platform/upstream/llvm.git
6 years ago[CMake] Rename `SANITIZER_HEADERS` to `SANITIZER_IMPL_HEADERS` under `lib/sanitizer_c...
Dan Liew [Thu, 12 Jul 2018 13:36:44 +0000 (13:36 +0000)]
[CMake] Rename `SANITIZER_HEADERS` to `SANITIZER_IMPL_HEADERS` under `lib/sanitizer_common`.

The variable name `SANITIZER_HEADERS` is already used for the list of
public headers in `include/CMakeLists.txt`.  Although the previous
implementation worked it's probably best to avoid shadowing global
variables to avoid confusion.

Differential Revision: https://reviews.llvm.org/D49176

llvm-svn: 336904

6 years ago[X86][SSE] Utilize ZeroableElements for canWidenShuffleElements
Simon Pilgrim [Thu, 12 Jul 2018 13:29:41 +0000 (13:29 +0000)]
[X86][SSE] Utilize ZeroableElements for canWidenShuffleElements

canWidenShuffleElements can do a better job if given a mask with ZeroableElements info. Apparently, ZeroableElements was being only used to identify AllZero candidates, but possibly we could plug it into more shuffle matchers.

Original Patch by Zvi Rackover @zvi

Differential Revision: https://reviews.llvm.org/D42044

llvm-svn: 336903

6 years ago[InstCombine]add testcases for folding more SPFofSPF pattern
Chen Zheng [Thu, 12 Jul 2018 13:28:20 +0000 (13:28 +0000)]
[InstCombine]add testcases for folding more SPFofSPF pattern

Differential Revision: https://reviews.llvm.org/D49222

llvm-svn: 336902

6 years ago[analyzer][UninitializedObjectChecker] Moved non-member functions out of the anonymou...
Kristof Umann [Thu, 12 Jul 2018 13:13:46 +0000 (13:13 +0000)]
[analyzer][UninitializedObjectChecker] Moved non-member functions out of the anonymous namespace

As the code for the checker grew, it became increasinly difficult to see
whether a function was global or statically defined. In this patch,
anything that isn't a type declaration or definition was moved out of the
anonymous namespace and is marked as static.

llvm-svn: 336901

6 years ago[X86][AVX] Use Zeroable mask to improve shuffle mask widening
Simon Pilgrim [Thu, 12 Jul 2018 13:03:58 +0000 (13:03 +0000)]
[X86][AVX] Use Zeroable mask to improve shuffle mask widening

Noticed while updating D42044, lowerV2X128VectorShuffle can improve the shuffle mask with the zeroable data to create a target shuffle mask to recognise more 'zero upper 128' patterns.

NOTE: lowerV4X128VectorShuffle could benefit as well but the code needs refactoring first to discriminate between SM_SentinelUndef and SM_SentinelZero for negative shuffle indices.

Differential Revision: https://reviews.llvm.org/D49092

llvm-svn: 336900

6 years ago[clangd] log request/response messages with method/ID/error at INFO level
Sam McCall [Thu, 12 Jul 2018 11:52:18 +0000 (11:52 +0000)]
[clangd] log request/response messages with method/ID/error at INFO level

Summary: Bodies are logged at VERBOSE level (since r336785), tweak the formatting.

Reviewers: hokein

Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, cfe-commits

Differential Revision: https://reviews.llvm.org/D49224

llvm-svn: 336899

6 years ago[ASTImporter] Fix infinite recursion on function import with struct definition in...
Gabor Marton [Thu, 12 Jul 2018 11:50:21 +0000 (11:50 +0000)]
[ASTImporter] Fix infinite recursion on function import with struct definition in parameters

Summary:
Importing a function having a struct definition in the parameter list
causes a crash in the importer via infinite recursion. This patch avoids
the crash and reports such functions as not supported. Unit tests make
sure that normal struct definitions inside function bodies work normally
on the other hand and LLDB-like type imports also do.

Reviewers: a.sidorin, martong

Differential Revision: https://reviews.llvm.org/D47946

Patch by Zoltan Gera!

llvm-svn: 336898

6 years ago[UnJ] Use SmallPtrSets for block collections. NFC
David Green [Thu, 12 Jul 2018 10:44:47 +0000 (10:44 +0000)]
[UnJ] Use SmallPtrSets for block collections. NFC

We no longer care about the order of blocks in these collections,
so can change to SmallPtrSets, making contains checks quicker.

Differential revision: https://reviews.llvm.org/D49060

llvm-svn: 336897

6 years ago[ASTImporter] Refactor Decl creation
Gabor Marton [Thu, 12 Jul 2018 09:42:05 +0000 (09:42 +0000)]
[ASTImporter] Refactor Decl creation

Summary:
Generalize the creation of Decl nodes during Import.  With this patch we do the
same things after and before a new AST node is created (::Create) The import
logic should be really simple, we create the node, then we mark that as
imported, then we recursively import the parts for that node and then set them
on that node.  However, the AST is actually a graph, so we have to handle
circles.  If we mark something as imported (`MapImported()`) then we return with
the corresponding `To` decl whenever we want to import that node again, this way
circles are handled.  In order to make this algorithm work we must ensure
things, which are handled in the generic CreateDecl<> template:
* There are no `Import()` calls in between any node creation (::Create)
and the `MapImported()` call.
* Before actually creating an AST node (::Create), we must check if
the Node had been imported already, if yes then return with that one.
One very important case for this is connected to templates: we may
start an import both from the templated decl of a template and from
the template itself.

Now, the virtual `Imported` function is called in `ASTImporter::Impor(Decl *)`,
but only once, when the `Decl` is imported.  One point of this refactor is to
separate responsibilities. The original `Imported()` had 3 responsibilities:
- notify subclasses when an import happened
- register the decl into `ImportedDecls`
- initialise the Decl (set attributes, etc)
Now all of these are in separate functions:
- `Imported`
- `MapImported`
- `InitializeImportedDecl`
I tried to check all the clients, I executed tests for `ExternalASTMerger.cpp`
and some unittests for lldb.

