platform/upstream/llvm.git
7 years agodocs: Fix Sphinx detection with out-of-tree builds
Tom Stellard [Tue, 9 May 2017 11:18:03 +0000 (11:18 +0000)]
docs: Fix Sphinx detection with out-of-tree builds

Adapt to changes made in r302499.

llvm-svn: 302517

7 years agodocs: Fix Sphinx detection with out-of-tree builds
Tom Stellard [Tue, 9 May 2017 11:11:52 +0000 (11:11 +0000)]
docs: Fix Sphinx detection with out-of-tree builds

Adapt to changes made in r302499.

llvm-svn: 302516

7 years ago[Polly][PPCGCodeGen] OpenCL now gets kernel argument size from PPCG CodeGen
Siddharth Bhat [Tue, 9 May 2017 10:45:52 +0000 (10:45 +0000)]
[Polly][PPCGCodeGen] OpenCL now gets kernel argument size from PPCG CodeGen

Summary: PPCGCodeGeneration now attaches the size of the kernel launch parameters at the end of the parameter list. For the existing CUDA Runtime, this gets ignored, but the OpenCL Runtime knows to check for kernel-argument size at the end of the parameter list. (The resulting parameters list is twice as long. This has been accounted for in the corresponding test cases).

Reviewers: grosser, Meinersbur, bollu

Reviewed By: bollu

Subscribers: nemanjai, yaxunl, Anastasia, pollydev, llvm-commits

Tags: #polly

Differential Revision: https://reviews.llvm.org/D32961

llvm-svn: 302515

7 years agoIntroduce experimental generic intrinsics for horizontal vector reductions.
Amara Emerson [Tue, 9 May 2017 10:43:25 +0000 (10:43 +0000)]
Introduce experimental generic intrinsics for horizontal vector reductions.

- This change allows targets to opt-in to using them instead of the log2
  shufflevector algorithm.
- The SLP and Loop vectorizers have the common code to do shuffle reductions
  factored out into LoopUtils, and now have a unified interface for generating
  reductions regardless of the preference of the target. LoopUtils now uses TTI
  to determine what kind of reductions the target wants to handle.
- For CodeGen, basic legalization support is added.

Differential Revision: https://reviews.llvm.org/D30086

llvm-svn: 302514

7 years ago[X86] Clang option -fuse-init-array has no effect when generating for MCU target
Nikolai Bozhenov [Tue, 9 May 2017 10:14:03 +0000 (10:14 +0000)]
[X86] Clang option -fuse-init-array has no effect when generating for MCU target

Reviewers: Eugene.Zelenko, dschuff, craig.topper

Reviewed By: craig.topper

Subscribers: ahatanak, aaboud, DavidKreitzer, llvm-commits, cfe-commits

Differential Revision: https://reviews.llvm.org/D32543
Patch by AndreiGrischenko <andrei.l.grischenko@intel.com>

llvm-svn: 302513

7 years ago[MIPS] Add support to match more patterns for DINS instruction
Strahinja Petrovic [Tue, 9 May 2017 10:02:00 +0000 (10:02 +0000)]
[MIPS] Add support to match more patterns for DINS instruction

This patch adds support for recognizing patterns to match
DINS instruction.

Differential Revision: https://reviews.llvm.org/D31465

llvm-svn: 302512

7 years ago[ARM GlobalISel] Remove hand-written G_FADD selection
Diana Picus [Tue, 9 May 2017 08:32:42 +0000 (08:32 +0000)]
[ARM GlobalISel] Remove hand-written G_FADD selection

Remove the code selecting G_FADD - now that TableGen can handle more
opcodes, it's not needed anymore.

llvm-svn: 302511

7 years ago[ConstantRange] Rewrite shl to avoid repeated calls to getUnsignedMax and avoid creat...
Craig Topper [Tue, 9 May 2017 07:04:04 +0000 (07:04 +0000)]
[ConstantRange] Rewrite shl to avoid repeated calls to getUnsignedMax and avoid creating the min APInt until we're sure we need it. Use inplace shift operations.

llvm-svn: 302510

7 years ago[ConstantRange] Combine the two adds max+1 in lshr into a single addition.
Craig Topper [Tue, 9 May 2017 07:04:02 +0000 (07:04 +0000)]
[ConstantRange] Combine the two adds max+1 in lshr into a single addition.

llvm-svn: 302509

7 years ago[ConstantRange] Use APInt::isNullValue in place of comparing with 0. The compiler...
Craig Topper [Tue, 9 May 2017 05:01:29 +0000 (05:01 +0000)]
[ConstantRange] Use APInt::isNullValue in place of comparing with 0. The compiler should be able to generate slightly better code for the former. NFC

llvm-svn: 302508

7 years agoFix PR32638 : Make sure we switch Sema's CurContext to the substituted FunctionDecl...
Faisal Vali [Tue, 9 May 2017 04:17:15 +0000 (04:17 +0000)]
Fix PR32638 : Make sure we switch Sema's CurContext to the substituted FunctionDecl when instantiating the exception specification.

This fixes the bug: https://bugs.llvm.org/show_bug.cgi?id=32638

  int main()
  {
    [](auto x) noexcept(noexcept(x)) { } (0);
  }

In the above code, prior to this patch, when substituting into the noexcept expression, i.e. transforming the DeclRefExpr that represents 'x' - clang attempts to capture 'x' because Sema's CurContext is still pointing to the pattern FunctionDecl (i.e. the templated-decl set in FinishTemplateArgumentDeduction) which does not match the substituted 'x's DeclContext, which leads to an attempt to capture and an assertion failure.

We fix this by adjusting Sema's CurContext to point to the substituted FunctionDecl under which the noexcept specifier's argument should be transformed, and so the ParmVarDecl that 'x' refers to has the same declcontext and no capture is attempted.

