platform/upstream/llvm.git
8 years ago[X86][AVX] Provide SubVectorBroadcast fallback if load fold fails (PR29133)
Simon Pilgrim [Thu, 25 Aug 2016 12:45:16 +0000 (12:45 +0000)]
[X86][AVX] Provide SubVectorBroadcast fallback if load fold fails (PR29133)

Fix for PR29133, matching the approach that was taken for AVX1 scalar broadcasts.

llvm-svn: 279735

8 years agoIntroduce unittests.
Michael Kruse [Thu, 25 Aug 2016 12:36:15 +0000 (12:36 +0000)]
Introduce unittests.

Add the infrastructure for unittests to Polly and two simple tests for
conversion between isl_val and APInt. In addition, a build target
check-polly-unittests is added to run only the unittests but not the regression
tests.

Clang's unittest mechanism served as as a blueprint which then was adapted to
Polly.

Differential Revision: https://reviews.llvm.org/D23833

llvm-svn: 279734

8 years agoUse configure_lit_site_cfg instead of configure_file.
Michael Kruse [Thu, 25 Aug 2016 12:03:33 +0000 (12:03 +0000)]
Use configure_lit_site_cfg instead of configure_file.

configure_lit_site_cfg defines some more parameters that are used in
lit.site.cfg.in. configure_file would leave those empty. These additional
definitions seem to be unimportant for regression tests, but unittests do not
work without them.

In case of out-of-tree builds, define the additional parameters with default
values. These may not take all configuration parameters into account, as
configure_lit_site_cfg would.

llvm-svn: 279733

8 years agoGVN-hoist: fix hoistingFromAllPaths for loops (PR29034)
Sebastian Pop [Thu, 25 Aug 2016 11:55:47 +0000 (11:55 +0000)]
GVN-hoist: fix hoistingFromAllPaths for loops (PR29034)

It is invalid to hoist stores or loads if they are not executed on all paths
from the hoisting point to the exit of the function. In the testcase, there are
paths in the loop that do not execute the stores or the loads, and so hoisting
them within the loop is unsafe.

The problem is that the current implementation of hoistingFromAllPaths is
incomplete: it walks all blocks dominated by the hoisting point, and does not
return false when the loop contains a path on which the hoisted ld/st is
not executed.

Differential Revision: https://reviews.llvm.org/D23843

llvm-svn: 279732

8 years agoRemove trailing WS [NFC]
Aditya Kumar [Thu, 25 Aug 2016 11:52:38 +0000 (11:52 +0000)]
Remove trailing WS [NFC]

llvm-svn: 279731

8 years agoAdd LLVM libdir to library search path in out-of-tree builds.
Michael Kruse [Thu, 25 Aug 2016 11:28:52 +0000 (11:28 +0000)]
Add LLVM libdir to library search path in out-of-tree builds.

This previously was not required because in an out-of-tree build Polly would
only build libraries (LLVMPolly, libPolly, libPollyISL, libPollyPPCG), but no
executables where the libraries would be linked to. This will change when adding
unittests in a follow-up commit.

llvm-svn: 279730

8 years agoRemoved trailing whitespace. NFC.
George Rimar [Thu, 25 Aug 2016 10:39:04 +0000 (10:39 +0000)]
Removed trailing whitespace. NFC.

llvm-svn: 279729

8 years agoAlso warn if llvm-lit is not available.
Michael Kruse [Thu, 25 Aug 2016 10:35:22 +0000 (10:35 +0000)]
Also warn if llvm-lit is not available.

The program 'llvm-lit', like 'not' and 'FileCheck' are necessary for running
check-polly. Warn of any of the three is not in LLVM_INSTALL_ROOT/bin directory.

llvm-svn: 279728

8 years agoclang-format: [JS] nested and tagged template strings.
Martin Probst [Thu, 25 Aug 2016 10:13:21 +0000 (10:13 +0000)]
clang-format: [JS] nested and tagged template strings.

JavaScript template strings can be nested arbitrarily:

    foo = `text ${es.map(e => { return `<${e}>`; })} text`;

This change lexes nested template strings using a stack of lexer states to
correctly switch back to template string lexing on closing braces.

Also, reuse the same stack for the token-stashed logic.

Reviewers: djasper

Subscribers: cfe-commits, klimek

Differential Revision: https://reviews.llvm.org/D22431

llvm-svn: 279727

8 years ago[ELF] - Implemented --oformat binary option.
George Rimar [Thu, 25 Aug 2016 09:05:47 +0000 (09:05 +0000)]
[ELF] - Implemented --oformat binary option.

-oformat output-format
`-oformat' option can be used to specify the binary format for the output object file.

Patch implements binary format output type.

