platform/upstream/llvm.git
5 years ago[HIP] Do not unbundle object files for -fno-gpu-rdc
Yaxun Liu [Tue, 5 Mar 2019 16:07:56 +0000 (16:07 +0000)]
[HIP] Do not unbundle object files for -fno-gpu-rdc

When -fno-gpu-rdc is set, device code is compiled, linked, and assembled into fat binary
and embedded as string in object files. The object files are normal object files which
can be linked by host linker. In the linking stage, the object files should not be unbundled
when -fno-gpu-rdc is set since they are normal object files, not bundles. The object files
only need to be unbundled when -fgpu-rdc is set.

Currently clang always unbundles object files, disregarding -fgpu-rdc option.

This patch fixes that.

Differential Revision: https://reviews.llvm.org/D58917

llvm-svn: 355410

5 years ago[libc++] Fix <atomic> failures on GCC
Louis Dionne [Tue, 5 Mar 2019 15:49:58 +0000 (15:49 +0000)]
[libc++] Fix <atomic> failures on GCC

Summary:
In https://reviews.llvm.org/D58201, we turned memory_order into an enum
class in C++20 mode. However, we were not casting memory_order to its
underlying type correctly for the GCC implementation, which broke the
build bots. I also fixed a test that was failing in C++17 mode on GCC 5.

Reviewers: EricWF, jfb, mclow.lists

Subscribers: zoecarver

Differential Revision: https://reviews.llvm.org/D58966

llvm-svn: 355409

5 years agoRevert "[GlobalISel][AArch64] Add selection support for G_EXTRACT_VECTOR_ELT"
Jessica Paquette [Tue, 5 Mar 2019 15:47:00 +0000 (15:47 +0000)]
Revert "[GlobalISel][AArch64] Add selection support for G_EXTRACT_VECTOR_ELT"

This broke test-suite::aarch64_neon_intrinsics.test

Reverting while I look into it.

Example failure:
http://lab.llvm.org:8011/builders/clang-cmake-aarch64-quick/builds/17740

llvm-svn: 355408

5 years ago[X86] Add SMULO/UMULO combine tests
Simon Pilgrim [Tue, 5 Mar 2019 15:36:45 +0000 (15:36 +0000)]
[X86] Add SMULO/UMULO combine tests

Include scalar and vector test variants covering the folds in DAGCombiner (vector isn't currently supported - PR40442)

llvm-svn: 355407

5 years agoRevert "Fix embedded Python initialization according to changes in version 3.7"
Tatyana Krasnukha [Tue, 5 Mar 2019 15:27:33 +0000 (15:27 +0000)]
Revert "Fix embedded Python initialization according to changes in version 3.7"

Testsuite hangs on Windows likely due to these changes.

llvm-svn: 355406

5 years agoFix typo in constant vector
Simon Pilgrim [Tue, 5 Mar 2019 15:06:01 +0000 (15:06 +0000)]
Fix typo in constant vector

llvm-svn: 355405

5 years ago[X86] Add SADDO/UADDO and SSUBO/USUBO combine tests
Simon Pilgrim [Tue, 5 Mar 2019 14:52:42 +0000 (14:52 +0000)]
[X86] Add SADDO/UADDO and SSUBO/USUBO combine tests

Include scalar and vector test variants covering the folds in DAGCombiner (vector isn't currently supported - PR40442)

llvm-svn: 355404

5 years ago[libc++] Change memory_order to an enum class
Louis Dionne [Tue, 5 Mar 2019 14:50:25 +0000 (14:50 +0000)]
[libc++] Change memory_order to an enum class

This implements P0439R0.

Thanks to Zoe Carver for the patch.
Differential Revision: https://reviews.llvm.org/D58201

llvm-svn: 355403

5 years agoAdds property to force enabling of GDB JIT loader for MacOS
Yury Delendik [Tue, 5 Mar 2019 14:23:53 +0000 (14:23 +0000)]
Adds property to force enabling of GDB JIT loader for MacOS

Summary:
Based on https://gist.github.com/thlorenz/30bf0a3f67b1d97b2945#patching-and-rebuilding

The functionality was disabled at https://github.com/llvm/llvm-project/commit/521c2278abb16f0148cef1bd061cadb01ef43192

Reviewers: jingham

Subscribers: lldb-commits

Tags: #lldb

Differential Revision: https://reviews.llvm.org/D57689

llvm-svn: 355402

5 years ago[clang-tidy] Fix bugprone-string-constructor crash
Alexander Kornienko [Tue, 5 Mar 2019 14:09:57 +0000 (14:09 +0000)]
[clang-tidy] Fix bugprone-string-constructor crash

llvm-svn: 355401

5 years ago[X86] Add test cases for D58874
Simon Pilgrim [Tue, 5 Mar 2019 13:52:09 +0000 (13:52 +0000)]
[X86] Add test cases for D58874

Add scalar and vector test cases for missing (add (add (xor a, -1), b), 1) -> (sub b, a) fold

llvm-svn: 355400

5 years ago[llvm-objcopy] - Simplify `isCompressable` and fix the issue relative.
George Rimar [Tue, 5 Mar 2019 13:07:43 +0000 (13:07 +0000)]
[llvm-objcopy] - Simplify `isCompressable` and fix the issue relative.

When --compress-debug-sections is given, llvm-objcopy do not compress
sections that have "ZLIB" header in data. Normally this signature is used
in zlib-gnu compression format. But if zlib-gnu used then the name of the compressed
section should start from .z* (e.g .zdebug_info). If it does not, then it is not
a zlib-gnu format and section should be treated as a normal uncompressed section.

Differential revision: https://reviews.llvm.org/D58908

llvm-svn: 355399

5 years agoOne more UserIDResolver fix
Pavel Labath [Tue, 5 Mar 2019 12:51:20 +0000 (12:51 +0000)]
One more UserIDResolver fix

The intention in r355323 has been to implement a no-op resolver in the
HostInfoBase class, which will then be shadowed a an implementation in
the HostInfoPosix class. However, I add the shadowing declaration in
HostInfoPosix.h, and instead had implemented the HostInfoBase function
in HostInfoPosix.cpp. This has lead to undefined symbols on windows, and
a subsequent implementation of a no-op resolver in HostInfoWindows
(r355329).