I briefly investigated whether the SwitchContext should occur right after VisitMethodDecl creates the new substituted FunctionDecl, instead of only during instantiating the exception specification - but seeing no other code that seemed to rely on that, I decided to leave it just for the duration of the exception specification instantiation.

llvm-svn: 302507

7 years ago[Sema] Make typeof(OverloadedFunctionName) not a pointer.
George Burgess IV [Tue, 9 May 2017 04:06:24 +0000 (04:06 +0000)]
[Sema] Make typeof(OverloadedFunctionName) not a pointer.

We were sometimes doing a function->pointer conversion in
Sema::CheckPlaceholderExpr, which isn't the job of CheckPlaceholderExpr.

So, when we saw typeof(OverloadedFunctionName), where
OverloadedFunctionName referenced a name with only one function that
could have its address taken, we'd give back a function pointer type
instead of a function type. This is incorrect.

I kept the logic for doing the function pointer conversion in
resolveAndFixAddressOfOnlyViableOverloadCandidate because it was more
consistent with existing ResolveAndFix* methods.

llvm-svn: 302506

7 years ago[ODRHash] Loosen checks on typedefs.
Richard Trieu [Tue, 9 May 2017 03:24:34 +0000 (03:24 +0000)]
[ODRHash] Loosen checks on typedefs.

When a type in a class is from a typedef, only check the canonical type.  Skip
checking the intermediate underlying types.  This is in response to PR 32965

llvm-svn: 302505

7 years agoRevert "Don't add DBG_VALUE instructions for static allocas in dbg.declare"
Reid Kleckner [Tue, 9 May 2017 01:57:44 +0000 (01:57 +0000)]
Revert "Don't add DBG_VALUE instructions for static allocas in dbg.declare"

This reverts commit r302461.

It appears to be causing failures compiling gtest with debug info on the
Linux sanitizer bot. I was unable to reproduce the failure locally,
however.

llvm-svn: 302504

7 years ago[Sema][ObjC] Clean up possible null dereference.
Akira Hatanaka [Tue, 9 May 2017 01:54:51 +0000 (01:54 +0000)]
[Sema][ObjC] Clean up possible null dereference.

It appears that the code is actually dead since unbridged-cast
placeholder types are created by calling CastOperation::complete and
ImplicitCastExprs are never passed to it.

Spotted by Vedant Kumar.

rdar://problem/31542226

llvm-svn: 302503

7 years agoFix code section prefix for proper layout
Teresa Johnson [Tue, 9 May 2017 01:43:24 +0000 (01:43 +0000)]
Fix code section prefix for proper layout

Summary:
r284533 added hot and cold section prefixes based on profile
information, to enable grouping of hot/cold functions at link time.
However, it used "cold" as the prefix for cold sections, but gold only
recognizes "unlikely" (which is used by gcc for cold sections).
Therefore, cold sections were not properly being grouped. Switch to
using "unlikely"

Reviewers: danielcdh, davidxl

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32983

llvm-svn: 302502

7 years agodocs: Fix Sphinx detection with out-of-tree builds
Tom Stellard [Tue, 9 May 2017 01:43:22 +0000 (01:43 +0000)]
docs: Fix Sphinx detection with out-of-tree builds

Adapt to changes made in r302499.

llvm-svn: 302501

7 years agodocs: Fix Sphinx detection with out-of-tree builds
Tom Stellard [Tue, 9 May 2017 01:42:33 +0000 (01:42 +0000)]
docs: Fix Sphinx detection with out-of-tree builds

Adapt to changes made in r302499.

llvm-svn: 302500

7 years agoRevert "Revert "CMake: Move sphinx detection into AddSphinxTarget.cmake""
Tom Stellard [Tue, 9 May 2017 01:41:28 +0000 (01:41 +0000)]
Revert "Revert "CMake: Move sphinx detection into AddSphinxTarget.cmake""

This reverts commit r302054.

Re-commit now that I have fixes for clang/lld.

llvm-svn: 302499

7 years ago[libFuzzer] update docs on -print_coverage/-dump_coverage
Kostya Serebryany [Tue, 9 May 2017 01:34:27 +0000 (01:34 +0000)]
[libFuzzer] update docs on -print_coverage/-dump_coverage

llvm-svn: 302498

7 years agoFix Windows tests when __config_site is present.
Ben Craig [Tue, 9 May 2017 01:34:12 +0000 (01:34 +0000)]
Fix Windows tests when __config_site is present.
Previously, the force includes would complain about a missing _DEBUG symbol.
Now we dump macros before adding the force includes to the command line.
Now with proper newlines.

llvm-svn: 302497

7 years agoRevert "Fix Windows tests when __config_site is present."
Ben Craig [Tue, 9 May 2017 01:26:39 +0000 (01:26 +0000)]
Revert "Fix Windows tests when __config_site is present."
It's 2017, and line endings are still an issue.

llvm-svn: 302496

7 years ago[CodeGen][ObjC] Emit @objc_retain at -O0 for variables captured by
Akira Hatanaka [Tue, 9 May 2017 01:20:05 +0000 (01:20 +0000)]
[CodeGen][ObjC] Emit @objc_retain at -O0 for variables captured by
blocks.

r302270 made changes to avoid emitting clang.arc.use at -O0 and instead
emit @objc_release. We also have to emit @objc_retain for the captured
variable at -O0 to match the @objc_release instead of just storing the
pointer to the capture field.

llvm-svn: 302495

7 years ago[libFuzzer] make sure the input data is not overwritten in the fuzz target (if it...
Kostya Serebryany [Tue, 9 May 2017 01:17:29 +0000 (01:17 +0000)]
[libFuzzer] make sure the input data is not overwritten in the fuzz target (if it is -- report an error)

llvm-svn: 302494

7 years agoRevert "Use the frame index side table for byval and inalloca arguments"
Reid Kleckner [Tue, 9 May 2017 01:14:39 +0000 (01:14 +0000)]
Revert "Use the frame index side table for byval and inalloca arguments"