Differential revision: https://reviews.llvm.org/D23769

llvm-svn: 279726

8 years agogdb-remote: Make the sequence mutex non-recursive
Pavel Labath [Thu, 25 Aug 2016 08:34:57 +0000 (08:34 +0000)]
gdb-remote: Make the sequence mutex non-recursive

Summary:
This is a preparatory commit for D22914, where I'd like to replace this mutex by an R/W lock
(which is also not recursive). This required a couple of changes:
- The only caller of Read/WriteRegister, GDBRemoteRegisterContext class, was already acquiring
  the mutex, so these functions do not need to. All functions which now do not take a lock, take
  an lock argument instead, to remind the caller of this fact.
- GetThreadSuffixSupported() was being called from locked and unlocked contexts (including
  contexts where the process was running, and the call would fail if it did not have the result
  cached). I have split this into two functions, one which computes the thread suffix support and
  caches it (this one always takes the lock), and another, which returns the cached value (and
  never needs to take the lock). This feels quite natural as ProcessGdbRemote was already
  pre-caching this value at the start.

Reviewers: clayborg

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D23802

llvm-svn: 279725

8 years agoFix warnings preventing copy elision.
Pavel Labath [Thu, 25 Aug 2016 08:22:14 +0000 (08:22 +0000)]
Fix warnings preventing copy elision.

Summary:
Moving a temporary object prevents copy elision, which is exactly
what clang points out by warning about this pattern.

The fix is simply removal of std::move applied to temporary objects.

Reviewers: clayborg

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D23825
Author: Taras Tsugrii <ttsugrii@fb.com>

llvm-svn: 279724

8 years agoamdgcn: Fix return type of get_num_groups
Matt Arsenault [Thu, 25 Aug 2016 07:31:40 +0000 (07:31 +0000)]
amdgcn: Fix return type of get_num_groups

llvm-svn: 279723

8 years agoFix memory leaks in clang-offload-bundler
Vitaly Buka [Thu, 25 Aug 2016 07:21:34 +0000 (07:21 +0000)]
Fix memory leaks in clang-offload-bundler

Summary:
1. Pair removed from StringMap was not destroyed
2. ObjectFile had no owner

Reviewers: sfantao

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D23865

llvm-svn: 279722

8 years ago[X86] Simplify getOperandBias as a bit. NFC
Craig Topper [Thu, 25 Aug 2016 04:16:10 +0000 (04:16 +0000)]
[X86] Simplify getOperandBias as a bit. NFC

There's no reason for it to return a signed type. Just return the operand bias in each if instead of starting from 0 and adding in the 'if'.

llvm-svn: 279720

8 years ago[X86] Fix indentation per coding standards. NFC
Craig Topper [Thu, 25 Aug 2016 04:16:08 +0000 (04:16 +0000)]
[X86] Fix indentation per coding standards. NFC

llvm-svn: 279719

8 years agoFixed comment
Vitaly Buka [Thu, 25 Aug 2016 03:44:36 +0000 (03:44 +0000)]
Fixed comment

llvm-svn: 279718

8 years ago[asan] Disable CreateSigAltStack from Unix/Signals.inc for asan builds
Vitaly Buka [Thu, 25 Aug 2016 03:32:49 +0000 (03:32 +0000)]
[asan] Disable CreateSigAltStack from Unix/Signals.inc for asan builds

Summary: Asan fails to UnsetAlternateSignalStack if it set by Unix/Signals.inc

Reviewers: kcc

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D23864

llvm-svn: 279717

8 years agoIf the user has specified target.memory-module-load-level 'minimal'
Jason Molenda [Thu, 25 Aug 2016 02:33:09 +0000 (02:33 +0000)]
If the user has specified target.memory-module-load-level 'minimal'
and we couldn't find a dyld binary on the debug system, override
that setting and read dyld out of memory - we need to put an
internal breakpoint on dyld to register binaries being loaded or
unloaded; the debugger won't work right without dyld symbols.

<rdar://problem/27857025>

llvm-svn: 279704

8 years agoMake buildbots happy.
George Burgess IV [Thu, 25 Aug 2016 02:15:54 +0000 (02:15 +0000)]
Make buildbots happy.

"warning: extra ‘;’ [-Wpedantic]"

llvm-svn: 279703

8 years agoRemove a pointless LLVM_CONSTEXPR. NFC.
George Burgess IV [Thu, 25 Aug 2016 01:54:37 +0000 (01:54 +0000)]
Remove a pointless LLVM_CONSTEXPR. NFC.

llvm-svn: 279702

8 years agoTailDuplication: Don't pass MMI separately from MF. NFC
Kyle Butt [Thu, 25 Aug 2016 01:37:07 +0000 (01:37 +0000)]
TailDuplication: Don't pass MMI separately from MF. NFC

MMI must match the function passed, and MF has a handle on MMI. Use that instead
of accepting it as separate argument. No Functional Change.