Reviewers: a.sidorin, balazske, xazax.hun, r.stahl

Subscribers: rnkovacs, dkrupp, cfe-commits

Differential Revision: https://reviews.llvm.org/D47632

llvm-svn: 336896

6 years agoFix -Wdocumentation warnings. NFCI.
Simon Pilgrim [Thu, 12 Jul 2018 09:10:55 +0000 (09:10 +0000)]
Fix -Wdocumentation warnings. NFCI.

llvm-svn: 336895

6 years ago[X86] Add UDIV by uniform/non-uniform constant tests
Simon Pilgrim [Thu, 12 Jul 2018 09:04:28 +0000 (09:04 +0000)]
[X86] Add UDIV by uniform/non-uniform constant tests

llvm-svn: 336894

6 years ago[mips] Mark standard encoded instructions as not being in MIPS16e
Simon Atanasyan [Thu, 12 Jul 2018 08:50:11 +0000 (08:50 +0000)]
[mips] Mark standard encoded instructions as not being in MIPS16e

Mark standard encoded instructions and pseudo "standard encoded"
as not being in MIPS16e by default.

Patch by Simon Dardis.

Differential revision: https://reviews.llvm.org/D48379

llvm-svn: 336893

6 years ago[ELF] - Simplify code. NFC.
George Rimar [Thu, 12 Jul 2018 08:33:02 +0000 (08:33 +0000)]
[ELF] - Simplify code. NFC.

Just use getDataAs for taking sections contents.

llvm-svn: 336892

6 years ago[ELF] - Eliminate dead code. NFC.
George Rimar [Thu, 12 Jul 2018 08:12:08 +0000 (08:12 +0000)]
[ELF] - Eliminate dead code. NFC.

Code is dead because caller of the isDuplicateArmExidxSex
(https://github.com/llvm-mirror/lld/blob/master/ELF/Writer.cpp#L1446)

explicitly does not pass sentinel. So no reason to check it.

llvm-svn: 336891

6 years ago[clangd] Simplify logging wrapper after r336888
Sam McCall [Thu, 12 Jul 2018 08:00:21 +0000 (08:00 +0000)]
[clangd] Simplify logging wrapper after r336888

llvm-svn: 336890

6 years ago[X86] Remove i128 type from FR128 regclass.
Craig Topper [Thu, 12 Jul 2018 07:30:01 +0000 (07:30 +0000)]
[X86] Remove i128 type from FR128 regclass.

i128 isn't a legal type in our x86 implementation today. So remove this and the few patterns that used it until it becomes necessary.

llvm-svn: 336889

6 years ago[Support] Require llvm::Error passed to formatv() to be wrapped in fmt_consume()
Sam McCall [Thu, 12 Jul 2018 07:11:28 +0000 (07:11 +0000)]
[Support] Require llvm::Error passed to formatv() to be wrapped in fmt_consume()

Summary:
Someone must be responsible for handling an Error. When formatv takes
ownership of an Error, the formatv_object destructor must take care of this.

Passing an error by value to formatv() is not considered explicit enough to mark
the error as handled (see D49013), so we require callers to use a format adapter
to confirm this intent.

Reviewers: zturner

Subscribers: llvm-commits, lhames

Differential Revision: https://reviews.llvm.org/D49170

llvm-svn: 336888

6 years agoFix few typos in comments (write access test commit)
Stefan Granitz [Thu, 12 Jul 2018 06:41:41 +0000 (06:41 +0000)]
Fix few typos in comments (write access test commit)

llvm-svn: 336887

6 years ago[Dominators] Add isUpdateLazy() method to the DomTreeUpdater
Chijun Sima [Thu, 12 Jul 2018 04:08:14 +0000 (04:08 +0000)]
[Dominators] Add isUpdateLazy() method to the DomTreeUpdater

Summary:
Previously, when people need to deal with DTU with different UpdateStrategy using different actions, they need to
```
if (DTU.getUpdateStrategy() == DomTreeUpdater::UpdateStrategy::Lazy) {
  ...
}
if (DTU.getUpdateStrategy() == DomTreeUpdater::UpdateStrategy::Eager) {
  ...
}
```
After the patch, they can avoid code patterns above
```
if (DTU.isUpdateLazy()){
  ...
}
if (!DTU.isUpdateLazy()){
  ...
}
```

Reviewers: kuhar, brzycki, dmgreen

Reviewed By: kuhar

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D49056

llvm-svn: 336886

6 years agoRemove the unused m_signal member variable, but leave the code that gets it out of...
Eric Christopher [Thu, 12 Jul 2018 03:52:46 +0000 (03:52 +0000)]
Remove the unused m_signal member variable, but leave the code that gets it out of the json.

llvm-svn: 336885

6 years agoRemove unused variable m_header as it hasn't been used since it was
Eric Christopher [Thu, 12 Jul 2018 03:52:45 +0000 (03:52 +0000)]
Remove unused variable m_header as it hasn't been used since it was
added in 2016.

llvm-svn: 336884

6 years ago[X86] Remove patterns and ISD nodes for the old scalar FMA intrinsic lowering.
Craig Topper [Thu, 12 Jul 2018 03:42:41 +0000 (03:42 +0000)]
[X86] Remove patterns and ISD nodes for the old scalar FMA intrinsic lowering.

We now use llvm.fma.f32/f64 or llvm.x86.fmadd.f32/f64 intrinsics that use scalar types rather than vector types. So we don't these special ISD nodes that operate on the lowest element of a vector.

llvm-svn: 336883

6 years ago[coff] remove_dots from /PDBPATH but not /PDBALTPATH.
Zachary Turner [Thu, 12 Jul 2018 03:22:39 +0000 (03:22 +0000)]
[coff] remove_dots from /PDBPATH but not /PDBALTPATH.

This more closely matches the behavior of link.exe, and also
simplifies the code slightly.

llvm-svn: 336882

6 years ago[InstSimplify] simplify add instruction if two operands are negative
Chen Zheng [Thu, 12 Jul 2018 03:06:04 +0000 (03:06 +0000)]
[InstSimplify] simplify add instruction if two operands are negative

Differential Revision: https://reviews.llvm.org/D49216

llvm-svn: 336881

6 years agoTurns out that wide literals U"xxx" and u"xxx" are c++11 and later.
Marshall Clow [Thu, 12 Jul 2018 02:55:01 +0000 (02:55 +0000)]
Turns out that wide literals U"xxx" and u"xxx" are c++11 and later.

llvm-svn: 336880

6 years ago[AsmParser] Fix inconsistent declaration parameter name
Fangrui Song [Thu, 12 Jul 2018 02:03:53 +0000 (02:03 +0000)]
[AsmParser] Fix inconsistent declaration parameter name

llvm-svn: 336879

6 years ago[XRay][compiler-rt] Fixup: require x86_64 for profiling mode tests
Dean Michael Berris [Thu, 12 Jul 2018 01:54:29 +0000 (01:54 +0000)]
[XRay][compiler-rt] Fixup: require x86_64 for profiling mode tests

This constrains the build environments we are testing/supporting for the runtime
tests until we can be sure xray works in more platforms.

llvm-svn: 336878

6 years agoTemporarily revert "Recommit r328307: [IPSCCP] Use constant range information for...
Eric Christopher [Thu, 12 Jul 2018 01:53:21 +0000 (01:53 +0000)]
Temporarily revert "Recommit r328307: [IPSCCP] Use constant range information for comparisons of parameters." as it's causing miscompiles.