This reverts r302483 and it's follow up fix.

llvm-svn: 302493

7 years ago[XRay] Add __xray_customeevent(...) as a clang-supported builtin
Dean Michael Berris [Tue, 9 May 2017 00:45:40 +0000 (00:45 +0000)]
[XRay] Add __xray_customeevent(...) as a clang-supported builtin

Summary:
We define the `__xray_customeevent` builtin that gets translated to
IR calls to the correct intrinsic. The default implementation of this is
a no-op function. The codegen side of this follows the following logic:

- When `-fxray-instrument` is not provided in the driver, we elide all
calls to `__xray_customevent`.
- When `-fxray-instrument` is enabled and a function is marked as "never
instrumented", we elide all calls to `__xray_customevent` in that
function; if either marked as "always instrumented" or subject to
threshold-based instrumentation, we emit a call to the
`llvm.xray.customevent` intrinsic from LLVM for each
`__xray_customevent` occurrence in the function.

This change depends on D27503 (to land in LLVM first).

Reviewers: echristo, rsmith

Subscribers: mehdi_amini, pelikan, lrl, cfe-commits

Differential Revision: https://reviews.llvm.org/D30018

llvm-svn: 302492

7 years ago[Modules] Allow umbrella frameworks to define private submodules for subframeworks
Bruno Cardoso Lopes [Tue, 9 May 2017 00:41:38 +0000 (00:41 +0000)]
[Modules] Allow umbrella frameworks to define private submodules for subframeworks

In r298391 we fixed the umbrella framework model to work when submodules
named "Private" are used. This complements the work by allowing the
umbrella framework model to work in general.

rdar://problem/31790067

llvm-svn: 302491

7 years agoRename a method. NFC.
Vedant Kumar [Tue, 9 May 2017 00:12:33 +0000 (00:12 +0000)]
Rename a method. NFC.

llvm-svn: 302490

7 years agoXFAIL noexcept member function throw/catch test under GCC.
Eric Fiselier [Tue, 9 May 2017 00:11:02 +0000 (00:11 +0000)]
XFAIL noexcept member function throw/catch test under GCC.

I'm still not exactly sure why the test fails, but I suspect it's
a bug in GCC. More investigation needed.

llvm-svn: 302489

7 years agoFix GCC 7 test failures.
Eric Fiselier [Tue, 9 May 2017 00:00:00 +0000 (00:00 +0000)]
Fix GCC 7 test failures.

This patch fixes the test failures and unexpected passes that occur
when testing against GCC 7. Specifically:

* don't mark __gcd as always inline because it's a recursive function. GCC diagnoses this.
* don't XFAIL the aligned allocation tests. GCC 7 supports them but not the -faligned-allocation option.
* Work around gcc.gnu.org/PR78489 in variants constructors.

llvm-svn: 302488

7 years ago[APInt] Use default constructor instead of explicitly creating a 1-bit APInt in udiv...
Craig Topper [Mon, 8 May 2017 23:49:54 +0000 (23:49 +0000)]
[APInt] Use default constructor instead of explicitly creating a 1-bit APInt in udiv and urem. NFC

The default constructor does the same thing.

llvm-svn: 302487

7 years ago[APInt] Remove 'else' after 'return' in udiv and urem. NFC
Craig Topper [Mon, 8 May 2017 23:49:49 +0000 (23:49 +0000)]
[APInt] Remove 'else' after 'return' in udiv and urem. NFC

llvm-svn: 302486

7 years agoIgnore !associated metadata with null argument.
Evgeniy Stepanov [Mon, 8 May 2017 23:46:20 +0000 (23:46 +0000)]
Ignore !associated metadata with null argument.

Fixes PR32577 (comment 10).
Such metadata may legitimately appear in LTO.

llvm-svn: 302485

7 years agoRelax Dwarf filecheck test for 32-bit hosts
Reid Kleckner [Mon, 8 May 2017 23:27:52 +0000 (23:27 +0000)]
Relax Dwarf filecheck test for 32-bit hosts

llvm-svn: 302484

7 years agoUse the frame index side table for byval and inalloca arguments
Reid Kleckner [Mon, 8 May 2017 23:20:27 +0000 (23:20 +0000)]
Use the frame index side table for byval and inalloca arguments

Summary:
For inalloca functions, this is a very common code pattern:

  %argpack = type <{ i32, i32, i32 }>
  define void @f(%argpack* inalloca %args) {
  entry:
    %a = getelementptr inbounds %argpack, %argpack* %args, i32 0, i32 0
    %b = getelementptr inbounds %argpack, %argpack* %args, i32 0, i32 1
    %c = getelementptr inbounds %argpack, %argpack* %args, i32 0, i32 2
    tail call void @llvm.dbg.declare(metadata i32* %a, ... "a")
    tail call void @llvm.dbg.declare(metadata i32* %c, ... "b")
    tail call void @llvm.dbg.declare(metadata i32* %b, ... "c")

Even though these GEPs can be simplified to a constant offset from EBP
or RSP, we don't do that at -O0, and each GEP is computed into a
register. Registers used to compute argument addresses are typically
spilled and clobbered very quickly after the initial computation, so
live debug variable tracking loses information very quickly if we use
DBG_VALUE instructions.

This change moves processing of dbg.declare between argument lowering
and basic block isel, so that we can ask if an argument has a frame
index or not. If the argument lives in a register as is the case for
byval arguments on some targets, then we don't put it in the side table
and during ISel we emit DBG_VALUE instructions.