llvm-svn: 279701

8 years agoTailDuplication: Save MF and reduce number of parameters. NFC
Kyle Butt [Thu, 25 Aug 2016 01:37:03 +0000 (01:37 +0000)]
TailDuplication: Save MF and reduce number of parameters. NFC

Save the function in the class, and then don't pass it around. This reduces the
number of parameters and makes calls to member functions simpler.
No Functional Change.

llvm-svn: 279700

8 years agoUpdate a comment.
George Burgess IV [Thu, 25 Aug 2016 01:29:55 +0000 (01:29 +0000)]
Update a comment.

r279696, which changed `LLVM_CONSTEXPR AliasAttr` to `const AliasAttr`,
made this comment make less sense.

llvm-svn: 279699

8 years agoMachineFunctionProperties/MIRParser: Rename AllVRegsAllocated->NoVRegs, compute it
Matthias Braun [Thu, 25 Aug 2016 01:27:13 +0000 (01:27 +0000)]
MachineFunctionProperties/MIRParser: Rename AllVRegsAllocated->NoVRegs, compute it

Rename AllVRegsAllocated to NoVRegs. This avoids the connotation of
running after register and simply describes that no vregs are used in
a machine function. With that we can simply compute the property and do
not need to dump/parse it in .mir files.

Differential Revision: http://reviews.llvm.org/D23850

llvm-svn: 279698

8 years ago[libFuzzer] simplify the code, NFC
Kostya Serebryany [Thu, 25 Aug 2016 01:25:03 +0000 (01:25 +0000)]
[libFuzzer] simplify the code, NFC

llvm-svn: 279697

8 years agoMake some LLVM_CONSTEXPR variables const. NFC.
George Burgess IV [Thu, 25 Aug 2016 01:05:08 +0000 (01:05 +0000)]
Make some LLVM_CONSTEXPR variables const. NFC.

This patch changes LLVM_CONSTEXPR variable declarations to const
variable declarations, since LLVM_CONSTEXPR expands to nothing if the
current compiler doesn't support constexpr. In all of the changed
cases, it looks like the code intended the variable to be const instead
of sometimes-constexpr sometimes-not.

llvm-svn: 279696

8 years agoFix some Clang-tidy modernize-use-using and Include What You Use warnings; other...
Eugene Zelenko [Thu, 25 Aug 2016 00:45:04 +0000 (00:45 +0000)]
Fix some Clang-tidy modernize-use-using and Include What You Use warnings; other minor fixes.

Differential revision: https://reviews.llvm.org/D23861

llvm-svn: 279695

8 years agoLazily load the ContextDecl for a lambda's DefinitionData, to fix a
Richard Smith [Thu, 25 Aug 2016 00:34:00 +0000 (00:34 +0000)]
Lazily load the ContextDecl for a lambda's DefinitionData, to fix a
deserialization cycle caused by the ContextDecl recursively importing members
of the lambda's closure type.

llvm-svn: 279694

8 years ago[Profile] Propagate branch metadata properly in instcombine
Xinliang David Li [Thu, 25 Aug 2016 00:26:32 +0000 (00:26 +0000)]
[Profile] Propagate branch metadata properly in instcombine

Differential Revision: http://reviews.llvm.org/D23590

llvm-svn: 279693

8 years agoStrip opencl.ocl.version metadata
Matt Arsenault [Thu, 25 Aug 2016 00:25:10 +0000 (00:25 +0000)]
Strip opencl.ocl.version metadata

This should be uniqued when linking, but right now it creates
a lot of metadata spam listing the same version. This should also
probably be reporting the compiled version of the user program,
which may differ from the library. Currently the library IR files report
1.0 while 1.1/1.2 are the default for user programs.

llvm-svn: 279692

8 years ago[Sema][Comments] Factor out function type loc logic. NFCI
Bruno Cardoso Lopes [Thu, 25 Aug 2016 00:22:08 +0000 (00:22 +0000)]
[Sema][Comments] Factor out function type loc logic. NFCI

This is in prepatation for @param TypeAliasTemplate support.

llvm-svn: 279691

8 years agoTest: Add REQUIRES: asserts to test that now requires stats.
Kyle Butt [Thu, 25 Aug 2016 00:06:52 +0000 (00:06 +0000)]
Test: Add REQUIRES: asserts to test that now requires stats.

Test was modified in r279670

llvm-svn: 279690

8 years agoRewrite the GetFileInSDK methods in PlatformRemoteiOS,
Jason Molenda [Wed, 24 Aug 2016 23:46:48 +0000 (23:46 +0000)]
Rewrite the GetFileInSDK methods in PlatformRemoteiOS,
PlatformRemoteAppleWatch, PlatformRemoteAppleTV and remove the
GetFileInSDKRoot method from those classes.

The rewrite uses the more modern FileSpec etc API to simplify,
and handles the case where an SDK Root is given to lldb with
the "/Symbols" directory name already appended.  The new version
will try appending "/Symbols" and "/Symbols.Internal" to the
sdk root directories, and will also try appending nothing to
the sdk root directory in case it's handed such an sdkroot.