A testcase was provided in the original review thread.

This reverts commit r336098.

llvm-svn: 336877

6 years ago[x86] Fix another trivial bug in x86 flags copy lowering that has been
Chandler Carruth [Thu, 12 Jul 2018 01:43:21 +0000 (01:43 +0000)]
[x86] Fix another trivial bug in x86 flags copy lowering that has been
there for a long time.

The boolean tracking whether we saw a kill of the flags was supposed to
be per-block we are scanning and instead was outside that loop and never
cleared. It requires a quite contrived test case to hit this as you have
to have multiple levels of successors and interleave them with kills.
I've included such a test case here.

This is another bug found testing SLH and extracted to its own focused
patch.

llvm-svn: 336876

6 years ago[X86] Add patterns to use VMOVSS/SD zero masking for scalar f32/f64 select with zero.
Craig Topper [Thu, 12 Jul 2018 00:54:40 +0000 (00:54 +0000)]
[X86] Add patterns to use VMOVSS/SD zero masking for scalar f32/f64 select with zero.

These showed up in some of the upgraded FMA code. We really need to improve these test cases more, but this helps for now.

llvm-svn: 336875

6 years ago[x86] Fix EFLAGS copy lowering to correctly handle walking past uses in
Chandler Carruth [Thu, 12 Jul 2018 00:52:50 +0000 (00:52 +0000)]
[x86] Fix EFLAGS copy lowering to correctly handle walking past uses in
multiple successors where some of the uses end up killing the EFLAGS
register.

There was a bug where rather than skipping to the next basic block
queued up with uses once we saw a kill, we stopped processing the blocks
entirely. =/

Test case produces completely nonsensical code w/o this tiny fix.

This was found testing Speculative Load Hardening and split out of that
work.

Differential Revision: https://reviews.llvm.org/D49211

llvm-svn: 336874

6 years ago[coff] Remove dots in path pointing to PDB file.
Zachary Turner [Thu, 12 Jul 2018 00:44:15 +0000 (00:44 +0000)]
[coff] Remove dots in path pointing to PDB file.

Some Microsoft tools (e.g. new versions of WPA) fail when the
COFF Debug Directory contains a path to the PDB that contains
dots, such as D:\foo\./bar.pdb.  Remove dots before writing this
path.

This fixes pr38126.

llvm-svn: 336873

6 years ago[IRInterpreter] Fix misevaluation of interpretation expressions with `urem`.
Davide Italiano [Thu, 12 Jul 2018 00:31:04 +0000 (00:31 +0000)]
[IRInterpreter] Fix misevaluation of interpretation expressions with `urem`.

Scalar::MakeUnsigned was implemented incorrectly so it didn't
really change the sign of the type (leaving signed types signed).
This showed up as a misevaluation when IR-interpreting urem but
it's likely to arise in other contexts.

This commit fixes the definition, and adds a test to make
sure this won't regress in future (hopefully).

Fixes rdar://problem/42038760 and LLVM PR38076

Differential Revision:  https://reviews.llvm.org/D49155

llvm-svn: 336872

6 years ago[X86] Remove and autoupgrade the scalar fma intrinsics with masking.
Craig Topper [Thu, 12 Jul 2018 00:29:56 +0000 (00:29 +0000)]
[X86] Remove and autoupgrade the scalar fma intrinsics with masking.

This converts them to what clang is now using for codegen. Unfortunately, there seem to be a few kinks to work out still. I'll try to address with follow up patches.

llvm-svn: 336871

6 years agoAdd -allow-deprecated-dag-overlap to one of the experimental webassembly target tests.
Eric Christopher [Thu, 12 Jul 2018 00:01:51 +0000 (00:01 +0000)]
Add -allow-deprecated-dag-overlap to one of the experimental webassembly target tests.

llvm-svn: 336870

6 years agoIR: Skip -print-*-all after -print-*
Duncan P. N. Exon Smith [Wed, 11 Jul 2018 23:30:25 +0000 (23:30 +0000)]
IR: Skip -print-*-all after -print-*

This changes `-print-*` from transformation passes to analysis passes so
that `-print-after-all` and `-print-before-all` don't trigger.  This
avoids some redundant output.

Patch by Son Tuan Vu!

llvm-svn: 336869

6 years ago[CodeGen] Emit more precise AssertZext/AssertSext nodes.
Eli Friedman [Wed, 11 Jul 2018 23:26:35 +0000 (23:26 +0000)]
[CodeGen] Emit more precise AssertZext/AssertSext nodes.

This is marginally helpful for removing redundant extensions, and the
code is easier to read, so it seems like an all-around win. In the new
test i8-phi-ext.ll, we used to emit an AssertSext i8; now we emit an
AssertZext i2, which allows the extension of the return value to be
eliminated.

Differential Revision: https://reviews.llvm.org/D49004

llvm-svn: 336868

6 years agoFix deduction for conversion function templates converting to reference
Richard Smith [Wed, 11 Jul 2018 23:19:41 +0000 (23:19 +0000)]
Fix deduction for conversion function templates converting to reference
types.

We previously tried to use the "parameter is a reference" logic here,
but that doesn't work because it gets P and A backwards. Instead, add
a separate implementation of the "deduced A can be less qualified than
A" rule.

This also exposes that we incorrectly stripped cv-qualifiers from the
referent of A if it was a reference. However, if we don't do that, we
get the wrong results when P is a reference. In an attempt to match
what sanity dictates and what other implementations are doing, we now
remove cv-qualifiers from A and P unless both are reference types. I've
brought this up on the core reflector too, to try to get the standard
fixed.

llvm-svn: 336867

6 years ago[libc++] Take 2: Replace uses of _LIBCPP_ALWAYS_INLINE by _LIBCPP_INLINE_VISIBILITY
Louis Dionne [Wed, 11 Jul 2018 23:14:33 +0000 (23:14 +0000)]
[libc++] Take 2: Replace uses of _LIBCPP_ALWAYS_INLINE by _LIBCPP_INLINE_VISIBILITY

Summary:
We never actually mean to always inline a function -- all the uses of
the macro I could find are actually attempts to control the visibility
of symbols. This is better described by _LIBCPP_INLINE_VISIBILITY, which
is actually always defined the same.

