOOL == bool. On others, BOOL == signed char.
This can cause differences in which bit patterns end up meaning YES or NO. In general, however, 0 == NO and 1 == YES.

To keep it simple, LLDB will now show "YES" and "NO" only for 1 and 0 respectively, and format other values as the plain numeric value instead.

Fixes rdar://24809994

llvm-svn: 263604

8 years agoMake it so that the data formatter for NSError can see through a variable of type...
Enrico Granata [Tue, 15 Mar 2016 23:20:10 +0000 (23:20 +0000)]
Make it so that the data formatter for NSError can see through a variable of type NSError**. Fixes rdar://25060684

llvm-svn: 263603

8 years agoOnly try to load the OS plug-in after a shared library load if we don't already have...
Greg Clayton [Tue, 15 Mar 2016 22:44:07 +0000 (22:44 +0000)]
Only try to load the OS plug-in after a shared library load if we don't already have one.

<rdar://problem/24162686>

llvm-svn: 263602

8 years agoFix ClangASTContext::GetFunctionArgumentAtIndex() to not water down the type to the...
Greg Clayton [Tue, 15 Mar 2016 22:43:26 +0000 (22:43 +0000)]
Fix ClangASTContext::GetFunctionArgumentAtIndex() to not water down the type to the canonical type before handing the type out for the function type.

llvm-svn: 263601

8 years agoUnbreak buildbot.
Davide Italiano [Tue, 15 Mar 2016 22:38:21 +0000 (22:38 +0000)]
Unbreak buildbot.

It passed on a case-insensitive filesystem, but
it was broken on a case-sensitive one.

llvm-svn: 263598

8 years agollvm-bcanalyzer: Fix handling of attribute group entries
Justin Bogner [Tue, 15 Mar 2016 22:37:25 +0000 (22:37 +0000)]
llvm-bcanalyzer: Fix handling of attribute group entries

These were printing as "UnknownCode3", since we were looking for them
inside PARAMATTR blocks instead of PARAMATTR_GROUP blocks.

llvm-svn: 263597

8 years agoMyriad: define __myriad2 macro automatically
Douglas Katzman [Tue, 15 Mar 2016 22:34:02 +0000 (22:34 +0000)]
Myriad: define __myriad2 macro automatically

llvm-svn: 263596

8 years agoTurn LoopLoadElimination on again
Adam Nemet [Tue, 15 Mar 2016 22:26:12 +0000 (22:26 +0000)]
Turn LoopLoadElimination on again

The latent bug that LLE exposed in the LoopVectorizer was resolved
(PR26952).

The pass can be disabled with -mllvm -enable-loop-load-elim=0

llvm-svn: 263595

8 years ago[ELF] Ignore _GLOBAL_OFFSET_TABLE_ symbol in shared libs.
Davide Italiano [Tue, 15 Mar 2016 22:24:58 +0000 (22:24 +0000)]
[ELF] Ignore _GLOBAL_OFFSET_TABLE_ symbol in shared libs.

This fixes PR26705.

Differential Revision:  http://reviews.llvm.org/D18166

llvm-svn: 263594

8 years agoDon't crash if the TypeSP is empty.
Greg Clayton [Tue, 15 Mar 2016 21:58:28 +0000 (21:58 +0000)]
Don't crash if the TypeSP is empty.

llvm-svn: 263593

8 years agoImprove the 'type lookup' command such that it guesses to use the current's frame...
Enrico Granata [Tue, 15 Mar 2016 21:50:51 +0000 (21:50 +0000)]
Improve the 'type lookup' command such that it guesses to use the current's frame language as the one to start searching from.

llvm-svn: 263592

8 years ago[libfuzzer] speeding up corpus load
Mike Aizatsky [Tue, 15 Mar 2016 21:47:21 +0000 (21:47 +0000)]
[libfuzzer] speeding up corpus load

llvm-svn: 263591

8 years agoFix EHFrame processing to add implicit references when needed.
Pete Cooper [Tue, 15 Mar 2016 21:33:10 +0000 (21:33 +0000)]
Fix EHFrame processing to add implicit references when needed.