<rdar://problem/28000054>

llvm-svn: 279688

8 years agoDebugInfo: Let -gsplit-dwarf and -gmlt compose if -fno-split-dwarf-inlining is used
David Blaikie [Wed, 24 Aug 2016 23:22:36 +0000 (23:22 +0000)]
DebugInfo: Let -gsplit-dwarf and -gmlt compose if -fno-split-dwarf-inlining is used

If the inline info is not duplicated into the skeleton CU, then there's
value in using -gsplit-dwarf and -gmlt together (to keep all those extra
subprograms out of the skeleton CU, while also producing smaller .dwo
files)

llvm-svn: 279687

8 years ago[libFuzzer] make a test more deterministic
Kostya Serebryany [Wed, 24 Aug 2016 23:10:17 +0000 (23:10 +0000)]
[libFuzzer] make a test more deterministic

llvm-svn: 279686

8 years ago[InstCombine] move foldICmpDivConstConst() contents to foldICmpDivConstant(); NFCI
Sanjay Patel [Wed, 24 Aug 2016 23:03:36 +0000 (23:03 +0000)]
[InstCombine] move foldICmpDivConstConst() contents to foldICmpDivConstant(); NFCI

There was no logic in foldICmpDivConstant, so no need for a separate function.
The code is directly copy/pasted, so further cleanups to follow.

llvm-svn: 279685

8 years agoThe patch improves ValueTracking on left shift with nsw flag.
Evgeny Stupachenko [Wed, 24 Aug 2016 23:01:33 +0000 (23:01 +0000)]
The patch improves ValueTracking on left shift with nsw flag.

Summary:
The patch fixes PR28946.

Reviewers: majnemer, sanjoy

Differential Revision: http://reviews.llvm.org/D23296

From: Li Huang
llvm-svn: 279684

8 years ago[WebAssembly] Change a comment line
Heejin Ahn [Wed, 24 Aug 2016 22:53:00 +0000 (22:53 +0000)]
[WebAssembly] Change a comment line

Test for commit access.

llvm-svn: 279683

8 years agoMIRYamlMapping cleanup
Matthias Braun [Wed, 24 Aug 2016 22:41:46 +0000 (22:41 +0000)]
MIRYamlMapping cleanup

Missed two lines got lost when cherry picking old commits to master.

llvm-svn: 279682

8 years ago[Hexagon] Check for block end when skipping debug instructions
Krzysztof Parzyszek [Wed, 24 Aug 2016 22:36:35 +0000 (22:36 +0000)]
[Hexagon] Check for block end when skipping debug instructions

llvm-svn: 279681

8 years agoMIRParser/MIRPrinter: Compute HasInlineAsm instead of printing/parsing it
Matthias Braun [Wed, 24 Aug 2016 22:34:06 +0000 (22:34 +0000)]
MIRParser/MIRPrinter: Compute HasInlineAsm instead of printing/parsing it

llvm-svn: 279680

8 years agoMissed a test in my last commit
Matthias Braun [Wed, 24 Aug 2016 22:32:11 +0000 (22:32 +0000)]
Missed a test in my last commit

llvm-svn: 279679

8 years ago[Hexagon] Change insertion of expand-condsets pass to avoid memory leaks
Krzysztof Parzyszek [Wed, 24 Aug 2016 22:27:36 +0000 (22:27 +0000)]
[Hexagon] Change insertion of expand-condsets pass to avoid memory leaks

llvm-svn: 279678

8 years ago[InstCombine] use m_APInt to allow icmp eq/ne (shr X, C2), C folds for splat constant...
Sanjay Patel [Wed, 24 Aug 2016 22:22:06 +0000 (22:22 +0000)]
[InstCombine] use m_APInt to allow icmp eq/ne (shr X, C2), C folds for splat constant vectors

llvm-svn: 279677

8 years agoMachineRegisterInfo/MIR: Initialize tracksSubRegLiveness early, do not print/parser it
Matthias Braun [Wed, 24 Aug 2016 22:17:45 +0000 (22:17 +0000)]
MachineRegisterInfo/MIR: Initialize tracksSubRegLiveness early, do not print/parser it

tracksSubRegLiveness only depends on the Subtarget and a cl::opt, there
is not need to change it or save/parse it in a .mir file.
Make the field const and move the initialization LiveIntervalAnalysis to the
MachineRegisterInfo constructor. Also cleanup some code and fix some
instances which better use MachineRegisterInfo::subRegLivenessEnabled() instead
of TargetSubtargetInfo::enableSubRegLiveness().

llvm-svn: 279676

8 years ago[CMake] Be more consistent about naming targets and components
Chris Bieneman [Wed, 24 Aug 2016 22:17:06 +0000 (22:17 +0000)]
[CMake] Be more consistent about naming targets and components

Summary:
The point of this patch is to have a consistent convention for naming build, check and install targets so that the targets can be constructed from the project name.