Reviewers: aprantl

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32980

llvm-svn: 302483

7 years agoAdd basic test case for -instnamer
Sanjoy Das [Mon, 8 May 2017 23:18:46 +0000 (23:18 +0000)]
Add basic test case for -instnamer

llvm-svn: 302482

7 years ago[InstNamer] Use range-for
Sanjoy Das [Mon, 8 May 2017 23:18:43 +0000 (23:18 +0000)]
[InstNamer] Use range-for

llvm-svn: 302481

7 years ago[InstNamer] Don't check type of arguments (they're never void)
Sanjoy Das [Mon, 8 May 2017 23:18:39 +0000 (23:18 +0000)]
[InstNamer] Don't check type of arguments (they're never void)

llvm-svn: 302480

7 years agoDelete trailing whitespace
Sanjoy Das [Mon, 8 May 2017 23:18:36 +0000 (23:18 +0000)]
Delete trailing whitespace

llvm-svn: 302479

7 years agoUpdate testcase for upstream LLVM changes.
Adrian Prantl [Mon, 8 May 2017 22:44:00 +0000 (22:44 +0000)]
Update testcase for upstream LLVM changes.

llvm-svn: 302476

7 years ago[InstCombine] add tests from D32285 to show current problems; NFC
Sanjay Patel [Mon, 8 May 2017 22:33:20 +0000 (22:33 +0000)]
[InstCombine] add tests from D32285 to show current problems; NFC

llvm-svn: 302475

7 years agoRefactor <locale> RAII guards to aid upcoming Windows locale changes.
Eric Fiselier [Mon, 8 May 2017 22:02:43 +0000 (22:02 +0000)]
Refactor <locale> RAII guards to aid upcoming Windows locale changes.

Previously <locale> used std::unique_ptr<remove_ptr<locale_t>, locale-mgmt-function>
as a scope guard for (A) creating new locales, and (B) setting the thread specific locale
in RAII safe manner.

However using unique_ptr has some problems, first it requires that locale_t is a pointer
type, which may not be the case (Windows will need a non-pointer locale_t type that emulates _locale_t).

The second problem is that users of the guards had to supply the locale management function to the custom
deleter at every call site. However these locale management functions don't exist natively Windows, making
a good Windows implementation of locale more difficult.

This patch creates distinct and simply RAII guards that replace unique_ptr. These guards handle calling
the correct locale management function so that callers don't have too. This simplification will
aid in upcoming Windows fixes.

llvm-svn: 302474

7 years agoFix possible loss of data warnings on amd64
Billy Robert O'Neal III [Mon, 8 May 2017 21:54:53 +0000 (21:54 +0000)]
Fix possible loss of data warnings on amd64

In T_size_size.pass, there is an explicit template argument to std::min to ask
for unsigned, to avoid type deduction errors. However, C1XX' warnings still
hate this use, because a 64 bit value (a size_t) is being passed to a function
accepting an unsigned (a 32 bit value).

Instead, change the tests to pass around std::size_t instances, and explicitly
narrow when constructing the string type under test. This also allows
removal of explicit template arguments to std::min.

llvm-svn: 302473

7 years agoResolve integer overflow warnings in GCD and LCM tests
Billy Robert O'Neal III [Mon, 8 May 2017 21:52:05 +0000 (21:52 +0000)]
Resolve integer overflow warnings in GCD and LCM tests

lcm.pass.cpp:
19: Update headers to that actually used in the test.
41: test0 was triggering narrowing warnings for all callers, because the
inputs were always ints, but some of the explicit template arguments were
smaller than that. Instead, have this function accept ints and static_cast
explicitly to the types we want before calling std::lcm.
47: Replace unnecessary ternary.
55: Use foo_t instead of typename foo<>::type
111/116: intX_t were not std::qualified but only <cfoo> headers were included.
141: C1XX has a bug where it interprets 2147483648 as unsigned int. Then the
negation trips "negation of unsigned value, result still unsigned" warnings.
Perma-workaround this issue by saying INT_MIN, which better documents the
intended behavior and avoids triggering warnings on C1XX.

gcd.pass.cpp:
Same changes as lcm.pass.cpp but for GCD.

llvm-svn: 302472

7 years agoAdd const to "DWARFDie &Die" in a few functions as they can't change the DWARFDie.
Greg Clayton [Mon, 8 May 2017 21:29:17 +0000 (21:29 +0000)]
Add const to "DWARFDie &Die" in a few functions as they can't change the DWARFDie.

llvm-svn: 302471

7 years agoFix typo
Eugene Zemtsov [Mon, 8 May 2017 21:20:53 +0000 (21:20 +0000)]
Fix typo

llvm-svn: 302470

7 years agoMake it illegal for two Functions to point to the same DISubprogram
Adrian Prantl [Mon, 8 May 2017 21:17:08 +0000 (21:17 +0000)]
Make it illegal for two Functions to point to the same DISubprogram

As recently discussed on llvm-dev [1], this patch makes it illegal for
two Functions to point to the same DISubprogram and updates
FunctionCloner to also clone the debug info of a function to conform
to the new requirement. To simplify the implementation it also factors
out the creation of inlineAt locations from the Inliner into a
general-purpose utility in DILocation.

[1] http://lists.llvm.org/pipermail/llvm-dev/2017-May/112661.html
<rdar://problem/31926379>

Differential Revision: https://reviews.llvm.org/D32975

llvm-svn: 302469

7 years ago[Driver] Don't enable -fsanitize-use-after-scope when ASan is disabled
Vedant Kumar [Mon, 8 May 2017 21:11:55 +0000 (21:11 +0000)]
[Driver] Don't enable -fsanitize-use-after-scope when ASan is disabled

When enabling any sanitizer, -fsanitize-use-after-scope is enabled by
default. This doesn't actually turn ASan on, because we've been getting
lucky and there are extra checks in BackendUtil that stop this from
happening.