Since now there is no point on having a no-op resolver in the base
class, I just remove the base declaration altogether, and have
HostInfoPosix implement the (newly-declared) HostInfoPosix version of
that function.

llvm-svn: 355398

5 years ago[sanitizers] Don't use Windows Trace Logging with Clang in MSVC mode either
Martin Storsjo [Tue, 5 Mar 2019 12:50:06 +0000 (12:50 +0000)]
[sanitizers] Don't use Windows Trace Logging with Clang in MSVC mode either

The TraceLoggingProvider.h header does work with clang-cl in general
these days with Win SDK 10.0.17763.0, but when compiled in 32 bit x86
mode, with the -Z7 flag, compilation fails with the following error:

fatal error: error in backend: assembler label '' can not be undefined

With older Win SDKs, there are other build failures (regardless of
architecture or the -Z7 flag).

Differential Revision: https://reviews.llvm.org/D58958

llvm-svn: 355397

5 years ago[analyzer] Fix taint propagation in GenericTaintChecker
Kristof Umann [Tue, 5 Mar 2019 12:42:59 +0000 (12:42 +0000)]
[analyzer] Fix taint propagation in GenericTaintChecker

The gets function has no SrcArgs. Because the default value for isTainted was
false, it didn't mark its DstArgs as tainted.

Patch by Gábor Borsik!

Differential Revision: https://reviews.llvm.org/D58828

llvm-svn: 355396

5 years ago[ASTMatchers] Improved formatting in a documentation comment
Dmitri Gribenko [Tue, 5 Mar 2019 12:38:18 +0000 (12:38 +0000)]
[ASTMatchers] Improved formatting in a documentation comment

llvm-svn: 355395

5 years ago[AMDGPU] Fix DPP operand order in atomic optimizer
Carl Ritson [Tue, 5 Mar 2019 12:21:44 +0000 (12:21 +0000)]
[AMDGPU] Fix DPP operand order in atomic optimizer

Summary:
Ensure order of operands in DPP atomic optimizer final WWM step is appropriate for sub instructions.

Change-Id: I631d050e1c00a3b4bc7c11a90437064403c4cf30

Reviewers: sheredom, tpr

Reviewed By: sheredom

Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, t-tye, jfb, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58900

llvm-svn: 355394

5 years ago[SCEV] Ensure that isHighCostExpansion takes into account what is being divided
David Green [Tue, 5 Mar 2019 12:12:18 +0000 (12:12 +0000)]
[SCEV] Ensure that isHighCostExpansion takes into account what is being divided

A SCEV is not low-cost just because you can divide it by a power of 2. We need to also
check what we are dividing to make sure it too is not a high-code expansion. This helps
to not expand the exit value of certain loops, helping not to bloat the code.

The change in no-iv-rewrite.ll is reverting back to what it was testing before rL194116,
and looks a lot like the other tests in replace-loop-exit-folds.ll.

Differential Revision: https://reviews.llvm.org/D58435

llvm-svn: 355393

5 years ago[lldb] Fix linux host build after r355342
Alexander Kornienko [Tue, 5 Mar 2019 12:05:35 +0000 (12:05 +0000)]
[lldb] Fix linux host build after r355342

llvm-svn: 355392

5 years ago[llvm-objcopy] - Report "no zlib available" error properly when --compress-debug...
George Rimar [Tue, 5 Mar 2019 11:32:14 +0000 (11:32 +0000)]
[llvm-objcopy] - Report "no zlib available" error properly when --compress-debug-sections is used.

If zlib is not available, and --compress-debug-sections is passed,
we want to report an error. Currently, it is only reported for
--compress_debug_sections= form of the option.

Fixes the https://bugs.llvm.org/show_bug.cgi?id=40886.

I do not think there is a way to write a test for this.

Differential revision: https://reviews.llvm.org/D58909

llvm-svn: 355391

5 years ago[ASTImporter] Fix redecl failures of Class and ClassTemplate
Gabor Marton [Tue, 5 Mar 2019 11:23:24 +0000 (11:23 +0000)]
[ASTImporter] Fix redecl failures of Class and ClassTemplate

Summary:
Redecl chains of classes and class templates are not handled well
currently. We want to handle them similarly to functions, i.e. try to
keep the structure of the original AST as much as possible. The aim is
to not squash a prototype with a definition, rather we create both and
put them in a redecl chain.

Reviewers: a_sidorin, shafik, a.sidorin

Subscribers: rnkovacs, dkrupp, Szelethus, gamesh411, jdoerfert, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D58502

llvm-svn: 355390

5 years ago[SCEV] Add some extra tests for IndVarSimplifys loop exit values. NFC.
David Green [Tue, 5 Mar 2019 11:18:55 +0000 (11:18 +0000)]
[SCEV] Add some extra tests for IndVarSimplifys loop exit values. NFC.

Add some tests for various loops of the form:
  while(S >= 32) {
    S -= 32;
    something();
  };
  return S;

llvm-svn: 355389

5 years agoFix embedded Python initialization according to changes in version 3.7
Tatyana Krasnukha [Tue, 5 Mar 2019 11:18:45 +0000 (11:18 +0000)]
Fix embedded Python initialization according to changes in version 3.7

Differential Revision: https://reviews.llvm.org/D58833

llvm-svn: 355388

5 years ago[WebAssembly] Rename a variable in LateEHPrepare (NFC)
Heejin Ahn [Tue, 5 Mar 2019 11:11:34 +0000 (11:11 +0000)]
[WebAssembly] Rename a variable in LateEHPrepare (NFC)

llvm-svn: 355387

5 years agoAdd wildcard support to all update_*_test_checks.py scripts (PR37500)
Simon Pilgrim [Tue, 5 Mar 2019 10:44:37 +0000 (10:44 +0000)]
Add wildcard support to all update_*_test_checks.py scripts (PR37500)

We can already update multiple files in each update call, this extends it to work with wildcards as well in the same way as update_mca_test_checks.py (to support shells that won't do this for us - windows command prompt etc.)

Differential Revision: https://reviews.llvm.org/D58817

llvm-svn: 355386

5 years ago[ARM] Fix select_cc lowering for fp16
Oliver Stannard [Tue, 5 Mar 2019 10:42:34 +0000 (10:42 +0000)]
[ARM] Fix select_cc lowering for fp16

When lowering a select_cc node where the true and false values are of type f16,
we can't use a general conditional move because the FP16 instructions do not
support conditional execution. Instead, we must ensure that the condition code
is one of the four supported by the VSEL instruction.

Differential revision: https://reviews.llvm.org/D58813

llvm-svn: 355385

5 years ago[AMDGPU] Omit KILL instructions from hazard recognizer
David Stuttard [Tue, 5 Mar 2019 10:25:16 +0000 (10:25 +0000)]
[AMDGPU] Omit KILL instructions from hazard recognizer

Summary:
In some cases the KILL was causing a hazard to be introduced as these were
scheduled into hazard slots, but don't result in an instruction.