This change renames a bunch of CMake components and targets from libcxx to cxx. For each renamed target I've added a convenience target that matches the old target name and depends on the new target. This will preserve function of the old targets so that the change doesn't break the world. We can evaluate if it is worth removing the extra targets later.

Reviewers: EricWF

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D23699

llvm-svn: 279675

8 years ago[lsan] give a test a bit more stack -- it started failing after r279664 on the debian...
Kostya Serebryany [Wed, 24 Aug 2016 22:10:35 +0000 (22:10 +0000)]
[lsan] give a test a bit more stack -- it started failing after r279664 on the debian bot, hopefully this is a fix.

llvm-svn: 279674

8 years ago[Order Files] On Darwin use DTrace's oneshot probe
Chris Bieneman [Wed, 24 Aug 2016 22:09:46 +0000 (22:09 +0000)]
[Order Files] On Darwin use DTrace's oneshot probe

The oneshot probe only gets executed the first time the probe is hit in the process. For order file generation this is really all we care about.

llvm-svn: 279673

8 years agofix darwin_log test errors on macOS < 10.12
Todd Fiala [Wed, 24 Aug 2016 21:40:29 +0000 (21:40 +0000)]
fix darwin_log test errors on macOS < 10.12

The newer event-based tests I added neglected to do the
macOS 10.12 check in the setup.  This caused earlier macOS
test suite runs to attempt to compile code that doesn't exist.

llvm-svn: 279672

8 years agoCodeGen: If Convert blocks that would form a diamond when tail-merged.
Kyle Butt [Wed, 24 Aug 2016 21:34:27 +0000 (21:34 +0000)]
CodeGen: If Convert blocks that would form a diamond when tail-merged.

The following function currently relies on tail-merging for if
conversion to succeed. The common tail of cond_true and cond_false is
extracted, and this then forms a diamond pattern that can be
successfully if converted.

If this block does not get extracted, either because tail-merging is
disabled or the threshold is higher, we should still recognize this
pattern and if-convert it.

Fixed a regression in the original commit. Need to un-reverse branches after
reversing them, or other conversions go awry.

define i32 @t2(i32 %a, i32 %b) nounwind {
entry:
        %tmp1434 = icmp eq i32 %a, %b           ; <i1> [#uses=1]
        br i1 %tmp1434, label %bb17, label %bb.outer

bb.outer:               ; preds = %cond_false, %entry
        %b_addr.021.0.ph = phi i32 [ %b, %entry ], [ %tmp10, %cond_false ]
        %a_addr.026.0.ph = phi i32 [ %a, %entry ], [ %a_addr.026.0, %cond_false ]
        br label %bb

bb:             ; preds = %cond_true, %bb.outer
        %indvar = phi i32 [ 0, %bb.outer ], [ %indvar.next, %cond_true ]
        %tmp. = sub i32 0, %b_addr.021.0.ph
        %tmp.40 = mul i32 %indvar, %tmp.
        %a_addr.026.0 = add i32 %tmp.40, %a_addr.026.0.ph
        %tmp3 = icmp sgt i32 %a_addr.026.0, %b_addr.021.0.ph
        br i1 %tmp3, label %cond_true, label %cond_false

cond_true:              ; preds = %bb
        %tmp7 = sub i32 %a_addr.026.0, %b_addr.021.0.ph
        %tmp1437 = icmp eq i32 %tmp7, %b_addr.021.0.ph
        %indvar.next = add i32 %indvar, 1
        br i1 %tmp1437, label %bb17, label %bb

cond_false:             ; preds = %bb
        %tmp10 = sub i32 %b_addr.021.0.ph, %a_addr.026.0
        %tmp14 = icmp eq i32 %a_addr.026.0, %tmp10
        br i1 %tmp14, label %bb17, label %bb.outer

bb17:           ; preds = %cond_false, %cond_true, %entry
        %a_addr.026.1 = phi i32 [ %a, %entry ], [ %tmp7, %cond_true ], [ %a_addr.026.0, %cond_false ]
        ret i32 %a_addr.026.1
}

Without tail-merging or diamond-tail if conversion:
LBB1_1:                                 @ %bb
                                        @ =>This Inner Loop Header: Depth=1
        cmp     r0, r1
        ble     LBB1_3
@ BB#2:                                 @ %cond_true
                                        @   in Loop: Header=BB1_1 Depth=1
        subs    r0, r0, r1
        cmp     r1, r0
        it      ne
        cmpne   r0, r1
        bgt     LBB1_4
LBB1_3:                                 @ %cond_false
                                        @   in Loop: Header=BB1_1 Depth=1
        subs    r1, r1, r0
        cmp     r1, r0
        bne     LBB1_1
LBB1_4:                                 @ %bb17
        bx      lr