This change is orthogonal to the decision of what we're actually going
to do with _LIBCPP_INLINE_VISIBILITY -- it just simplifies things by
having one canonical way of doing things.

Note that this commit had originally been applied in r336369 and then
reverted in r336382 because of unforeseen problems. Both of these problems
have now been fixed.

Reviewers: EricWF, mclow.lists

Subscribers: christof, dexonsmith, erikvanderpoel

Differential Revision: https://reviews.llvm.org/D48892

llvm-svn: 336866

6 years ago[windows] Fix out-of-memory failure in some of the tests
Stella Stamenova [Wed, 11 Jul 2018 22:47:35 +0000 (22:47 +0000)]
[windows] Fix out-of-memory failure in some of the tests

Summary: When ReadProcessMemory fails, bytes_read is sometimes set to a large garbage value. In that case, we need to set it back to zero before returning or the garbage value will be used to allocate memory later causing LLDB to crash with an out of memory error.

Reviewers: asmith, zturner

Reviewed By: zturner

Subscribers: zturner, asmith, stella.stamenova, llvm-commits

Differential Revision: https://reviews.llvm.org/D49159

llvm-svn: 336865

6 years ago[LoopIdiomRecognize] Don't convert a do while loop to ctlz.
Craig Topper [Wed, 11 Jul 2018 22:35:28 +0000 (22:35 +0000)]
[LoopIdiomRecognize] Don't convert a do while loop to ctlz.

This commit suppresses turning loops like this into "(bitwidth - ctlz(input))".

unsigned foo(unsigned input) {
  unsigned num = 0;
  do {
    ++num;
    input >>= 1;
  } while (input != 0);
  return num;
}

The loop version returns a value of 1 for both an input of 0 and an input of 1. Converting to a naive ctlz does not preserve that.

Theoretically we could do better if we checked isKnownNonZero or we could insert a select to handle the divergence. But until we have motivating cases for that, this is the easiest solution.

llvm-svn: 336864

6 years agoos_log: When there are multiple privacy annotations in the format
Akira Hatanaka [Wed, 11 Jul 2018 22:19:14 +0000 (22:19 +0000)]
os_log: When there are multiple privacy annotations in the format
string, choose the strictest one instead of the last.

Also fix an undefined behavior. Move the pointer update to a later point to
avoid adding StringRef::npos to the pointer.

rdar://problem/40706280

llvm-svn: 336863

6 years ago[LoopIdiomRecognize] Add a test case showing a loop we turn into ctlz that we shouldn't.
Craig Topper [Wed, 11 Jul 2018 22:17:26 +0000 (22:17 +0000)]
[LoopIdiomRecognize] Add a test case showing a loop we turn into ctlz that we shouldn't.

This loop executes one iteration without checking the input value. This produces a count of 1 for an input of 0 and 1. We are turning this into 32 - ctlz(n), but that returns 0 if n is 0.

llvm-svn: 336862

6 years agoAMDGPU/SI: Initialize InstrInfo before TargetLoweringInfo in GCNSubtarget
Tom Stellard [Wed, 11 Jul 2018 22:15:15 +0000 (22:15 +0000)]
AMDGPU/SI: Initialize InstrInfo before TargetLoweringInfo in GCNSubtarget

SITargetLowering queries SIInstrInfo in its constructor, so SIInstrInfo
must be initialized first.  This fixes msan buildbot failures and was
introduced by r336851.

llvm-svn: 336861

6 years ago[MemorySSA] Add APIs to move memory accesses between blocks, following CFG changes.
Alina Sbirlea [Wed, 11 Jul 2018 22:11:46 +0000 (22:11 +0000)]
[MemorySSA] Add APIs to move memory accesses between blocks, following CFG changes.

Summary:
The move APIs added in this patch will be used to update MemorySSA when CFG changes merge or split blocks, by moving memory accesses accordingly in MemorySSA's internal data structures.
[Split from D45299 for easier review]

Reviewers: george.burgess.iv

Subscribers: sanjoy, jlebar, Prazek, llvm-commits

Differential Revision: https://reviews.llvm.org/D48897

llvm-svn: 336860

6 years ago[FileCheck] Add -allow-deprecated-dag-overlap to another compiler-rt test
Joel E. Denny [Wed, 11 Jul 2018 22:07:31 +0000 (22:07 +0000)]
[FileCheck] Add -allow-deprecated-dag-overlap to another compiler-rt test

See https://reviews.llvm.org/D47106 for details.

llvm-svn: 336859

6 years agoTemporarily reverting.
Bill Wendling [Wed, 11 Jul 2018 21:47:55 +0000 (21:47 +0000)]
Temporarily reverting.

llvm-svn: 336858

6 years ago[NFC][InstCombine] Tests for x & (-1 >> y) != x -> x u> (-1 >> y) fold
Roman Lebedev [Wed, 11 Jul 2018 21:28:42 +0000 (21:28 +0000)]
[NFC][InstCombine] Tests for  x & (-1 >> y) != x  ->  x u> (-1 >> y)  fold

https://bugs.llvm.org/show_bug.cgi?id=38123
https://rise4fun.com/Alive/Rny

llvm-svn: 336857

6 years agoSame reversed ifdef happened twice. Test fix only, NFC to the library.
Marshall Clow [Wed, 11 Jul 2018 21:22:13 +0000 (21:22 +0000)]
Same reversed ifdef happened twice. Test fix only, NFC to the library.

llvm-svn: 336856

6 years agoFix a test #ifdef that was reversed. NFC to the library.
Marshall Clow [Wed, 11 Jul 2018 21:20:42 +0000 (21:20 +0000)]
Fix a test #ifdef that was reversed. NFC to the library.

llvm-svn: 336855

6 years agoAMDGPU: Remove duplicate call to initializeSubtargetDependencies()
Tom Stellard [Wed, 11 Jul 2018 21:12:03 +0000 (21:12 +0000)]
AMDGPU: Remove duplicate call to initializeSubtargetDependencies()

This was added in r336851.

llvm-svn: 336853

6 years agoFix determination of whether one set of cvr-qualifiers is compatible
Richard Smith [Wed, 11 Jul 2018 21:07:04 +0000 (21:07 +0000)]
Fix determination of whether one set of cvr-qualifiers is compatible
with another in template argument deduction.