KILL shouldn't be considered for hazard recognition.

Change-Id: Ib6d2a2160f8c94cd0ce611ab198c7e4f46aeffcf

Subscribers: arsenm, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, tpr, t-tye, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58898

llvm-svn: 355384

5 years ago[ELF][test] chmod u+w to prevent annoying rm prompt
Fangrui Song [Tue, 5 Mar 2019 09:56:50 +0000 (09:56 +0000)]
[ELF][test] chmod u+w to prevent annoying rm prompt

If the permission does not permit writing and the standard input is a
terminal, rm gives an annoying prompt. chmod u+w to make the output
directory easier to delete.

llvm-svn: 355382

5 years ago[NFC] Simplify BufferedStackTrace::UnwindImpl
Vitaly Buka [Tue, 5 Mar 2019 05:52:34 +0000 (05:52 +0000)]
[NFC] Simplify BufferedStackTrace::UnwindImpl

llvm-svn: 355381

5 years ago[NFC] Move isUnwinding check into ScopedUnwinding
Vitaly Buka [Tue, 5 Mar 2019 05:40:05 +0000 (05:40 +0000)]
[NFC] Move isUnwinding check into ScopedUnwinding

llvm-svn: 355380

5 years ago[LangRef] Add 'callbr' instruction to the 'blockaddress' section.
Craig Topper [Tue, 5 Mar 2019 05:23:37 +0000 (05:23 +0000)]
[LangRef] Add 'callbr' instruction to the 'blockaddress' section.

llvm-svn: 355379

5 years ago[PowerPC] fix killed/dead flag after convert x-form to d-form tranformation.
Chen Zheng [Tue, 5 Mar 2019 04:56:54 +0000 (04:56 +0000)]
[PowerPC] fix killed/dead flag after convert x-form to d-form tranformation.
Differential Revision: https://reviews.llvm.org/D58428

llvm-svn: 355378

5 years ago[NFC] Move ScopedUnwinding from .h to .cc
Vitaly Buka [Tue, 5 Mar 2019 04:36:56 +0000 (04:36 +0000)]
[NFC] Move ScopedUnwinding from .h to .cc

llvm-svn: 355377

5 years ago[NFC] Reorder ifs in BufferedStackTrace::UnwindImpl
Vitaly Buka [Tue, 5 Mar 2019 04:36:49 +0000 (04:36 +0000)]
[NFC] Reorder ifs in BufferedStackTrace::UnwindImpl

llvm-svn: 355376

5 years ago[Expression] Remove unused parameter from Evaluate
Alex Langford [Tue, 5 Mar 2019 03:33:34 +0000 (03:33 +0000)]
[Expression] Remove unused parameter from Evaluate

llvm-svn: 355375

5 years ago[ARM][MC] Update one test case in 'test/MC/Disassembler/ARM/invalid-armv7.txt'
Xing GUO [Tue, 5 Mar 2019 03:07:56 +0000 (03:07 +0000)]
[ARM][MC] Update one test case in 'test/MC/Disassembler/ARM/invalid-armv7.txt'

Summary:
Instruction `[0xfe 0xf0 0x20 0xe3]` is a valid instruction on ARM-v7, which is `dbg #14`. See:
https://www.cl.cam.ac.uk/research/srg/han/ACS-P35/zynq/ARMv7-A-R-manual.pdf
(Page: 377)

```
Encoding A1:
DBG<c> #<option>

|31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16|15 14 13 12 11 10 09 08 07 06 05 04 03 02 01 00|
|      cond | 0  0  1  1  0| 0| 1  0| 0  0  0  0| 1  1  1  1| 0  0  0  0| 1  1  1  1|    option |
```

Reviewers: fhahn, efriedma

Reviewed By: efriedma

Subscribers: javed.absar, kristof.beyls, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58873

llvm-svn: 355374

5 years ago[AMDGPU] Implement AMDGPUMCInstrAnalysis
Scott Linder [Tue, 5 Mar 2019 03:02:00 +0000 (03:02 +0000)]
[AMDGPU] Implement AMDGPUMCInstrAnalysis

Implement MCInstrAnalysis for AMDGPU, with default implementations save
for `evaluateBranch`.

Differential Revision: https://reviews.llvm.org/D58400

llvm-svn: 355373

5 years ago[NFC] Return on SANITIZER_MIPS && !IsValidFrame
Vitaly Buka [Tue, 5 Mar 2019 02:51:04 +0000 (02:51 +0000)]
[NFC] Return on SANITIZER_MIPS && !IsValidFrame

llvm-svn: 355372

5 years ago[NFC] Put bp into Unwind(.. false) for consistency
Vitaly Buka [Tue, 5 Mar 2019 02:50:57 +0000 (02:50 +0000)]
[NFC] Put bp into Unwind(.. false) for consistency

It's not used there anyway

llvm-svn: 355371

5 years ago[NFC] Move asan_inited and size reset on top of ::UnwindImpl
Vitaly Buka [Tue, 5 Mar 2019 02:50:49 +0000 (02:50 +0000)]
[NFC] Move asan_inited and size reset on top of  ::UnwindImpl

llvm-svn: 355370

5 years agoRevert "[NFC][Sanitizer] Cleanup ASan's GetStackTrace implementation"
Vitaly Buka [Tue, 5 Mar 2019 02:35:49 +0000 (02:35 +0000)]
Revert "[NFC][Sanitizer] Cleanup ASan's GetStackTrace implementation"

I've missed that UnwindSlow was removed from ScopedUnwinding.

This reverts commit 4ce918e3942f0333ccb7d65d6265f4fc5f5324be.

llvm-svn: 355369

5 years agoReplace clang::FileData with llvm::vfs::Status
Harlan Haskins [Tue, 5 Mar 2019 02:27:12 +0000 (02:27 +0000)]
Replace clang::FileData with llvm::vfs::Status

Summary:
FileData was only ever used as a container for the values in
llvm::vfs::Status, so they might as well be consolidated.

The `InPCH` member was also always set to false, and unused.

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D58924

llvm-svn: 355368

5 years agoFix -fsanitize=vptr badness in <__debug>
Eric Fiselier [Tue, 5 Mar 2019 02:10:31 +0000 (02:10 +0000)]
Fix -fsanitize=vptr badness in <__debug>

Summary:

This patch fixes a lifetime bug when inserting a new container into the debug database. It is
diagnosed by UBSAN when debug mode is enabled. This patch corrects how nodes are constructed
during insertion.