However, this has been causing a behavior change: extra lifetime markers
are emitted in some cases where they aren't needed or expected.

llvm-svn: 302468

7 years agoSTLExtras: Fix enumerate() documentation
Matthias Braun [Mon, 8 May 2017 21:10:58 +0000 (21:10 +0000)]
STLExtras: Fix enumerate() documentation

llvm-svn: 302467

7 years agoFix typo "veify" to "verify".
Greg Clayton [Mon, 8 May 2017 20:53:00 +0000 (20:53 +0000)]
Fix typo "veify" to "verify".

llvm-svn: 302466

7 years ago[InstCombine] add folds for not-of-shift-right
Sanjay Patel [Mon, 8 May 2017 20:49:59 +0000 (20:49 +0000)]
[InstCombine] add folds for not-of-shift-right

This is another step towards getting rid of dyn_castNotVal,
so we can recommit:
https://reviews.llvm.org/rL300977

As the tests show, we were missing the lshr case for constants
and both ashr/lshr vector splat folds. The ashr case with constant
was being performed inefficiently in 2 steps. It's also possible
there was a latent bug in that case because we can't do that fold
if the constant is positive:
http://rise4fun.com/Alive/Bge

llvm-svn: 302465

7 years ago[PartialInlining] Capture by reference rather than by value.
Davide Italiano [Mon, 8 May 2017 20:44:01 +0000 (20:44 +0000)]
[PartialInlining] Capture by reference rather than by value.

llvm-svn: 302464

7 years agoIf we are building a module, and we read a second description of the same
Richard Smith [Mon, 8 May 2017 20:30:47 +0000 (20:30 +0000)]
If we are building a module, and we read a second description of the same
module from a different module map, ignore it.

This happens during builds of preprocessed modules (where it is harmless).

llvm-svn: 302463

7 years agoARM: use divmod libcalls on embedded MachO platforms too.
Tim Northover [Mon, 8 May 2017 20:00:14 +0000 (20:00 +0000)]
ARM: use divmod libcalls on embedded MachO platforms too.

The separated libcalls are implemented in terms of __divmodsi4 and __udivmodsi4
anyway, so we should always use them if possible.

llvm-svn: 302462

7 years agoDon't add DBG_VALUE instructions for static allocas in dbg.declare
Reid Kleckner [Mon, 8 May 2017 19:58:15 +0000 (19:58 +0000)]
Don't add DBG_VALUE instructions for static allocas in dbg.declare

Summary:
An llvm.dbg.declare of a static alloca is always added to the
MachineFunction dbg variable map, so these values are entirely
redundant. They survive all the way through codegen to be ignored by
DWARF emission.

Effectively revert r113967

Two bugpoint-reduced test cases from 2012 broke as a result of this
change. Despite my best efforts, I haven't been able to rewrite the test
case using dbg.value. I'm not too concerned about the lost coverage
because these were reduced from the test-suite, which we still run.

Reviewers: aprantl, dblaikie

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32920

llvm-svn: 302461

7 years agoAdd some useful helper methods / operators to TypeIndex.
Zachary Turner [Mon, 8 May 2017 19:46:37 +0000 (19:46 +0000)]
Add some useful helper methods / operators to TypeIndex.

llvm-svn: 302460

7 years agoUpdate instructions for using the experimental monorepo
Reid Kleckner [Mon, 8 May 2017 19:45:57 +0000 (19:45 +0000)]
Update instructions for using the experimental monorepo

llvm-svn: 302459

7 years ago[BitVector] Make find_prev member function const.
Zachary Turner [Mon, 8 May 2017 19:45:55 +0000 (19:45 +0000)]
[BitVector] Make find_prev member function const.

NFC.

llvm-svn: 302458

7 years agoAdd a blurb to the release notes about the WeakVH -> WeakTrackingVH transition
Sanjoy Das [Mon, 8 May 2017 19:15:06 +0000 (19:15 +0000)]
Add a blurb to the release notes about the WeakVH -> WeakTrackingVH transition

llvm-svn: 302456

7 years ago[CodeView] Add support for random access type visitors.
Zachary Turner [Mon, 8 May 2017 18:38:43 +0000 (18:38 +0000)]
[CodeView] Add support for random access type visitors.

Previously type visitation was done strictly sequentially, and
TypeIndexes were computed by incrementing the TypeIndex of the
last visited record.  This works fine for situations like dumping,
but not when you want to visit types in random order.  For example,
in a debug session someone might lookup a symbol by name, find that
it has TypeIndex 10,000 and then want to go straight to TypeIndex
10,000.

In order to make this work, the visitation framework needs a mode
where it can plumb TypeIndices through the callback pipeline.  This
patch adds such a mode.  In doing so, it is necessary to provide
an alternative implementation of TypeDatabase that supports random
access, so that is done as well.

Nothing actually uses these random access capabilities yet, but
this will be done in subsequent patches.

Differential Revision: https://reviews.llvm.org/D32928

llvm-svn: 302454

7 years ago[AArch64][RegisterBankInfo] Change the default mapping of fp loads.
Quentin Colombet [Mon, 8 May 2017 18:16:31 +0000 (18:16 +0000)]
[AArch64][RegisterBankInfo] Change the default mapping of fp loads.

This fixes PR32550, in a way that does not imply running the greedy
mode at O0.

The fix consists in checking if a load is used by any floating point
instruction and if yes, we return a default mapping with FPR instead
of GPR.

llvm-svn: 302453

7 years ago[AArch64][RegisterBankInfo] Fix mapping cost for GPR.
Quentin Colombet [Mon, 8 May 2017 18:16:23 +0000 (18:16 +0000)]
[AArch64][RegisterBankInfo] Fix mapping cost for GPR.

In r292478, we changed the order of the enum that is referenced by
PMI_FirstXXX. This had the side effect of changing the cost of the
mapping of all the loads, instead of just the FPRs ones.