We happened to typically get away with getting this wrong, because the
cases where we'd produce a bogus deduction were caught by the final
"deduced A is compatible with A" check.

llvm-svn: 336852

6 years agoAMDGPU: Refactor Subtarget classes
Tom Stellard [Wed, 11 Jul 2018 20:59:01 +0000 (20:59 +0000)]
AMDGPU: Refactor Subtarget classes

Summary:
This is a follow-up to r335942.
- Merge SISubtarget into AMDGPUSubtarget and rename to GCNSubtarget
- Rename AMDGPUCommonSubtarget to AMDGPUSubtarget
- Merge R600Subtarget::Generation and GCNSubtarget::Generation into
  AMDGPUSubtarget::Generation.

Reviewers: arsenm, jvesely

Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, javed.absar, llvm-commits

Differential Revision: https://reviews.llvm.org/D49037

llvm-svn: 336851

6 years ago[Documentation] Fix incorrect documentation references, new checks order in Release...
Eugene Zelenko [Wed, 11 Jul 2018 20:56:26 +0000 (20:56 +0000)]
[Documentation] Fix incorrect documentation references, new checks order in Release Notes

llvm-svn: 336850

6 years ago[Documentation] Link format and order of Clang-tidy changes in Release Notes
Eugene Zelenko [Wed, 11 Jul 2018 20:41:16 +0000 (20:41 +0000)]
[Documentation] Link format and order of Clang-tidy changes in Release Notes

llvm-svn: 336849

6 years agofinish: [FileCheck] Add -allow-deprecated-dag-overlap to failing llvm tests
Joel E. Denny [Wed, 11 Jul 2018 20:31:51 +0000 (20:31 +0000)]
finish: [FileCheck] Add -allow-deprecated-dag-overlap to failing llvm tests

Differential Revision: https://reviews.llvm.org/D47171

This contains the portions of that patch that could not be committed
using the git monorepo because of dos line ending problems.

llvm-svn: 336848

6 years ago[FileCheck] Don't permit overlapping CHECK-DAG
Joel E. Denny [Wed, 11 Jul 2018 20:27:27 +0000 (20:27 +0000)]
[FileCheck] Don't permit overlapping CHECK-DAG

That is, make CHECK-DAG skip matches that overlap the matches of any
preceding consecutive CHECK-DAG directives.  This change makes
CHECK-DAG more consistent with other directives, and there is evidence
it makes CHECK-DAG more intuitive and less error-prone.  See the RFC
discussion starting at:

  http://lists.llvm.org/pipermail/llvm-dev/2018-May/123010.html

Moreover, this behavior enables CHECK-DAG groups for unordered,
non-unique strings or patterns.  For example, it is useful for
verifying output or logs from a parallel program, such as the OpenMP
runtime.

This patch also implements the command-line option
-allow-deprecated-dag-overlap, which reverts CHECK-DAG to the old
overlapping behavior.  This option should not be used in new tests.
It is meant only for the existing tests that are broken by this change
and that need time to update.

See the following bugzilla issue for tracking of such tests:

  https://bugs.llvm.org/show_bug.cgi?id=37532

Patches to add -allow-deprecated-dag-overlap to those tests will
follow immediately.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D47106

llvm-svn: 336847

6 years ago[FileCheck] Add -allow-deprecated-dag-overlap to failing lldb tests
Joel E. Denny [Wed, 11 Jul 2018 20:27:05 +0000 (20:27 +0000)]
[FileCheck] Add -allow-deprecated-dag-overlap to failing lldb tests

See https://reviews.llvm.org/D47106 for details.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D49192

llvm-svn: 336846

6 years ago[FileCheck] Add -allow-deprecated-dag-overlap to failing compiler-rt tests
Joel E. Denny [Wed, 11 Jul 2018 20:26:44 +0000 (20:26 +0000)]
[FileCheck] Add -allow-deprecated-dag-overlap to failing compiler-rt tests

See https://reviews.llvm.org/D47106 for details.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D47326

llvm-svn: 336845

6 years ago[FileCheck] Add -allow-deprecated-dag-overlap to failing clang tests
Joel E. Denny [Wed, 11 Jul 2018 20:26:20 +0000 (20:26 +0000)]
[FileCheck] Add -allow-deprecated-dag-overlap to failing clang tests

See https://reviews.llvm.org/D47106 for details.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D47172

llvm-svn: 336844

6 years ago[FileCheck] Add -allow-deprecated-dag-overlap to failing llvm tests
Joel E. Denny [Wed, 11 Jul 2018 20:25:49 +0000 (20:25 +0000)]
[FileCheck] Add -allow-deprecated-dag-overlap to failing llvm tests

See https://reviews.llvm.org/D47106 for details.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D47171

This commit drops that patch's changes to:

  llvm/test/CodeGen/NVPTX/f16x2-instructions.ll
  llvm/test/CodeGen/NVPTX/param-load-store.ll

For some reason, the dos line endings there prevent me from commiting
via the monorepo.  A follow-up commit (not via the monorepo) will
finish the patch.

llvm-svn: 336843

6 years agoFix setting of empty implicit-section-name attribute
Petr Pavlu [Wed, 11 Jul 2018 20:17:54 +0000 (20:17 +0000)]
Fix setting of empty implicit-section-name attribute

Code in `CodeGenModule::SetFunctionAttributes()` could set an empty
attribute `implicit-section-name` on a function that is affected by
`#pragma clang text="section"`. This is incorrect because the attribute
should contain a valid section name. If the function additionally also
used `__attribute__((section("section")))` then this could result in
emitting the function in a section with an empty name.

The patch fixes the issue by removing the problematic code that sets
empty `implicit-section-name` from
`CodeGenModule::SetFunctionAttributes()` because it is sufficient to set
this attribute only from a similar code in `setNonAliasAttributes()`
when the function is emitted.

Differential Revision: https://reviews.llvm.org/D48916

llvm-svn: 336842

6 years agoRevert "[docs] As of binutils 2.21.51.0.2, ld.bfd supports plugins too, represent...
Teresa Johnson [Wed, 11 Jul 2018 20:08:32 +0000 (20:08 +0000)]
Revert "[docs] As of binutils 2.21.51.0.2, ld.bfd supports plugins too, represent this in docs"

This reverts commit r306102.