With diamond-tail if conversion, but without tail-merging:
@ BB#0:                                 @ %entry
        cmp     r0, r1
        it      eq
        bxeq    lr
LBB1_1:                                 @ %bb
                                        @ =>This Inner Loop Header: Depth=1
        cmp     r0, r1
        ite     le
        suble   r1, r1, r0
        subgt   r0, r0, r1
        cmp     r1, r0
        bne     LBB1_1
@ BB#2:                                 @ %bb17
        bx      lr

llvm-svn: 279671

8 years agoIfConversion: Rescan diamonds.
Kyle Butt [Wed, 24 Aug 2016 21:34:24 +0000 (21:34 +0000)]
IfConversion: Rescan diamonds.

The cost of predicating a diamond is only the instructions that are not shared
between the two branches. Additionally If a predicate clobbering instruction
occurs in the shared portion of the branches (e.g. a cond move), it may still
be possible to if convert the sub-cfg. This change handles these two facts by
rescanning the non-shared portion of a diamond sub-cfg to recalculate both the
predication cost and whether both blocks are pred-clobbering.

Fixed 2 bugs before recommitting. Branch instructions must be compared and found
identical before diamond conversion. Also, predicate-clobbering instructions in
the shared prefix disqualifies a potential diamond conversion. Includes tests
for both.

llvm-svn: 279670

8 years ago[StreamExecutor] Rename Executor to Device
Jason Henline [Wed, 24 Aug 2016 21:31:53 +0000 (21:31 +0000)]
[StreamExecutor] Rename Executor to Device

Summary: This more clearly describes what the class is.

Reviewers: jlebar

Subscribers: jprice, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D23851

llvm-svn: 279669

8 years agoDisable test under asan: it uses a lot of stack, and asan increases the
Richard Smith [Wed, 24 Aug 2016 21:30:00 +0000 (21:30 +0000)]
Disable test under asan: it uses a lot of stack, and asan increases the
per-frame stack usage enough to cause it to hit our stack limit. This is not
ideal; we should find a better way of dealing with this, such as increasing
our stack allocation when built with ASan.

llvm-svn: 279668

8 years agoPR29097: add an update record when we instantiate the default member
Richard Smith [Wed, 24 Aug 2016 21:25:37 +0000 (21:25 +0000)]
PR29097: add an update record when we instantiate the default member
initializer of an imported field.

llvm-svn: 279667

8 years ago[clang-tidy misc-move-const-arg] more specific messages + suggest alternative solution
Alexander Kornienko [Wed, 24 Aug 2016 21:23:24 +0000 (21:23 +0000)]
[clang-tidy misc-move-const-arg] more specific messages + suggest alternative solution

llvm-svn: 279666

8 years agoARM: don't diagnose cbz/cbnz to Thumb functions.
Tim Northover [Wed, 24 Aug 2016 21:21:29 +0000 (21:21 +0000)]
ARM: don't diagnose cbz/cbnz to Thumb functions.

A branch-distance to a Thumb function shouldn't be forced to be odd for
CBZ/CBNZ instructions because (assuming it's within range), it's going to be a
valid, even offset.

llvm-svn: 279665

8 years ago[sanitizer] re-apply r279572 and r279595 reverted in r279643: change the 64-bit alloc...
Kostya Serebryany [Wed, 24 Aug 2016 21:20:10 +0000 (21:20 +0000)]
[sanitizer] re-apply r279572 and r279595 reverted in r279643: change the 64-bit allocator to use a single array for free-d chunks instead of a lock-free linked list of tranfer batches. This change simplifies the code, makes the allocator more 'hardened', and will allow simpler code to release RAM to OS. This may also slowdown malloc stress tests due to lock contension, but I did not observe noticeable slowdown on various real multi-threaded benchmarks.

llvm-svn: 279664

8 years ago[Sema][Comments] Support @param with c++ 'using' keyword
Bruno Cardoso Lopes [Wed, 24 Aug 2016 21:11:43 +0000 (21:11 +0000)]
[Sema][Comments] Support @param with c++ 'using' keyword

Give appropriate warnings with -Wdocumentation for @param comments
that refer to function aliases defined with 'using'. Very similar
to typedef's behavior. This does not add support for
TypeAliasTemplateDecl yet.

Differential Revision: https://reviews.llvm.org/D23783

rdar://problem/27300695

llvm-svn: 279662

8 years ago[ubsan] fix the test to me more resistent against changes in the sanitizer allocator
Kostya Serebryany [Wed, 24 Aug 2016 21:03:28 +0000 (21:03 +0000)]
[ubsan] fix the test to me more resistent against changes in the sanitizer allocator

llvm-svn: 279661

8 years agoAMDGCN/SI: Implement readlane/readfirstlane intrinsics
Changpeng Fang [Wed, 24 Aug 2016 20:35:23 +0000 (20:35 +0000)]
AMDGCN/SI: Implement readlane/readfirstlane intrinsics

Summary:
  This patch implements readlane/readfirstlane intrinsics.
TODO: need to define a new register class to consider the case
that the source could be a vector register or M0.