Reinstate the higher cost for all but GPR loads.
Note: This did not have any external visible effects:
- For Fast mode, the cost would have been higher, but we don't care
  because we don't try to use alternative mappings.
- For Greedy mode, the higher cost of the GPR loads, would have
  triggered the use of the supposedly alternative mapping, that
  would be in fact the same GPR mapping but with a lower cost.

llvm-svn: 302452

7 years ago[InstCombine] move/add tests for not(shr (not X), Y); NFC
Sanjay Patel [Mon, 8 May 2017 18:16:04 +0000 (18:16 +0000)]
[InstCombine] move/add tests for not(shr (not X), Y); NFC

llvm-svn: 302451

7 years ago[ARM] Use a Changed flag to avoid making a pass's return value dependent on a compare...
Craig Topper [Mon, 8 May 2017 18:02:51 +0000 (18:02 +0000)]
[ARM] Use a Changed flag to avoid making a pass's return value dependent on a compare with a Statistic object.

Statistic compile to always be 0 in release build so this compare would always return false. And in the debug builds Statistic are global variables and remember their values across pass runs. So this compare returns true anytime the pass runs after the first time it modifies something.

This was found after reviewing all usages of comparison operators on a Statistic object. We had some internal code that did a compare with a statistic that caused a mismatch in output between debug and release builds. So we did an audit out of paranoia.

llvm-svn: 302450

7 years ago[SCEV] Make setRange take ConstantRange by value instead of rvalue reference so we...
Craig Topper [Mon, 8 May 2017 17:39:08 +0000 (17:39 +0000)]
[SCEV] Make setRange take ConstantRange by value instead of rvalue reference so we don't force anything on the caller.

llvm-svn: 302449

7 years ago[SCEV] Don't use std::move on both inputs to APInt::operator+ or operator-. It might...
Craig Topper [Mon, 8 May 2017 17:39:01 +0000 (17:39 +0000)]
[SCEV] Don't use std::move on both inputs to APInt::operator+ or operator-. It might be confusing to the reader. NFC

llvm-svn: 302448

7 years agoConstantFold: Handle gep nonnull, undef as well
Daniel Berlin [Mon, 8 May 2017 17:37:33 +0000 (17:37 +0000)]
ConstantFold: Handle gep nonnull, undef as well

llvm-svn: 302447

7 years agoConstantFold: Fold getelementptr (i32, i32* null, i64 undef) to null.
Daniel Berlin [Mon, 8 May 2017 17:37:29 +0000 (17:37 +0000)]
ConstantFold: Fold  getelementptr (i32, i32* null, i64 undef) to null.
Transforms/IndVarSimplify/2011-10-27-lftrnull will fail if this regresses.
Transforms/GVN/PRE/2011-06-01-NonLocalMemdepMiscompile.ll has been changed to still test what it was
trying to test.

llvm-svn: 302446

7 years ago[X86][LWP] Add __LWP__ macro tests
Simon Pilgrim [Mon, 8 May 2017 17:25:48 +0000 (17:25 +0000)]
[X86][LWP] Add __LWP__ macro tests

Missed in rL302418

Differential Revision: https://reviews.llvm.org/D32770

llvm-svn: 302445

7 years ago[ValueTracking] Use KnownOnes to provide a better bound on known zeros for ctlz/cttz...
Craig Topper [Mon, 8 May 2017 17:22:34 +0000 (17:22 +0000)]
[ValueTracking] Use KnownOnes to provide a better bound on known zeros for ctlz/cttz intrinics

This patch uses KnownOnes of the input of ctlz/cttz to bound the value that can be returned from these intrinsics. This makes these intrinsics more similar to the handling for ctpop which already uses known bits to produce a similar bound.

Differential Revision: https://reviews.llvm.org/D32521

llvm-svn: 302444

7 years agoFix grammar in comment. NFC
Jonathan Roelofs [Mon, 8 May 2017 17:06:17 +0000 (17:06 +0000)]
Fix grammar in comment. NFC

llvm-svn: 302443

7 years ago[X86] Split test configurations. NFC.
Zvi Rackover [Mon, 8 May 2017 16:54:25 +0000 (16:54 +0000)]
[X86] Split test configurations. NFC.

Split test that includes reproducer for pr32967 to KNL and SKX.

llvm-svn: 302442

7 years agoSimplify orphan section positioning.
Rafael Espindola [Mon, 8 May 2017 16:49:20 +0000 (16:49 +0000)]
Simplify orphan section positioning.

The code following this one already considers every possible insertion
point for orphan sections, there is no point in sorting them before.

llvm-svn: 302441

7 years ago[AST] Fix copy&paste error in comment. NFC.
Malcolm Parsons [Mon, 8 May 2017 16:43:29 +0000 (16:43 +0000)]
[AST] Fix copy&paste error in comment. NFC.

llvm-svn: 302440

7 years ago[InstSimplify] fix typo; NFC
Sanjay Patel [Mon, 8 May 2017 16:35:02 +0000 (16:35 +0000)]
[InstSimplify] fix typo; NFC

llvm-svn: 302439

7 years ago[InstCombine] use local variable to reduce code duplication; NFCI
Sanjay Patel [Mon, 8 May 2017 16:33:42 +0000 (16:33 +0000)]
[InstCombine] use local variable to reduce code duplication; NFCI

llvm-svn: 302438

7 years ago[ValueTracking] Introduce a version of computeKnownBits that returns a KnownBits...
Craig Topper [Mon, 8 May 2017 16:22:48 +0000 (16:22 +0000)]
[ValueTracking] Introduce a version of computeKnownBits that returns a KnownBits struct. Begin using it to replace internal usages of ComputeSignBit

This introduces a new interface for computeKnownBits that returns the KnownBits object instead of requiring it to be pre-constructed and passed in by reference.

This is a much more convenient interface as it doesn't require the caller to figure out the BitWidth to pre-construct the object. It's so convenient that I believe we can use this interface to remove the special ComputeSignBit flavor of computeKnownBits.