The fix requires unconditionally breaking the debug mode ABI. Users should not expect ABI
stability from debug mode.

Reviewers: ldionne, serge-sans-paille, EricWF

Reviewed By: EricWF

Subscribers: mclow.lists, christof, libcxx-commits

Tags: #libc

Differential Revision: https://reviews.llvm.org/D58011

llvm-svn: 355367

5 years agoMake VCRuntime ABI configuration a first-class option.
Eric Fiselier [Tue, 5 Mar 2019 01:57:01 +0000 (01:57 +0000)]
Make VCRuntime ABI configuration a first-class option.

Summary:
On Windows we currently provide two separate ABI configurations. One which defers to `vcruntime` to provide the C++ runtime and another which doesn't.
Using `vcruntime` allows interoperability which programs compiled against the MSVC STL, and should be preferred whenever possible.

When deferring to `vcruntime` much of the ABI we provide changes. Including the layout of `<stdexcept>` types, their vtables, and how the linkage of their members.

This patch introduces the `_LIBCPP_ABI_VCRUNTIME` macro to denote this configuration. It also cleans up the existing configuration for using `vcruntime`.

This cleanup lays the groundwork for fixing a number of ABI and interoperability bugs in  `<stdexcept>`.

Reviewers: thomasanderson, ldionne, smeenai

Reviewed By: smeenai

Subscribers: jdoerfert, libcxx-commits, #libc

Differential Revision: https://reviews.llvm.org/D58942

llvm-svn: 355366

5 years ago[clang-format][docs][NFC] Fix example for Allman brace breaking style
Jan Korous [Tue, 5 Mar 2019 01:45:31 +0000 (01:45 +0000)]
[clang-format][docs][NFC] Fix example for Allman brace breaking style

I assume the example is wrong as it's clearly missing line-breaks before
braces.

I just ran the example through clang-format with .clang-format like
this:
BreakBeforeBraces: Allman

Differential Revision: https://reviews.llvm.org/D58941

llvm-svn: 355365

5 years agoRevert "[testsuite] Port crashlog and dependencies to Python 3."
Davide Italiano [Tue, 5 Mar 2019 01:34:47 +0000 (01:34 +0000)]
Revert "[testsuite] Port crashlog and dependencies to Python 3."

This revert the commit because it broke the bots. I need to find
a way that works with both versions.

llvm-svn: 355364

5 years agoRevert compiler-rt diffs for order file instrumentation to get bot green!
Manman Ren [Tue, 5 Mar 2019 01:21:40 +0000 (01:21 +0000)]
Revert compiler-rt diffs for order file instrumentation to get bot green!

This caused issues on Linux/Windows and other platforms.

r355343 355350 355350

llvm-svn: 355363

5 years agoPHI nodes are not `FPMathOperator` s
Sanjoy Das [Tue, 5 Mar 2019 01:15:08 +0000 (01:15 +0000)]
PHI nodes are not `FPMathOperator` s

Reviewers: chandlerc, arsenm

Reviewed By: arsenm

Subscribers: wdng, arsenm, mcrosier, jlebar, bixia, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58887

llvm-svn: 355362

5 years ago[X86] Reduce some patterns by using FP instructions for integer types even when AVX2...
Craig Topper [Tue, 5 Mar 2019 01:14:25 +0000 (01:14 +0000)]
[X86] Reduce some patterns by using FP instructions for integer types even when AVX2 is available and execution domain fixing will do the right thing

We have quite a few cases of using FP instructions for integer operations when only AVX1 is available. Then we switch to integer instructions with AVX2. In a lot of these cases execution domain fixing will take care of turning FP instructions into integer if its profitable.

With this patch we just keep on using the FP instructions even with AVX2. I've only handled some cases that don't require messing with patterns that are defined in the instruction definition. Those will require more subtle multiclass work possibly involving null_frag, hasSideEffects = 0, etc.

Differential Revision: https://reviews.llvm.org/D58470

llvm-svn: 355361

5 years ago[BPF] Do not generate BTF sections unnecessarily
Yonghong Song [Tue, 5 Mar 2019 01:01:21 +0000 (01:01 +0000)]
[BPF] Do not generate BTF sections unnecessarily

If There is no types/non-empty strings, do not generate
.BTF section. If there is no func_info/line_info, do
not generate .BTF.ext section.

Signed-off-by: Yonghong Song <yhs@fb.com>
Differential Revision: https://reviews.llvm.org/D58936

llvm-svn: 355360

5 years ago[testsuite] Port crashlog and dependencies to Python 3.
Davide Italiano [Tue, 5 Mar 2019 00:53:38 +0000 (00:53 +0000)]
[testsuite] Port crashlog and dependencies to Python 3.

Fixes three tests in the testsuite.

llvm-svn: 355359

5 years agoFix the Xcode project for UserIDResolver.
Jim Ingham [Tue, 5 Mar 2019 00:52:12 +0000 (00:52 +0000)]
Fix the Xcode project for UserIDResolver.

llvm-svn: 355358

5 years agoAttemp to fix windows profile-rt build breakage.
Manman Ren [Tue, 5 Mar 2019 00:52:00 +0000 (00:52 +0000)]
Attemp to fix windows profile-rt build breakage.

Followup to D57530.

llvm-svn: 355357

5 years ago[lldbtest] Check against the correct name for libcxxabi (macOS).
Davide Italiano [Tue, 5 Mar 2019 00:47:15 +0000 (00:47 +0000)]
[lldbtest] Check against the correct name for libcxxabi (macOS).

llvm-svn: 355356

5 years ago[NFC][Sanitizer] Cleanup ASan's GetStackTrace implementation
Julian Lettner [Tue, 5 Mar 2019 00:41:15 +0000 (00:41 +0000)]
[NFC][Sanitizer] Cleanup ASan's GetStackTrace implementation

Cleanup ASan's  __sanitizer::BufferedStackTrace::UnwindImpl (formerly
GetStackTrace) implementation. Start with ASan because it is the most
complex implementation.

GetStackTrace implementations seem to have started out as exact copies
of the original implementation in ASan, but have diverged in subtle
ways. My goal is to parameterize this algorithm (via templating or
callbacks) so we can share the implementation and get rid of the
inversed dependency (sanitizer_common depends on concrete
implementations in asan, ubsan, etc.). This should also help us to avoid
those pesky linker errors caused by undefined, duplicate, and weak
symbols on Windows.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D58861

llvm-svn: 355355

5 years ago[cmake] Create exports for umbrella library targets
Shoaib Meenai [Tue, 5 Mar 2019 00:38:32 +0000 (00:38 +0000)]
[cmake] Create exports for umbrella library targets

When using the umbrella llvm-libraries and clang-libraries targets, we
should export all library targets, otherwise they'll be part of our
distribution but not usable from the CMake package.