This change was made without any review, and has a couple of issues.
First, AFAIK we do not test the combination of the LLVM gold plugin with
ld.bfd. Second, the change removed documentation for how to build gold
and replaced it with instructions for building ld.bfd.

llvm-svn: 336841

6 years ago[NFC] typo
JF Bastien [Wed, 11 Jul 2018 19:51:40 +0000 (19:51 +0000)]
[NFC] typo

llvm-svn: 336840

6 years agoRemove ppc64 BE XFAILs now that gcov profiling works, after starting a clean
Zaara Syeda [Wed, 11 Jul 2018 19:17:43 +0000 (19:17 +0000)]
Remove ppc64 BE XFAILs now that gcov profiling works, after starting a clean
build this time.

llvm-svn: 336839

6 years agogold: Add ability to toggle function/data sections
Bill Wendling [Wed, 11 Jul 2018 19:13:26 +0000 (19:13 +0000)]
gold: Add ability to toggle function/data sections

Some programs (e.g. Linux) aren't able to handle function/data sections when
LTO is used. Thus they need a way to disable it. That can be done with these
plugin options:

    -plugin-opt=-function-sections=0
    -plugin-opt=-data-sections=0

llvm-svn: 336838

6 years ago[DebugInfo] Fix getPreviousSibling after r336823
Fangrui Song [Wed, 11 Jul 2018 19:09:37 +0000 (19:09 +0000)]
[DebugInfo] Fix getPreviousSibling after r336823

llvm-svn: 336837

6 years ago[NFC] Replace usage of QualType.getTypePtr()-> with operator->
Erich Keane [Wed, 11 Jul 2018 19:09:21 +0000 (19:09 +0000)]
[NFC] Replace usage of QualType.getTypePtr()-> with operator->

llvm-svn: 336836

6 years ago[analyzer] Track multiple raw pointer symbols in DanglingInternalBufferChecker.
Reka Kovacs [Wed, 11 Jul 2018 19:08:02 +0000 (19:08 +0000)]
[analyzer] Track multiple raw pointer symbols in DanglingInternalBufferChecker.

Previously, the checker only tracked one raw pointer symbol for each
container object. But member functions returning a pointer to the
object's inner buffer may be called on the object several times. These
pointer symbols are now collected in a set inside the program state map
and thus all of them is checked for use-after-free problems.

Differential Revision: https://reviews.llvm.org/D49057

llvm-svn: 336835

6 years ago[InstCombine] Fold x & (-1 >> y) == x to x u<= (-1 >> y)
Roman Lebedev [Wed, 11 Jul 2018 19:05:04 +0000 (19:05 +0000)]
[InstCombine] Fold  x & (-1 >> y) == x  to  x u<= (-1 >> y)

Summary:
https://bugs.llvm.org/show_bug.cgi?id=38123

This pattern will be produced by Implicit Integer Truncation sanitizer,
https://reviews.llvm.org/D48958
https://bugs.llvm.org/show_bug.cgi?id=21530
in unsigned case, therefore it is probably a good idea to improve it.

https://rise4fun.com/Alive/Rny
^ there are more opportunities for folds, i will follow up with them afterwards.

Caveat: this somehow exposes a missing opportunities
in `test/Transforms/InstCombine/icmp-logical.ll`
It seems, the problem is in `foldLogOpOfMaskedICmps()` in `InstCombineAndOrXor.cpp`.
But i'm not quite sure what is wrong, because it calls `getMaskedTypeForICmpPair()`,
which calls `decomposeBitTestICmp()` which should already work for these cases...
As @spatel notes in https://reviews.llvm.org/D49179#1158760,
that code is a rather complex mess, so we'll let it slide.

Reviewers: spatel, craig.topper

Reviewed By: spatel

Subscribers: yamauchi, majnemer, t.p.northover, llvm-commits

Differential Revision: https://reviews.llvm.org/D49179

llvm-svn: 336834

6 years agoRevert r336830: [FileCheck] Don't permit overlapping CHECK-DAG
Joel E. Denny [Wed, 11 Jul 2018 19:03:00 +0000 (19:03 +0000)]
Revert r336830: [FileCheck] Don't permit overlapping CHECK-DAG

Companion patches are failing to commit, and this patch alone breaks
many tests.

llvm-svn: 336833

6 years agoQuick fix for some Windows bots
Paul Robinson [Wed, 11 Jul 2018 18:51:15 +0000 (18:51 +0000)]
Quick fix for some Windows bots

llvm-svn: 336832

6 years agoRevert "[VirtualFileSystem] InMemoryFileSystem::status: Return a Status with the...
Eric Liu [Wed, 11 Jul 2018 18:43:07 +0000 (18:43 +0000)]
Revert "[VirtualFileSystem] InMemoryFileSystem::status: Return a Status with the requested name"

This reverts commit r336807. This breaks users of
ClangTool::mapVirtualFile. Will try to investigate a fix. See also the
discussion on https://reviews.llvm.org/D48903

llvm-svn: 336831

6 years ago[FileCheck] Don't permit overlapping CHECK-DAG
Joel E. Denny [Wed, 11 Jul 2018 18:42:58 +0000 (18:42 +0000)]
[FileCheck] Don't permit overlapping CHECK-DAG

That is, make CHECK-DAG skip matches that overlap the matches of any
preceding consecutive CHECK-DAG directives.  This change makes
CHECK-DAG more consistent with other directives, and there is evidence
it makes CHECK-DAG more intuitive and less error-prone.  See the RFC
discussion starting at:

  http://lists.llvm.org/pipermail/llvm-dev/2018-May/123010.html

Moreover, this behavior enables CHECK-DAG groups for unordered,
non-unique strings or patterns.  For example, it is useful for
verifying output or logs from a parallel program, such as the OpenMP
runtime.

This patch also implements the command-line option
-allow-deprecated-dag-overlap, which reverts CHECK-DAG to the old
overlapping behavior.  This option should not be used in new tests.
It is meant only for the existing tests that are broken by this change
and that need time to update.

See the following bugzilla issue for tracking of such tests:

  https://bugs.llvm.org/show_bug.cgi?id=37532

Patches to add -allow-deprecated-dag-overlap to those tests will
follow immediately.

Reviewed By: probinson

Differential Revision: https://reviews.llvm.org/D47106

llvm-svn: 336830

6 years agoRevert "[llvm-objdump] Add -demangle (-C) option"
Paul Semel [Wed, 11 Jul 2018 18:09:52 +0000 (18:09 +0000)]
Revert "[llvm-objdump] Add -demangle (-C) option"

This reverts commit 3a44ccd156e0edd2e89226f8ed63928e227900bb.
This reverts commit d5cfc836bb5552e20507d3612d13ff66ff9e36a0.

llvm-svn: 336829

6 years ago[X86] Remove patterns for inserting a load into a zero vector.
Craig Topper [Wed, 11 Jul 2018 18:09:04 +0000 (18:09 +0000)]
[X86] Remove patterns for inserting a load into a zero vector.

We can instead block the load folding isProfitableToFold. Then isel will emit a register->register move for the zeroing part and a separate load. The PostProcessISelDAG should be able to remove the register->register move.