As a step towards that idea, this patch replaces all of the internal usages of ComputeSignBit with this new interface. As you can see from the patch there were a couple places where we called ComputeSignBit which really called computeKnownBits, and then called computeKnownBits again directly. I've reduced those places to only making one call to computeKnownBits. I bet there are probably external users that do it too.

A future patch will update the external users and remove the ComputeSignBit interface. I'll also working on moving more locations to the KnownBits returning interface for computeKnownBits.

Differential Revision: https://reviews.llvm.org/D32848

llvm-svn: 302437

7 years ago[InstCombine/InstSimplify] add comments about code duplication; NFC
Sanjay Patel [Mon, 8 May 2017 16:21:55 +0000 (16:21 +0000)]
[InstCombine/InstSimplify] add comments about code duplication; NFC

llvm-svn: 302436

7 years ago[Sema] Fix typos handling in an overloadable call.
Anastasia Stulova [Mon, 8 May 2017 16:05:54 +0000 (16:05 +0000)]
[Sema] Fix typos handling in an overloadable call.

In C typos in arguments in a call of an overloadable function lead
to a failure of construction of CallExpr and following recovery does
not handle created delayed typos. This causes an assertion fail in
Sema::~Sema since Sema::DelayedTypos remains not empty.

The patch fixes that behavior by handling a call with arguments
having dependant types in the way that C++ does.

Differential Revision: https://reviews.llvm.org/D31764

Patch by Dmitry Borisenkov!

llvm-svn: 302435

7 years ago[InstCombine] add another test for PR32949; NFC
Sanjay Patel [Mon, 8 May 2017 15:58:57 +0000 (15:58 +0000)]
[InstCombine] add another test for PR32949; NFC

A patch for the InstSimplify variant of this bug is up for review here:
https://reviews.llvm.org/D32954

llvm-svn: 302434

7 years agoInstructionSimplify: Refactor foldIdentityShuffles. NFC.
Zvi Rackover [Mon, 8 May 2017 15:46:58 +0000 (15:46 +0000)]
InstructionSimplify: Refactor foldIdentityShuffles. NFC.

Summary:
Minor refactoring of foldIdentityShuffles() which allows the removal of a
ConstantDataVector::get() in SimplifyShuffleVectorInstruction.

Reviewers: spatel

Reviewed By: spatel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32955

Conflicts:
lib/Analysis/InstructionSimplify.cpp

llvm-svn: 302433

7 years agoFix comment typos.
Geoff Berry [Mon, 8 May 2017 15:33:08 +0000 (15:33 +0000)]
Fix comment typos.

llvm-svn: 302432

7 years ago[clang-tidy] Fix readability-implicit-bool-cast false positives
Alexander Kornienko [Mon, 8 May 2017 15:22:09 +0000 (15:22 +0000)]
[clang-tidy] Fix readability-implicit-bool-cast false positives

The patch makes the check treat binary conditional operator (`x ?: y`), `while`
and regular `for` loops as conditional statements for the purpose of
AllowConditional*Cast options.

llvm-svn: 302431

7 years ago[powerpc] Remove XFAIL for sanitizer_coverage_no_prune.cc on powerpc64
Bill Seurer [Mon, 8 May 2017 15:17:43 +0000 (15:17 +0000)]
[powerpc] Remove XFAIL for sanitizer_coverage_no_prune.cc on powerpc64

This test case works fine on powerpc64 (both BE and LE).

llvm-svn: 302430

7 years agoclang-tidy: add IgnoreMacros option to modernize-use-default-member-init
Miklos Vajna [Mon, 8 May 2017 15:13:31 +0000 (15:13 +0000)]
clang-tidy: add IgnoreMacros option to modernize-use-default-member-init

Summary:
And also enable it by default to be consistent with e.g.
modernize-use-using.

This helps e.g. when running this check on cppunit client code where the
macro is provided by the system, so there is no easy way to modify it.

Reviewers: alexfh, malcolm.parsons

Reviewed By: malcolm.parsons

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D32945

llvm-svn: 302429

7 years ago[clang-format] Convert AlignEscapedNewlinesLeft to an enum, adding
Daniel Jasper [Mon, 8 May 2017 15:08:00 +0000 (15:08 +0000)]
[clang-format] Convert AlignEscapedNewlinesLeft to an enum, adding
DontAlign

This converts the clang-format option AlignEscapedNewlinesLeft from a
boolean to an enum, named AlignEscapedNewlines, with options Left (prev.
true), Right (prev. false), and a new option DontAlign.

When set to DontAlign, the backslashes are placed just after the last token in each line:
  #define EXAMPLE \
    do { \
        int x = aaaaa; \
        int b; \
        int dddddddddd; \
    } while (0)

Patch by jtbandes. Thank you!

llvm-svn: 302428

7 years ago[clang-format] Don’t propagate AvoidBinPacking into argument
Daniel Jasper [Mon, 8 May 2017 15:07:52 +0000 (15:07 +0000)]
[clang-format] Don’t propagate AvoidBinPacking into argument
subexpressions

This is an attempt to fix the issue described in a recent email:
http://lists.llvm.org/pipermail/cfe-dev/2017-April/053632.html

Patch by jtbandes. Thank you!
Review: https://reviews.llvm.org/D32475

llvm-svn: 302427

7 years agoAdding reproducer for pr32967. NFC.
Zvi Rackover [Mon, 8 May 2017 14:47:32 +0000 (14:47 +0000)]
Adding reproducer for pr32967. NFC.

llvm-svn: 302426

7 years ago[clang-tidy] Ignore private =deleted methods in macros.
Alexander Kornienko [Mon, 8 May 2017 14:17:27 +0000 (14:17 +0000)]
[clang-tidy] Ignore private =deleted methods in macros.

modernize-use-equals-delete is extremely noisy in code using
DISALLOW_COPY_AND_ASSIGN-style macros and there's no easy way to automatically
fix the warning when macros are in play.

llvm-svn: 302425

7 years ago[X86][SSE] Improve combineLogicBlendIntoPBLENDV to use general masks.
Simon Pilgrim [Mon, 8 May 2017 14:16:39 +0000 (14:16 +0000)]
[X86][SSE] Improve combineLogicBlendIntoPBLENDV to use general masks.