The current code for processCIE and processFDE returns out if it sees
any references.  The problem with this is that some references could be
explicit in the binary, while others are implicit as they can be
inferred from the content of the EHFrame itself.

This change walks the references we have against the references we
need, and verifies that all explicit references are in the correct place,
and generates any missing implicit ones.

Reviewed by Lang Hames and Nick Kledzik.

Differential Revision: http://reviews.llvm.org/D15439

llvm-svn: 263590

8 years agoRevert commit http://reviews.llvm.org/D17877 to fix tests on x86.
Arpith Chacko Jacob [Tue, 15 Mar 2016 21:26:34 +0000 (21:26 +0000)]
Revert commit reviews.llvm.org/D17877 to fix tests on x86.

llvm-svn: 263589

8 years agoRemove unnecessary <limits> includes.
Jim Ingham [Tue, 15 Mar 2016 21:11:02 +0000 (21:11 +0000)]
Remove unnecessary <limits> includes.

llvm-svn: 263588

8 years ago[OpenMP] Base support for target directive codegen on NVPTX device.
Arpith Chacko Jacob [Tue, 15 Mar 2016 21:04:57 +0000 (21:04 +0000)]
[OpenMP] Base support for target directive codegen on NVPTX device.

Summary:
This patch adds base support for codegen of the target directive on the NVPTX device.

Reviewers: ABataev

Differential Revision: http://reviews.llvm.org/D17877

llvm-svn: 263587

8 years agoFix spelling error in comment
Jonathan Peyton [Tue, 15 Mar 2016 20:59:10 +0000 (20:59 +0000)]
Fix spelling error in comment

llvm-svn: 263586

8 years agoAlso handle the new Rust pers fn to isCatchAll()
Bjorn Steinbrink [Tue, 15 Mar 2016 20:57:07 +0000 (20:57 +0000)]
Also handle the new Rust pers fn to isCatchAll()

llvm-svn: 263585

8 years agoMove the fixit for -Wformat-security to a note.
Bob Wilson [Tue, 15 Mar 2016 20:56:38 +0000 (20:56 +0000)]
Move the fixit for -Wformat-security to a note.

r263299 added a fixit for the -Wformat-security warning, but that runs
into complications with our guideline that error recovery should be done
as-if the fixit had been applied. Putting the fixit on a note avoids that.

llvm-svn: 263584

8 years ago[STATS] Print "Unknown" for frequency if it wasn't able to be parsed
Jonathan Peyton [Tue, 15 Mar 2016 20:55:32 +0000 (20:55 +0000)]
[STATS] Print "Unknown" for frequency if it wasn't able to be parsed

llvm-svn: 263583

8 years ago[STATS] Fix comments in kmp_stats.h
Jonathan Peyton [Tue, 15 Mar 2016 20:49:01 +0000 (20:49 +0000)]
[STATS] Fix comments in kmp_stats.h

llvm-svn: 263582

8 years agoAdd Rust's personality function to the list of known personality functions
Bjorn Steinbrink [Tue, 15 Mar 2016 20:35:45 +0000 (20:35 +0000)]
Add Rust's personality function to the list of known personality functions

Reviewers: majnemer

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D18192

llvm-svn: 263581

8 years ago[STATS] Add header information to stats print out
Jonathan Peyton [Tue, 15 Mar 2016 20:28:47 +0000 (20:28 +0000)]
[STATS] Add header information to stats print out

This change adds a header to the printout of the statistics which includes the
time, machine name, and processor info if available. This change also includes
some cosmetic changes like using enum casting for timer and counter iteration.

Differential Revision: http://reviews.llvm.org/D18153

llvm-svn: 263580

8 years ago[msan] Don't put module constructors in comdats.
Evgeniy Stepanov [Tue, 15 Mar 2016 20:25:47 +0000 (20:25 +0000)]
[msan] Don't put module constructors in comdats.