Differential Revision: https://reviews.llvm.org/D58862

llvm-svn: 355354

5 years ago[Host] Fix the build (and the modules build).
Davide Italiano [Tue, 5 Mar 2019 00:37:40 +0000 (00:37 +0000)]
[Host] Fix the build (and the modules build).

-> Add a missing include to find the base class.
-> Add a missing out-of-line declaration for a member function.

llvm-svn: 355353

5 years agoRevert "[DataFormatters] Fix regression in libc++ std::atomic formatter caused by...
Shafik Yaghmour [Tue, 5 Mar 2019 00:29:58 +0000 (00:29 +0000)]
Revert "[DataFormatters] Fix regression in libc++ std::atomic formatter caused by https://reviews.llvm.org/D56913"

This reverts commit r355351.

llvm-svn: 355352

5 years ago[DataFormatters] Fix regression in libc++ std::atomic formatter caused by https:...
Shafik Yaghmour [Tue, 5 Mar 2019 00:17:18 +0000 (00:17 +0000)]
[DataFormatters] Fix regression in libc++ std::atomic formatter caused by https://reviews.llvm.org/D56913

rdar://problem/48568543

llvm-svn: 355351

5 years agoAttemp to fix build brokage due to D57530.
Manman Ren [Mon, 4 Mar 2019 23:41:57 +0000 (23:41 +0000)]
Attemp to fix build brokage due to D57530.

By adding implementations for __llvm_profile_begin_orderfile for non-Darwin platforms.

llvm-svn: 355350

5 years agoFix invalid target triples in tests. (NFC)
Florian Hahn [Mon, 4 Mar 2019 23:37:41 +0000 (23:37 +0000)]
Fix invalid target triples in tests. (NFC)

llvm-svn: 355349

5 years ago[msan] Instrument x86 BMI intrinsics.
Evgeniy Stepanov [Mon, 4 Mar 2019 22:58:20 +0000 (22:58 +0000)]
[msan] Instrument x86 BMI intrinsics.

Summary:
They simply shuffle bits. MSan needs to do the same with shadow bits,
after making sure that the shuffle mask is fully initialized.

Reviewers: pcc, vitalybuka

Subscribers: hiraditya, #sanitizers, llvm-commits

Tags: #sanitizers, #llvm

Differential Revision: https://reviews.llvm.org/D58858

llvm-svn: 355348

5 years ago[sanitizer] Intercept bzero.
Evgeniy Stepanov [Mon, 4 Mar 2019 22:58:11 +0000 (22:58 +0000)]
[sanitizer] Intercept bzero.

Summary:
Intercept bzero and enable existing __bzero interceptor in Linux.
bzero is deprecated but still used occasionally.

Reviewers: vitalybuka

Subscribers: srhines, kubamracek, #sanitizers, llvm-commits

Tags: #sanitizers, #llvm

Differential Revision: https://reviews.llvm.org/D58850

llvm-svn: 355347

5 years ago[NFC] Fix PGO link error in shared libs build
Jordan Rupprecht [Mon, 4 Mar 2019 22:54:44 +0000 (22:54 +0000)]
[NFC] Fix PGO link error in shared libs build

llvm-svn: 355346

5 years ago[CodeGenPrepare] avoid crashing on non-canonical/degenerate code
Sanjay Patel [Mon, 4 Mar 2019 22:47:13 +0000 (22:47 +0000)]
[CodeGenPrepare] avoid crashing on non-canonical/degenerate code

The test is reduced from an example in the post-commit thread for:
rL354746
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20190304/632396.html

While we must avoid dying here, the real question should be:
Why is non-canonical and/or degenerate code making it to CGP when
using the new pass manager?

llvm-svn: 355345

5 years ago[GlobalISel][AArch64] Add selection support for G_EXTRACT_VECTOR_ELT
Jessica Paquette [Mon, 4 Mar 2019 22:35:32 +0000 (22:35 +0000)]
[GlobalISel][AArch64] Add selection support for G_EXTRACT_VECTOR_ELT

This adds instruction selection support for G_EXTRACT_VECTOR_ELT for cases
where the index is defined by a G_CONSTANT.

It also factos out the lane copy opcode selection part into its own function,
`getLaneCopyOpcode`. This is used by both `selectUnmergeValues` and
`selectExtractElt`.

Differential Revision: https://reviews.llvm.org/D58469

llvm-svn: 355344

5 years agoOrder File Instrumentation: dump the data in compiler-rt
Manman Ren [Mon, 4 Mar 2019 22:28:38 +0000 (22:28 +0000)]
Order File Instrumentation: dump the data in compiler-rt

The profile data will be dumped in a file default_xxx.profraw.order.

Differential Revision: https://reviews.llvm.org/D57530

llvm-svn: 355343

5 years agoMove ProcessInfo from Host to Utility.
Zachary Turner [Mon, 4 Mar 2019 21:51:03 +0000 (21:51 +0000)]
Move ProcessInfo from Host to Utility.

There are set of classes in Target that describe the parameters of a
process - e.g. it's PID, name, user id, and similar. However, since it
is a bare description of a process and contains no actual functionality,
there's nothing specifically that makes this appropriate for being in
Target -- it could just as well be describing a process on the host, or
some hypothetical virtual process that doesn't even exist.

To cement this, I'm moving these classes to Utility. It's possible that
we can find a better place for it in the future, but as it is neither
Host specific nor Target specific, Utility seems like the most appropriate
place for the time being.

After this there is only 2 remaining references to Target from Host,
which I'll address in a followup.

Differential Revision: https://reviews.llvm.org/D58842

llvm-svn: 355342

5 years ago[build.py] Allow clang-cl to build files starting with '/U'
Alex Langford [Mon, 4 Mar 2019 21:36:49 +0000 (21:36 +0000)]
[build.py] Allow clang-cl to build files starting with '/U'

Summary:
clang-cl tries to match cl's interface, and treats /U as "Removes a
predefined macro" as cl does. When you feed clang-cl a file that begins with
'/U' (e.g. /Users/xiaobai/foo.c), clang-cl will emit a warning and in some cases
an error, like so:

clang-9: warning: '/Users/xiaobai/foo.c' treated as the '/U' option [-Wslash-u-filename]
clang-9: note: Use '--' to treat subsequent arguments as filenames
clang-9: error: no input files

If you're using clang-cl, make sure '--' is passed before the source file.