Reviewed by:
  arsenm and tstellarAMD

Differential Revision:
  http://reviews.llvm.org/D22489

llvm-svn: 279660

8 years agoClang-tidy documentation style. Two Google checks are aliases.
Eugene Zelenko [Wed, 24 Aug 2016 20:05:36 +0000 (20:05 +0000)]
Clang-tidy documentation style. Two Google checks are aliases.

Differential revision: https://reviews.llvm.org/D23815

llvm-svn: 279659

8 years ago[StreamExecutor] Fix allocateDeviceMemory
Jason Henline [Wed, 24 Aug 2016 19:42:03 +0000 (19:42 +0000)]
[StreamExecutor] Fix allocateDeviceMemory

Summary:
The return value from PlatformExecutor::allocateDeviceMemory needs to be
converted from Expected<GlobalDeviceMemoryBase> to
Expected<GlobalDeviceMemory<T>> in Executor::allocateDeviceMemory.

A similar bug is also fixed for Executor::allocateHostMemory.

Thanks to jprice for identifying this bug.

Reviewers: jprice, jlebar

Subscribers: parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D23849

llvm-svn: 279658

8 years agoAdd %loadPolly to test command line.
Michael Kruse [Wed, 24 Aug 2016 19:12:48 +0000 (19:12 +0000)]
Add %loadPolly to test command line.

Required for out-of-tree builds of Polly.

llvm-svn: 279657

8 years agoamdgcn: Also correct get_local_size type for HSA
Matt Arsenault [Wed, 24 Aug 2016 19:11:52 +0000 (19:11 +0000)]
amdgcn: Also correct get_local_size type for HSA

llvm-svn: 279656

8 years agoUse isTargetMachO instead of isTargetDarwin.
Rafael Espindola [Wed, 24 Aug 2016 19:02:29 +0000 (19:02 +0000)]
Use isTargetMachO instead of isTargetDarwin.

llvm-svn: 279655

8 years ago[StreamExecutor] Clean up device copy comments
Jason Henline [Wed, 24 Aug 2016 18:56:26 +0000 (18:56 +0000)]
[StreamExecutor] Clean up device copy comments

Summary:
Consolidate Executor::synchronousCopy* and Stream::thenCopy* methods into
Doxygen method groups and combine all their comments into one section.

Also a "doc" target to the build files to use Doxygen to build the
documentation.

Reviewers: jlebar

Subscribers: jprice, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D23845

llvm-svn: 279654

8 years agoFix offload bundler tests so that diagnostic can start with caps.
Samuel Antao [Wed, 24 Aug 2016 18:52:18 +0000 (18:52 +0000)]
Fix offload bundler tests so that diagnostic can start with caps.

Windows require that.

llvm-svn: 279653

8 years ago[X86][SSE] Add MINSD/MAXSD/MINSS/MAXSS intrinsic scalar load folding support
Simon Pilgrim [Wed, 24 Aug 2016 18:40:53 +0000 (18:40 +0000)]
[X86][SSE] Add MINSD/MAXSD/MINSS/MAXSS intrinsic scalar load folding support

These are no different in load behaviour to the existing ADD/SUB/MUL/DIV scalar ops but were missing from isNonFoldablePartialRegisterLoad

llvm-svn: 279652

8 years agoDebugInfo: Add flag to CU to disable emission of inline debug info into the skeleton CU
David Blaikie [Wed, 24 Aug 2016 18:29:58 +0000 (18:29 +0000)]
DebugInfo: Add flag to CU to disable emission of inline debug info into the skeleton CU

In cases where .dwo/.dwp files are guaranteed to be available, skipping
the extra online (in the .o file) inline info can save a substantial
amount of space - see the original r221306 for more details there.

llvm-svn: 279651

8 years agoDebugInfo: Add flag to CU to disable emission of inline debug info into the skeleton CU
David Blaikie [Wed, 24 Aug 2016 18:29:49 +0000 (18:29 +0000)]
DebugInfo: Add flag to CU to disable emission of inline debug info into the skeleton CU

In cases where .dwo/.dwp files are guaranteed to be available, skipping
the extra online (in the .o file) inline info can save a substantial
amount of space - see the original r221306 for more details there.

llvm-svn: 279650

8 years ago[LV] Unify vector and scalar maps
Matthew Simpson [Wed, 24 Aug 2016 18:23:17 +0000 (18:23 +0000)]
[LV] Unify vector and scalar maps

This patch unifies the data structures we use for mapping instructions from the
original loop to their corresponding instructions in the new loop. Previously,
we maintained two distinct maps for this purpose: WidenMap and ScalarIVMap.
WidenMap maintained the vector values each instruction from the old loop was
represented with, and ScalarIVMap maintained the scalar values each scalarized
induction variable was represented with. With this patch, all values created
for the new loop are maintained in VectorLoopValueMap.