There is something strange going on with debug info (.eh_frame_hdr)
disappearing when msan.module_ctor are placed in comdat sections.

Moving this functionality under flag, disabled by default.

llvm-svn: 263579

8 years ago[cfi] Don't emit checks for disabled CFI kinds.
Evgeniy Stepanov [Tue, 15 Mar 2016 20:19:29 +0000 (20:19 +0000)]
[cfi] Don't emit checks for disabled CFI kinds.

In the cross-DSO CFI mode clang emits __cfi_check_fail that handles
errors triggered from other modules with targets in the current
module. With this change, __cfi_check_fail will handle errors for
CFI kinds that are not enabled in the current module as if they
have the trapping behaviour (-fsanitize-trap=...).

This fixes a bug where some combinations of -fsanitize* flags may
result in a link failure due to a missing sanitizer runtime library
for the diagnostic calls in __cfi_check_fail.

llvm-svn: 263578

8 years agoAdd missing error handling in llvm-lto
Mehdi Amini [Tue, 15 Mar 2016 20:17:55 +0000 (20:17 +0000)]
Add missing error handling in llvm-lto

Annoyingly, ErrorOr allows to *not check* the error when things go
well. It will crash badly when there is an error though. It should
runtime assert when it is used without being checked!

From: Mehdi Amini <mehdi.amini@apple.com>
llvm-svn: 263577

8 years ago[ThinLTO] Record all global variable defs in the summary
Teresa Johnson [Tue, 15 Mar 2016 19:35:45 +0000 (19:35 +0000)]
[ThinLTO] Record all global variable defs in the summary

Record all variable defs with a summary record to aid in building a
complete reference graph and locating constant variable defs to import.

llvm-svn: 263576

8 years ago[ASan] Make sure unittest works if sized-delete is enabled by default.
Alexey Samsonov [Tue, 15 Mar 2016 19:07:56 +0000 (19:07 +0000)]
[ASan] Make sure unittest works if sized-delete is enabled by default.

llvm-svn: 263575

8 years agoTest commit - Remove some trailing whitespace
Pablo Barrio [Tue, 15 Mar 2016 19:03:09 +0000 (19:03 +0000)]
Test commit - Remove some trailing whitespace

llvm-svn: 263574

8 years ago[MIR] Add a test case for the diagnostic of a wrongly typed generic instruction
Quentin Colombet [Tue, 15 Mar 2016 18:31:29 +0000 (18:31 +0000)]
[MIR] Add a test case for the diagnostic of a wrongly typed generic instruction

llvm-svn: 263573

8 years ago[AArch64] Move GlobalISel test cases into a GlobalISel subdirectory
Quentin Colombet [Tue, 15 Mar 2016 18:30:00 +0000 (18:30 +0000)]
[AArch64] Move GlobalISel test cases into a GlobalISel subdirectory

llvm-svn: 263572

8 years ago[CMake] Updating Apple build configurations
Chris Bieneman [Tue, 15 Mar 2016 18:27:28 +0000 (18:27 +0000)]
[CMake] Updating Apple build configurations

This updates Apple build configurations to adapt to r263566 & r263570, which added a PACKAGE_VENDOR variable.

llvm-svn: 263571

8 years ago[CMake] Defaulting CLANG_VENDOR to PACKAGE_VENDOR
Chris Bieneman [Tue, 15 Mar 2016 18:25:36 +0000 (18:25 +0000)]
[CMake] Defaulting CLANG_VENDOR to PACKAGE_VENDOR

LLVM r263566 adds a generic PACKAGE_VENDOR configuration which can be used to specify the vendor for LLVM toolchain tools. This change defaults the CLANG_VENDOR to the PACKAGE_VENDOR so that you don't have to specify both when building a package.

llvm-svn: 263570

8 years agoELF: Fix use-after-free problem.
Rui Ueyama [Tue, 15 Mar 2016 18:20:50 +0000 (18:20 +0000)]
ELF: Fix use-after-free problem.