Differential Revision: https://reviews.llvm.org/D58860

llvm-svn: 355341

5 years ago[build] Rename clang-headers to clang-resource-headers
Shoaib Meenai [Mon, 4 Mar 2019 21:19:53 +0000 (21:19 +0000)]
[build] Rename clang-headers to clang-resource-headers

Summary:
The current install-clang-headers target installs clang's resource
directory headers. This is different from the install-llvm-headers
target, which installs LLVM's API headers. We want to introduce the
corresponding target to clang, and the natural name for that new target
would be install-clang-headers. Rename the existing target to
install-clang-resource-headers to free up the install-clang-headers name
for the new target, following the discussion on cfe-dev [1].

I didn't find any bots on zorg referencing install-clang-headers. I'll
send out another PSA to cfe-dev to accompany this rename.

[1] http://lists.llvm.org/pipermail/cfe-dev/2019-February/061365.html

Reviewers: beanz, phosek, tstellar, rnk, dim, serge-sans-paille

Subscribers: mgorny, javed.absar, jdoerfert, #sanitizers, openmp-commits, lldb-commits, cfe-commits, llvm-commits

Tags: #clang, #sanitizers, #lldb, #openmp, #llvm

Differential Revision: https://reviews.llvm.org/D58791

llvm-svn: 355340

5 years ago[GlobalISel][AArch64] Legalize vector G_SELECT
Jessica Paquette [Mon, 4 Mar 2019 21:12:46 +0000 (21:12 +0000)]
[GlobalISel][AArch64] Legalize vector G_SELECT

Just scalarize it, and add a test showing it works.

Differential Revision: https://reviews.llvm.org/D58747

llvm-svn: 355339

5 years agoFix wrong enum value in switch.
Evgeniy Stepanov [Mon, 4 Mar 2019 21:00:28 +0000 (21:00 +0000)]
Fix wrong enum value in switch.

llvm-svn: 355338

5 years ago[ConstantHoisting] avoid hang/crash from unreachable blocks (PR40930)
Sanjay Patel [Mon, 4 Mar 2019 20:57:14 +0000 (20:57 +0000)]
[ConstantHoisting] avoid hang/crash from unreachable blocks (PR40930)

I'm not too familiar with this pass, so there might be a better
solution, but this appears to fix the degenerate:
PR40930
PR40931
PR40932
PR40934
...without affecting any real-world code.

As we've seen in several other passes, when we have unreachable blocks,
they can contain semi-bogus IR and/or cause unexpected conditions. We
would not typically expect these patterns to make it this far, but we
have to guard against them anyway.

llvm-svn: 355337

5 years ago[libc++] Remove old CMake workaround
Louis Dionne [Mon, 4 Mar 2019 20:52:14 +0000 (20:52 +0000)]
[libc++] Remove old CMake workaround

We haven't had any complaints so far, and I don't think anybody builds
libc++ from source for that old platform anymore.

llvm-svn: 355336

5 years agoInstall GCC 5 on buildbots to replace GCC 4.9.
Eric Fiselier [Mon, 4 Mar 2019 20:49:18 +0000 (20:49 +0000)]
Install GCC 5 on buildbots to replace GCC 4.9.

LLVM is dropping support for GCC 4.9. This patch adds
a GCC 5 installation to the buildbot image so we can upgrade
the 4.9 bot to GCC 5.

As a temporary workaround until zorg updates, we install GCC 5
as GCC 4.9.

llvm-svn: 355334

5 years agoOrder File Instrumentation: add clang support for -forder-file-instrumentation
Manman Ren [Mon, 4 Mar 2019 20:30:30 +0000 (20:30 +0000)]
Order File Instrumentation: add clang support for -forder-file-instrumentation

When -forder-file-instrumentation is on, we pass llvm flag to enable the order file instrumentation pass.

https://reviews.llvm.org/D58751

llvm-svn: 355333

5 years ago[ASTImporter] Handle built-in when importing SourceLocation and FileID
Shafik Yaghmour [Mon, 4 Mar 2019 20:25:54 +0000 (20:25 +0000)]
[ASTImporter] Handle built-in when importing SourceLocation and FileID

Summary:
Currently when we see a built-in we try and import the include location. Instead what we do now is find the buffer like we do for the invalid case and copy that over to the to context.

Differential Revision: https://reviews.llvm.org/D58743

llvm-svn: 355332

5 years ago[PGO] Clang part of change for context-sensitive PGO (part1)
Rong Xu [Mon, 4 Mar 2019 20:21:31 +0000 (20:21 +0000)]
[PGO] Clang part of change for context-sensitive PGO (part1)

Part 1 of CSPGO change in Clang. This includes changes in clang options
and calls to llvm PassManager. Tests will be committed in part2.
This change needs the PassManager change in llvm.

Differential Revision: https://reviews.llvm.org/D54176

llvm-svn: 355331

5 years ago[PGO] Context sensitive PGO (part 3)
Rong Xu [Mon, 4 Mar 2019 20:21:27 +0000 (20:21 +0000)]
[PGO] Context sensitive PGO (part 3)

Part 3 of CSPGO changes (mostly related to PassMananger).

Differential Revision: https://reviews.llvm.org/D54175

llvm-svn: 355330

5 years agoFix Windows build after UserIDResolver patch.
Zachary Turner [Mon, 4 Mar 2019 19:57:04 +0000 (19:57 +0000)]
Fix Windows build after UserIDResolver patch.

That patch added a function to HostInfo that returns an instance
of UserIDResolver, but this function was unimplemented on Windows,
leading to linker errors.  For now, just return a dummy implementation
that doesn't resolve user ids to get the build green.

llvm-svn: 355329

5 years ago[InstCombine] Add tests for add nsw + sadd.with.overflow; NFC
Nikita Popov [Mon, 4 Mar 2019 19:35:46 +0000 (19:35 +0000)]
[InstCombine] Add tests for add nsw + sadd.with.overflow; NFC

Baseline tests for D58881, which fixes part of PR38146.

Patch by Dan Robertson.

llvm-svn: 355328

5 years ago[Subtarget] Follow up to r355167, add another set of curly braces to FeatureBitArray...
Craig Topper [Mon, 4 Mar 2019 19:23:37 +0000 (19:23 +0000)]
[Subtarget] Follow up to r355167, add another set of curly braces to FeatureBitArray initialization to satisfy older versions of clang.