This saves us patterns and fixes the fact that we only had unaligned load patterns. The test changes show places where we should have been using an aligned load.

llvm-svn: 336828

6 years ago[TargetTransformInfo] Add pow2 analysis for scalar constants
Simon Pilgrim [Wed, 11 Jul 2018 17:51:27 +0000 (17:51 +0000)]
[TargetTransformInfo] Add pow2 analysis for scalar constants

Add ConstantInt analysis to getOperandInfo so we get more realistic div/rem expansion costs comparable to the vector costs.

llvm-svn: 336827

6 years agoAlso search BitcodeFiles for exclude-lib symbols
Yi Kong [Wed, 11 Jul 2018 17:45:28 +0000 (17:45 +0000)]
Also search BitcodeFiles for exclude-lib symbols

Archives created with ThinLTO are bitcodes, they also need to be searched for excluded symbols.

Differential Revision: https://reviews.llvm.org/D48857

llvm-svn: 336826

6 years agoAMDGPU/NFC: Use already available explicit kernarg
Konstantin Zhuravlyov [Wed, 11 Jul 2018 17:27:17 +0000 (17:27 +0000)]
AMDGPU/NFC: Use already available explicit kernarg
size instead of calculating it again when filling
out the metadata.

llvm-svn: 336825

6 years agoAllow specifying an exit code for the 'quit' command
Raphael Isemann [Wed, 11 Jul 2018 17:18:01 +0000 (17:18 +0000)]
Allow specifying an exit code for the 'quit' command

Summary:
This patch adds the possibility to specify an exit code when calling quit.
We accept any int, even though it depends on the user what happens if the int is
out of the range of what the operating system supports as exit codes.

Fixes rdar://problem/38452312

Reviewers: davide, jingham, clayborg

Reviewed By: jingham

Subscribers: clayborg, jingham, lldb-commits

Differential Revision: https://reviews.llvm.org/D48659

llvm-svn: 336824

6 years ago[DebugInfo] Make children iterator bidirectional
Jonas Devlieghere [Wed, 11 Jul 2018 17:11:11 +0000 (17:11 +0000)]
[DebugInfo] Make children iterator bidirectional

Make the DIE iterator bidirectional so we can move to the previous
sibling of a DIE.

Differential revision: https://reviews.llvm.org/D49173

llvm-svn: 336823

6 years ago[InstSimplify] add/move tests for add folds; NFC
Sanjay Patel [Wed, 11 Jul 2018 16:52:18 +0000 (16:52 +0000)]
[InstSimplify] add/move tests for add folds; NFC

isKnownNegation() is currently proposed as part of D48754,
but it could be used to make InstSimplify stronger independently
of any abs() improvements.

llvm-svn: 336822

6 years agoFix llvm-objdump demangle test (added triple option)
Paul Semel [Wed, 11 Jul 2018 16:31:33 +0000 (16:31 +0000)]
Fix llvm-objdump demangle test (added triple option)

llvm-svn: 336821

6 years agoLink to the correct bug number about the Mac failure for instrprof-shared-gcov-flush...
Marco Castelluccio [Wed, 11 Jul 2018 15:44:15 +0000 (15:44 +0000)]
Link to the correct bug number about the Mac failure for instrprof-shared-gcov-flush.test.

llvm-svn: 336820

6 years agoRevert 336811, there are still some problems with the tests.
Zaara Syeda [Wed, 11 Jul 2018 15:37:19 +0000 (15:37 +0000)]
Revert 336811, there are still some problems with the tests.

llvm-svn: 336819

6 years ago[X86] Fix MayLoad/HasSideEffect flag for (V)MOVLPSrm instructions.
Andrea Di Biagio [Wed, 11 Jul 2018 15:27:50 +0000 (15:27 +0000)]
[X86] Fix MayLoad/HasSideEffect flag for (V)MOVLPSrm instructions.

Before revision 336728, the "mayLoad" flag for instruction (V)MOVLPSrm was
inferred directly from the "default" pattern associated with the instruction
definition.

r336728 removed special node X86Movlps, and all the patterns associated to it.
Now instruction (V)MOVLPSrm doesn't have a pattern associated to it, and the
'mayLoad/hasSideEffects' flags are left unset.

When the instruction info is emitted by tablegen, method
CodeGenDAGPatterns::InferInstructionFlags() sees that (V)MOVLPSrm doesn't have a
pattern, and flags are undefined. So, it conservatively sets the
"hasSideEffects" flag for it.

As a consequence, we were losing the 'mayLoad' flag, and we were gaining a
'hasSideEffect' flag in its place.
This patch fixes the issue (originally reported by Michael Holmen).

The mca tests show the differences in the instruction info flags.  Instructions
that were affected by this problem were: MOVLPSrm/VMOVLPSrm/VMOVLPSZ128rm.

Differential Revision: https://reviews.llvm.org/D49182

llvm-svn: 336818

6 years ago[AST] Fix for structural equivalence tests in rL336776.
Balazs Keri [Wed, 11 Jul 2018 15:26:26 +0000 (15:26 +0000)]
[AST] Fix for structural equivalence tests in rL336776.

llvm-svn: 336817

6 years ago[llvm-objdump] Add -demangle (-C) option
Paul Semel [Wed, 11 Jul 2018 15:25:39 +0000 (15:25 +0000)]
[llvm-objdump] Add -demangle (-C) option

Differential Revision: https://reviews.llvm.org/D49043

llvm-svn: 336816

6 years ago[ELF] - Simplify code. NFC.
George Rimar [Wed, 11 Jul 2018 15:23:33 +0000 (15:23 +0000)]
[ELF] - Simplify code. NFC.

This looks a bit simpler IMO.

llvm-svn: 336815

6 years ago[ELF] - Simplify. NFCI.
George Rimar [Wed, 11 Jul 2018 15:18:23 +0000 (15:18 +0000)]
[ELF] - Simplify. NFCI.

It does not look possible to end up with empty Sections
at this place. And this knowledge allows simplifying the code.

llvm-svn: 336814

6 years ago[ELF] - Add classof() member for ARMExidxSentinelSection.
George Rimar [Wed, 11 Jul 2018 15:11:13 +0000 (15:11 +0000)]
[ELF] - Add classof() member for ARMExidxSentinelSection.

Or code uses constructions like isa<ARMExidxSentinelSection>:
https://github.com/llvm-mirror/lld/blob/master/ELF/Writer.cpp#L1428
https://github.com/llvm-mirror/lld/blob/master/ELF/SyntheticSections.cpp#L2944

That is confusing, because without ARMExidxSentinelSection::classof()
these lines are equal to isa<SyntheticSection> and the code does not really do
the same what it expected to. I found no good way to break it though, but it is not nice.