Currently combineLogicBlendIntoPBLENDV can only match ASHR to detect sign splatting of a bit mask, this patch generalises this to use computeNumSignBits instead.

This is a first step in several things we can do to improve PBLENDV support:

 * Better matching of X86ISD::ANDNP patterns.
 * Handle floating point cases.
 * Better vector and bitcast support in computeNumSignBits.
 * Recognise that PBLENDV only uses the sign bit of the mask, we should be able strip away sign splats (ASHR, PCMPGT isNeg tests etc.).

Differential Revision: https://reviews.llvm.org/D32953

llvm-svn: 302424

7 years ago[Polly][GPUJIT] Fixed OpenCL 2.0 min requirement for Error codes
Siddharth Bhat [Mon, 8 May 2017 14:10:37 +0000 (14:10 +0000)]
[Polly][GPUJIT] Fixed OpenCL 2.0 min requirement for Error codes

Summary: Removed OpenCL error code identifiers introduced in version 2.0.

Reviewers: grosser, bollu

Reviewed By: bollu

Subscribers: yaxunl, Anastasia, pollydev, llvm-commits

Tags: #polly

Differential Revision: https://reviews.llvm.org/D32962

llvm-svn: 302423

7 years agoNormalize line endings. NFCI,
Simon Pilgrim [Mon, 8 May 2017 13:32:34 +0000 (13:32 +0000)]
Normalize line endings. NFCI,

llvm-svn: 302422

7 years agoFix Windows tests when __config_site is present.
Ben Craig [Mon, 8 May 2017 13:15:22 +0000 (13:15 +0000)]
Fix Windows tests when __config_site is present.
Previously, the force includes would complain about a missing _DEBUG symbol.
Now we dump macros before adding the force includes to the command line.

llvm-svn: 302421

7 years agoIR: Add a shufflevector mask commutation helper function. NFC.
Zvi Rackover [Mon, 8 May 2017 12:40:18 +0000 (12:40 +0000)]
IR: Add a shufflevector mask commutation helper function. NFC.

Summary:
Following up on Sanjay's suggetion in D32955, move this functionality
into ShuffleVectornstruction.

Reviewers: spatel, RKSimon

Reviewed By: RKSimon

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D32956

llvm-svn: 302420

7 years agoCorrect the attribute spelling for guarded_var and pt_guarded_var.
Aaron Ballman [Mon, 8 May 2017 12:39:17 +0000 (12:39 +0000)]
Correct the attribute spelling for guarded_var and pt_guarded_var.

Patch by Roman Lebedev.

llvm-svn: 302419

7 years ago[X86][LWP] Add clang support for LWP instructions.
Simon Pilgrim [Mon, 8 May 2017 12:09:45 +0000 (12:09 +0000)]
[X86][LWP] Add clang support for LWP instructions.

This patch adds support for the the LightWeight Profiling (LWP) instructions which are available on all AMD Bulldozer class CPUs (bdver1 to bdver4).

Differential Revision: https://reviews.llvm.org/D32770

llvm-svn: 302418

7 years ago[ARM][NEON] Add support for ISD::ABS lowering
Simon Pilgrim [Mon, 8 May 2017 10:37:34 +0000 (10:37 +0000)]
[ARM][NEON] Add support for ISD::ABS lowering

Update NEON int_arm_neon_vabs intrinsic to use the ISD::ABS opcode directly

Added constant folding tests.

Differential Revision: https://reviews.llvm.org/D32938

llvm-svn: 302417

7 years ago[ARM] Clear the constant pool cache on explicit .ltorg directives
Martin Storsjo [Mon, 8 May 2017 10:26:24 +0000 (10:26 +0000)]
[ARM] Clear the constant pool cache on explicit .ltorg directives

Multiple ldr pseudoinstructions with the same constant value will
reuse the same constant pool entry. However, if the constant pool
is explicitly flushed with a .ltorg directive, we should not try
to reference constants in the previous pool any longer, since they
may be out of range.

This fixes assembling hand-written assembler source which repeatedly
loads the same constant value, across a binary size larger than the
pc-relative fixup range for ldr instructions (4096 bytes). Such
assembler source already uses explicit .ltorg instructions to emit
constant pools with regular intervals. However if we try to reuse
constants emitted in earlier pools, they end up out of range.

This makes the output of the testcase match what binutils gas does
(prior to this patch, it would fail to assemble).

Differential Revision: https://reviews.llvm.org/D32847

llvm-svn: 302416

7 years ago[AARCH64][NEON] Add support for ISD::ABS lowering
Simon Pilgrim [Mon, 8 May 2017 10:25:18 +0000 (10:25 +0000)]
[AARCH64][NEON] Add support for ISD::ABS lowering

Update int_aarch64_neon_abs intrinsic to use the ISD::ABS opcode directly

Differential Revision: https://reviews.llvm.org/D32940

llvm-svn: 302415

7 years ago[ELF] - Set DF_STATIC_TLS flag for i386 target.
George Rimar [Mon, 8 May 2017 10:24:38 +0000 (10:24 +0000)]
[ELF] - Set DF_STATIC_TLS flag for i386 target.

This is PR32437.

DF_STATIC_TLS
If set in a shared object or executable, this flag instructs the
dynamic linker to reject attempts to load this file dynamically.
It indicates that the shared object or executable contains code
using a static thread-local storage scheme. Implementations need
not support any form of thread-local storage.

Patch checks if IE/LE relocations were used to check if code uses
static model. If so it sets the DF_STATIC_TLS flag.

Differential revision: https://reviews.llvm.org/D32354

llvm-svn: 302414