Apparently older versions of clang like 3.6 require an extra set of curly braces around std::array initializations. I'm told the C++ language was changed regarding this by CWG 1270.

llvm-svn: 355327

5 years agoRe-commit r355104: "[AArch64][GlobalISel] Add support for 64 bit vector shuffle using...
Amara Emerson [Mon, 4 Mar 2019 19:16:00 +0000 (19:16 +0000)]
Re-commit r355104: "[AArch64][GlobalISel] Add support for 64 bit vector shuffle using TBL1."

The code to materialize a mask from a constant pool load tried to use a 128 bit
LDR to load a 64 bit constant pool entry, which was 8 byte aligned. This resulted
in a link failure in the NEON tests in the test suite since the LDR address was
unaligned. This change fixes that to instead emit a 64 bit LDR if the entry is
64 bit, before converting back to a 128 bit register for the TBL.

llvm-svn: 355326

5 years ago[MC] Teach ELFObjectWriter that parse-time variables do not appear in
Nirav Dave [Mon, 4 Mar 2019 19:12:56 +0000 (19:12 +0000)]
[MC] Teach ELFObjectWriter that parse-time variables do not appear in
symbol table.

llvm-svn: 355325

5 years ago[DAGCombiner][X86][SystemZ][AArch64] Combine some cases of (bitcast (build_vector...
Craig Topper [Mon, 4 Mar 2019 19:12:16 +0000 (19:12 +0000)]
[DAGCombiner][X86][SystemZ][AArch64] Combine some cases of (bitcast (build_vector constants)) between legalize types and legalize dag.

This patch enables combining integer bitcasts of integer build vectors when the new scalar type is legal. I've avoided floating point because the implementation bitcasts float to int along the way and we would need to check the intermediate types for legality

Differential Revision: https://reviews.llvm.org/D58884

llvm-svn: 355324

5 years agoRefactor user/group name resolving code
Pavel Labath [Mon, 4 Mar 2019 18:48:00 +0000 (18:48 +0000)]
Refactor user/group name resolving code

Summary:
This creates an abstract base class called "UserIDResolver", which can
be implemented to provide user/group ID resolution capabilities for
various objects. Posix host implement a PosixUserIDResolver, which does
that using posix apis (getpwuid and friends).  PlatformGDBRemote
forwards queries over the gdb-remote link, etc. ProcessInstanceInfo
class is refactored to make use of this interface instead of taking a
platform pointer as an argument. The base resolver class already
implements caching and thread-safety, so implementations don't have to
worry about that.

The main motivating factor for this was to remove external dependencies
from the ProcessInstanceInfo class (so it can be put next to
ProcessLaunchInfo and friends), but it has other benefits too:
- ability to test the user name caching code
- ability to test ProcessInstanceInfo dumping code
- consistent interface for user/group resolution between Platform and
  Host classes.

Reviewers: zturner, clayborg, jingham

Subscribers: mgorny, lldb-commits

Differential Revision: https://reviews.llvm.org/D58167

llvm-svn: 355323

5 years agoEnable _rotl, _lrotl, _rotr, _lrotr on all platforms.
Erich Keane [Mon, 4 Mar 2019 18:47:21 +0000 (18:47 +0000)]
Enable _rotl, _lrotl, _rotr, _lrotr on all platforms.

The above builtins are currently implemented for MSVC mode, however GCC
also implements these.  This patch enables them for all platforms.

Additionally, this corrects the type for these builtins to always be
'long int' to match the specification in the Intel Intrinsics Guide.

Change-Id: Ida34be98078709584ef5136c8761783435ec02b1
llvm-svn: 355322

5 years ago[WebAssembly] Add support for data sections in the assembler.
Wouter van Oortmerssen [Mon, 4 Mar 2019 17:18:04 +0000 (17:18 +0000)]
[WebAssembly] Add support for data sections in the assembler.

Summary:
This is quite minimal so far, introduce them with .section,
fill them with .int8 or .asciz, end with .size

Reviewers: dschuff, sbc100, aheejin

Subscribers: jgravelle-google, sunfish, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58660

llvm-svn: 355321

5 years ago[lldb] [test] Mark failing tests XFAIL on NetBSD
Michal Gorny [Mon, 4 Mar 2019 16:54:06 +0000 (16:54 +0000)]
[lldb] [test] Mark failing tests XFAIL on NetBSD

Add a convenience 'expectedFailureNetBSD' decorator and mark all tests
currently failing on NetBSD with it.  Also skip a few tests that hang
the test suite.  This should establish a baseline for the test suite
and get us closer to enabling tests on buildbot.  This will help us
catch regressions while we still have a lot of work to do to get tests
working.

It seems that there are also some flaky tests.  I am going to address
them later on.

Differential Revision: https://reviews.llvm.org/D58527

llvm-svn: 355320

5 years agoRevert duplicate check for DragonFly BSD
Kamil Rytarowski [Mon, 4 Mar 2019 15:51:02 +0000 (15:51 +0000)]
Revert duplicate check for DragonFly BSD

Summary: Revert duplicate check for DragonFly BSD

Submitted by tuxillo.

Reviewers: krytarowski

Reviewed By: krytarowski

Subscribers: mgorny, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D58907

llvm-svn: 355319

5 years ago[libc++] decoupling Freestanding atomic<T> from libatomic.a
Louis Dionne [Mon, 4 Mar 2019 15:26:27 +0000 (15:26 +0000)]
[libc++] decoupling Freestanding atomic<T> from libatomic.a

This patch introduces non-lockfree atomics that do not require using
an external libatomic. This work is done with the long-term goal of
allowing the use of <atomic> in freestanding environments.

Thanks to Olivier Giroux for the patch.
Differential Revision: https://reviews.llvm.org/D56913

llvm-svn: 355318

5 years agoGive builtins and alloc/dealloc operators the default calling convention.
Erich Keane [Mon, 4 Mar 2019 14:54:52 +0000 (14:54 +0000)]
Give builtins and alloc/dealloc operators the default calling convention.

On SPIR targets, the default calling convention is SpirFunction.
However, operator new/delete and builtins were being created with CC_C.
The result is indirect references to new/delete (or builtins that are permitted
to be called indirectly have a mismatched type, as well as questionable codegen
in some cases.

This patch sets both to the default calling convention, so that it
properly matches the calling convention of the target.

Differential Revision: https://reviews.llvm.org/D58844

Change-Id: I52065bb00bc2655945caea8f29c409ba1e0ac24a
llvm-svn: 355317

5 years ago[X86] Regenerate illegal type load test with non-undef load address.
Simon Pilgrim [Mon, 4 Mar 2019 14:49:02 +0000 (14:49 +0000)]
[X86] Regenerate illegal type load test with non-undef load address.