Fixes pr26908. This patch is based on Filipe Cabecinhas'
patch (http://reviews.llvm.org/D18167)

http://reviews.llvm.org/D18169

llvm-svn: 263569

8 years agoFix a bot I broke.
Chris Bieneman [Tue, 15 Mar 2016 18:17:48 +0000 (18:17 +0000)]
Fix a bot I broke.

The builtin library isn't added by the driver unless it exists, so we shouldn't check for it. I've marked this as a FIXME, because we probably should have a way to test this.

llvm-svn: 263568

8 years ago[Driver] [Darwin] Fix linking libclang_rt.profile_*sim.a
Chris Bieneman [Tue, 15 Mar 2016 18:08:20 +0000 (18:08 +0000)]
[Driver] [Darwin] Fix linking libclang_rt.profile_*sim.a

Summary: isTarget*() calls are order-dependent. This is because iOS Sim *is* iOS. This means checks for the simulator version of the platform must always be ahead of checks for the embedded platform.

Reviewers: zaks.anna, bogner

Subscribers: cfe-commits

Differential Revision: http://reviews.llvm.org/D17947

llvm-svn: 263567

8 years ago[CMake] Add PACKAGE_VENDOR for customizing version output
Chris Bieneman [Tue, 15 Mar 2016 18:07:46 +0000 (18:07 +0000)]
[CMake] Add PACKAGE_VENDOR for customizing version output

Summary: This change adds a PACKAGE_VENDOR variable. When set it makes the version output more closely resemble the clang version output.

Reviewers: aprantl, bogner

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D18159

llvm-svn: 263566

8 years ago[LV] Preserve LoopInfo when store predication is used
Adam Nemet [Tue, 15 Mar 2016 18:06:20 +0000 (18:06 +0000)]
[LV] Preserve LoopInfo when store predication is used

This was a latent bug that got exposed by the change to add LoopSimplify
as a dependence to LoopLoadElimination.  Since LoopInfo was corrupted
after LV, LoopSimplify mis-compiled nbench in the test-suite (more
details in the PR).

The problem was that when we create the blocks for predicated stores we
didn't add those to any loops.

The original testcase for store predication provides coverage for this
assuming we verify LI on the way out of LV.

Fixes PR26952.

llvm-svn: 263565

8 years ago[MC] Rename TLSDESC as it's not ARM specific.
Davide Italiano [Tue, 15 Mar 2016 17:29:52 +0000 (17:29 +0000)]
[MC] Rename TLSDESC as it's not ARM specific.

Similarly to what was done for TLSCALL in r263515.

llvm-svn: 263564

8 years agoAMDGPU/SI: Implement GroupStaticSize Intrinsic for Dynamic LDS
Changpeng Fang [Tue, 15 Mar 2016 17:28:44 +0000 (17:28 +0000)]
AMDGPU/SI: Implement GroupStaticSize Intrinsic for Dynamic LDS

Summary:
  Static LDS size is saved in MachineFunctionInfo::LDSSize,
We define a pseudo instruction with usesCustomInserter bit set. Then, in EmitInstrWithCustomInserter,
we replace this pseudo instruction with a mov of MachineFunctionInfo::LDSSize.

Reviewers:
    arsenm
    tstellarAMD

Subscribers
    llvm-commits, arsenm

Differential Revision:
   http://reviews.llvm.org/D18064

llvm-svn: 263563

8 years ago[llvm-readobj] Impl GNU style printing of sections and relocations
Hemant Kulkarni [Tue, 15 Mar 2016 17:25:31 +0000 (17:25 +0000)]
[llvm-readobj] Impl GNU style printing of  sections and relocations

Differential Revision: http://reviews.llvm.org/D17523

llvm-svn: 263561