Patch adds ARMExidxSentinelSection::classof().

llvm-svn: 336813

6 years ago[SLPVectorizer] Add initial alternate opcode support for cast instructions. (REAPPLIED)
Simon Pilgrim [Wed, 11 Jul 2018 15:05:10 +0000 (15:05 +0000)]
[SLPVectorizer] Add initial alternate opcode support for cast instructions. (REAPPLIED)

We currently only support binary instructions in the alternate opcode shuffles.

This patch is an initial attempt at adding cast instructions as well, this raises several issues that we probably want to address as we continue to generalize the alternate mechanism:

1 - Duplication of cost determination - we should probably add scalar/vector costs helper functions and get BoUpSLP::getEntryCost to use them instead of determining costs directly.
2 - Support alternate instructions with the same opcode (e.g. casts with different src types) - alternate vectorization of calls with different IntrinsicIDs will require this.
3 - Allow alternates to be a different instruction type - mixing binary/cast/call etc.
4 - Allow passthrough of unsupported alternate instructions - related to PR30787/D28907 'copyable' elements.

Reapplied with fix to only accept 2 different casts if they come from the same source type.

Differential Revision: https://reviews.llvm.org/D49135

llvm-svn: 336812

6 years agoRemove ppc64 BE XFAILs now that gcov profiling works.
Zaara Syeda [Wed, 11 Jul 2018 14:55:19 +0000 (14:55 +0000)]
Remove ppc64 BE XFAILs now that gcov profiling works.

llvm-svn: 336811

6 years ago[clangd] Uprank delcarations when "using q::name" is present in the main file
Kirill Bobyrev [Wed, 11 Jul 2018 14:49:49 +0000 (14:49 +0000)]
[clangd] Uprank delcarations when "using q::name" is present in the main file

Having `using qualified::name;` for some symbol is an important signal
for clangd code completion as the user is more likely to use such
symbol.  This patch helps to uprank the relevant symbols by saving
UsingShadowDecl in the new field of CodeCompletionResult and checking
whether the corresponding UsingShadowDecl is located in the main file
later in ClangD code completion routine. While the relative importance
of such signal is a subject to change in the future, this patch simply
bumps DeclProximity score to the value of 1.0 which should be enough for
now.

The patch was tested using

`$ ninja check-clang check-clang-tools`

No unexpected failures were noticed after running the relevant testsets.

Reviewers: sammccall, ioeric

Subscribers: MaskRay, jkorous, cfe-commits

Differential Revision: https://reviews.llvm.org/D49012

llvm-svn: 336810

6 years ago[SLPVectorizer] Ensure alternate/passthrough doesn't vectorize sdiv with undef elts
Simon Pilgrim [Wed, 11 Jul 2018 14:34:43 +0000 (14:34 +0000)]
[SLPVectorizer] Ensure alternate/passthrough doesn't vectorize sdiv with undef elts

llvm-svn: 336809

6 years ago[SLPVectorizer] Add some additional alternate cast tests
Simon Pilgrim [Wed, 11 Jul 2018 14:29:13 +0000 (14:29 +0000)]
[SLPVectorizer] Add some additional alternate cast tests

Initial attempt at D49135 failed as we weren't correctly handling casts with different source types.

llvm-svn: 336808

6 years ago[VirtualFileSystem] InMemoryFileSystem::status: Return a Status with the requested...
Simon Marchi [Wed, 11 Jul 2018 14:08:17 +0000 (14:08 +0000)]
[VirtualFileSystem] InMemoryFileSystem::status: Return a Status with the requested name

Summary:
InMemoryFileSystem::status behaves differently than
RealFileSystem::status.  The Name contained in the Status returned by
RealFileSystem::status will be the path as requested by the caller,
whereas InMemoryFileSystem::status returns the normalized path.

For example, when requested the status for "../src/first.h",
RealFileSystem returns a Status with "../src/first.h" as the Name.
InMemoryFileSystem returns "/absolute/path/to/src/first.h".

The reason for this change is that I want to make a unit test in the
clangd testsuite (where we use an InMemoryFileSystem) to reproduce a
bug I get with the clangd program (where a RealFileSystem is used).
This difference in behavior "hides" the bug in the unit test version.

In general, I guess it's good if InMemoryFileSystem works as much as
possible like RealFileSystem.

Doing so made the FileEntry::RealPathName value (assigned in
FileManager::getFile) wrong when using the InMemoryFileSystem.  That's
because it assumes that vfs::File::getName will always return the real
path.  I changed to to use FileSystem::getRealPath instead.

Subscribers: ilya-biryukov, ioeric, cfe-commits

Differential Revision: https://reviews.llvm.org/D48903

llvm-svn: 336807

6 years agoRevert rL336804: [SLPVectorizer] Add initial alternate opcode support for cast instru...
Simon Pilgrim [Wed, 11 Jul 2018 14:08:16 +0000 (14:08 +0000)]
Revert rL336804: [SLPVectorizer] Add initial alternate opcode support for cast instructions.

Reverting due to buildbot failures

llvm-svn: 336806

6 years agoRecommit r334887: [SmallSet] Add SmallSetIterator.
Florian Hahn [Wed, 11 Jul 2018 13:39:59 +0000 (13:39 +0000)]
Recommit r334887: [SmallSet] Add SmallSetIterator.

This version now uses the subset of is_trivially_XXX provided by
GCC 4.8 and llvm/Support/type_traits.h

llvm-svn: 336805

6 years ago[SLPVectorizer] Add initial alternate opcode support for cast instructions.
Simon Pilgrim [Wed, 11 Jul 2018 13:34:09 +0000 (13:34 +0000)]
[SLPVectorizer] Add initial alternate opcode support for cast instructions.

We currently only support binary instructions in the alternate opcode shuffles.

This patch is an initial attempt at adding cast instructions as well, this raises several issues that we probably want to address as we continue to generalize the alternate mechanism:

1 - Duplication of cost determination - we should probably add scalar/vector costs helper functions and get BoUpSLP::getEntryCost to use them instead of determining costs directly.
2 - Support alternate instructions with the same opcode (e.g. casts with different src types) - alternate vectorization of calls with different IntrinsicIDs will require this.
3 - Allow alternates to be a different instruction type - mixing binary/cast/call etc.
4 - Allow passthrough of unsupported alternate instructions - related to PR30787/D28907 'copyable' elements.

Differential Revision: https://reviews.llvm.org/D49135

llvm-svn: 336804