This would be affected by an upcoming patch without undoing some of the bugpoint reduction.

llvm-svn: 355316

5 years agoUnbreak shared library linkage on DragonFlyBSD.
Kamil Rytarowski [Mon, 4 Mar 2019 14:36:43 +0000 (14:36 +0000)]
Unbreak shared library linkage on DragonFlyBSD.

Patch submitted by rimvydas.

Reviewers: llvm-commits, krytarowski, mgorny

Reviewed By: krytarowski, mgorny

Subscribers: mgorny

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D35125

llvm-svn: 355315

5 years ago[MCA] Remove unused methods. NFC
Andrea Di Biagio [Mon, 4 Mar 2019 13:34:56 +0000 (13:34 +0000)]
[MCA] Remove unused methods. NFC

llvm-svn: 355314

5 years ago[AMDGPU][MC] Enable lds_direct operand for v_readfirstlane_b32, v_readlane_b32 and...
Dmitry Preobrazhensky [Mon, 4 Mar 2019 12:48:32 +0000 (12:48 +0000)]
[AMDGPU][MC] Enable lds_direct operand for v_readfirstlane_b32, v_readlane_b32 and v_writelane_b32

See bug 40662: https://bugs.llvm.org/show_bug.cgi?id=40662

Reviewers: artem.tamazov, arsenm, rampitec

Differential Revision: https://reviews.llvm.org/D58713

llvm-svn: 355312

5 years ago[MCA] Correctly initialize struct SummaryView::BackPressureInfo.
Andrea Di Biagio [Mon, 4 Mar 2019 12:23:05 +0000 (12:23 +0000)]
[MCA] Correctly initialize struct SummaryView::BackPressureInfo.

This should appease the buildbots.

llvm-svn: 355309

5 years ago[MCA] Highlight kernel bottlenecks in the summary view.
Andrea Di Biagio [Mon, 4 Mar 2019 11:52:34 +0000 (11:52 +0000)]
[MCA] Highlight kernel bottlenecks in the summary view.

This patch adds a new flag named -bottleneck-analysis to print out information
about throughput bottlenecks.

MCA knows how to identify and classify dynamic dispatch stalls. However, it
doesn't know how to analyze and highlight kernel bottlenecks.  The goal of this
patch is to teach MCA how to correlate increases in backend pressure to backend
stalls (and therefore, the loss of throughput).

From a Scheduler point of view, backend pressure is a function of the scheduler
buffer usage (i.e. how the number of uOps in the scheduler buffers changes over
time). Backend pressure increases (or decreases) when there is a mismatch
between the number of opcodes dispatched, and the number of opcodes issued in
the same cycle.  Since buffer resources are limited, continuous increases in
backend pressure would eventually leads to dispatch stalls. So, there is a
strong correlation between dispatch stalls, and how backpressure changed over
time.

This patch teaches how to identify situations where backend pressure increases
due to:
 - unavailable pipeline resources.
 - data dependencies.

Data dependencies may delay execution of instructions and therefore increase the
time that uOps have to spend in the scheduler buffers. That often translates to
an increase in backend pressure which may eventually lead to a bottleneck.
Contention on pipeline resources may also delay execution of instructions, and
lead to a temporary increase in backend pressure.

Internally, the Scheduler classifies instructions based on whether register /
memory operands are available or not.

An instruction is marked as "ready to execute" only if data dependencies are
fully resolved.
Every cycle, the Scheduler attempts to execute all instructions that are ready
to execute. If an instruction cannot execute because of unavailable pipeline
resources, then the Scheduler internally updates a BusyResourceUnits mask with
the ID of each unavailable resource.

ExecuteStage is responsible for tracking changes in backend pressure. If backend
pressure increases during a cycle because of contention on pipeline resources,
then ExecuteStage sends a "backend pressure" event to the listeners.
That event would contain information about instructions delayed by resource
pressure, as well as the BusyResourceUnits mask.

Note that ExecuteStage also knows how to identify situations where backpressure
increased because of delays introduced by data dependencies.

The SummaryView observes "backend pressure" events and prints out a "bottleneck
report".

Example of bottleneck report:

```
Cycles with backend pressure increase [ 99.89% ]
Throughput Bottlenecks:
  Resource Pressure       [ 0.00% ]
  Data Dependencies:      [ 99.89% ]
   - Register Dependencies [ 0.00% ]
   - Memory Dependencies   [ 99.89% ]
```

A bottleneck report is printed out only if increases in backend pressure
eventually caused backend stalls.

About the time complexity:

Time complexity is linear in the number of instructions in the
Scheduler::PendingSet.

The average slowdown tends to be in the range of ~5-6%.
For memory intensive kernels, the slowdown can be significant if flag
-noalias=false is specified. In the worst case scenario I have observed a
slowdown of ~30% when flag -noalias=false was specified.

We can definitely recover part of that slowdown if we optimize class LSUnit (by
doing extra bookkeeping to speedup queries). For now, this new analysis is
disabled by default, and it can be enabled via flag -bottleneck-analysis. Users
of MCA as a library can enable the generation of pressure events through the
constructor of ExecuteStage.

This patch partially addresses https://bugs.llvm.org/show_bug.cgi?id=37494

Differential Revision: https://reviews.llvm.org/D58728

llvm-svn: 355308

5 years ago[X86] Avoid codegen changes when DBG_VALUE appears between lowered selects
Jeremy Morse [Mon, 4 Mar 2019 10:56:02 +0000 (10:56 +0000)]
[X86] Avoid codegen changes when DBG_VALUE appears between lowered selects

X86TargetLowering::EmitLoweredSelect presently detects sequences of CMOV pseudo
instructions without accounting for debug intrinsics. This leads to different
codegen with and without option -g, if a DBG_VALUE instruction lands in the
middle of several lowered selects.

Work around this by skipping over debug instructions when looking for CMOV
sequences, and sinking those debug insts into the EmitLoweredSelect sunk block.
This might slightly shift where variables appear in the instruction sequence,
but won't re-order assignments.

Differential Revision: https://reviews.llvm.org/D58672

llvm-svn: 355307

5 years ago[LLD] add test for line number info related to differrent .text sections
Alexey Lapshin [Mon, 4 Mar 2019 10:42:54 +0000 (10:42 +0000)]
[LLD] add test for line number info related to differrent .text sections

   There is a bug for llvm-objdump which incorrectly reported line number
   info for object file if several .text sections presented.
   https://bugs.llvm.org/show_bug.cgi?id=40703 .
   This patch adds test for the same situation.

   Differential revision: https://reviews.llvm.org/D58357

llvm-svn: 355306