The change allows for several simplifications. Previously, when an instruction
was scalarized, we had to insert the scalar values into vectors in order to
maintain the mapping in WidenMap. Then, if a user of the scalarized value was
also scalar, we had to extract the scalar values from the temporary vector we
created. We now aovid these unnecessary scalar-to-vector-to-scalar conversions.
If a scalarized value is used by a scalar instruction, the scalar value is used
directly. However, if the scalarized value is needed by a vector instruction,
we generate the needed insertelement instructions on-demand.

A common idiom in several locations in the code (including the scalarization
code), is to first get the vector values an instruction from the original loop
maps to, and then extract a particular scalar value. This patch adds
getScalarValue for this purpose along side getVectorValue as an interface into
VectorLoopValueMap. These functions work together to return the requested
values if they're available or to produce them if they're not.

The mapping has also be made less permissive. Entries can be added to
VectorLoopValue map with the new initVector and initScalar functions.
getVectorValue has been modified to return a constant reference to the mapped
entries.

There's no real functional change with this patch; however, in some cases we
will generate slightly different code. For example, instead of an insertelement
sequence following the definition of an instruction, it will now precede the
first use of that instruction. This can be seen in the test case changes.

Differential Revision: https://reviews.llvm.org/D23169

llvm-svn: 279649

8 years ago[AArch64] Adjust the feature set for Exynos M1.
Evandro Menezes [Wed, 24 Aug 2016 18:17:30 +0000 (18:17 +0000)]
[AArch64] Adjust the feature set for Exynos M1.

Enable zero cycle zeroing.

llvm-svn: 279648

8 years ago[SCCP] Don't delete side-effecting instructions
Sanjoy Das [Wed, 24 Aug 2016 18:10:21 +0000 (18:10 +0000)]
[SCCP] Don't delete side-effecting instructions

I'm not sure if the `!isa<CallInst>(Inst) &&
!isa<TerminatorInst>(Inst))` bit is correct either, but this fixes the
case we know is broken.

llvm-svn: 279647

8 years ago[X86][SSE] Add support for combining VZEXT_MOVL target shuffles
Simon Pilgrim [Wed, 24 Aug 2016 18:07:53 +0000 (18:07 +0000)]
[X86][SSE] Add support for combining VZEXT_MOVL target shuffles

Includes adding more general support for the pattern: VZEXT_MOVL(VZEXT_LOAD(ptr)) -> VZEXT_LOAD(ptr)

This has unearthed a couple of latent poor codegen issues (MINSS/MAXSS scalar load folding and MOVDDUP/BROADCAST load folding patterns), which will be fixed shortly.

Its also reduced a couple of tests so that they no longer reach the instruction threshold necessary to be combined to PSHUFB (see PR26183).

llvm-svn: 279646

8 years agoamdgcn: Fix return type for get_global_size
Matt Arsenault [Wed, 24 Aug 2016 17:52:04 +0000 (17:52 +0000)]
amdgcn: Fix return type for get_global_size

llvm-svn: 279644

8 years agoRevert r279572 "[sanitizer] change the 64-bit..." because of failures on ubsan
Vitaly Buka [Wed, 24 Aug 2016 17:40:29 +0000 (17:40 +0000)]
Revert r279572 "[sanitizer] change the 64-bit..." because of failures on ubsan

This reverts commit r279572 and r279595.

llvm-svn: 279643

8 years ago[Hexagon] Enable subregister liveness tracking
Krzysztof Parzyszek [Wed, 24 Aug 2016 17:17:39 +0000 (17:17 +0000)]
[Hexagon] Enable subregister liveness tracking

llvm-svn: 279642

8 years agoclang-offload-bundler: Update libdeps.
NAKAMURA Takumi [Wed, 24 Aug 2016 17:05:48 +0000 (17:05 +0000)]
clang-offload-bundler: Update libdeps.

llvm-svn: 279641

8 years ago[StreamExecutor] Executor add synchronous methods
Jason Henline [Wed, 24 Aug 2016 16:58:20 +0000 (16:58 +0000)]
[StreamExecutor] Executor add synchronous methods

Summary:
Add Executor methods that block the host until completion. Since these
methods are host-synchronous, they don't require Stream arguments.

Reviewers: jlebar

Subscribers: jprice, parallel_libs-commits

Differential Revision: https://reviews.llvm.org/D23577

llvm-svn: 279640

8 years agofix typo "varaible"
Nico Weber [Wed, 24 Aug 2016 16:37:21 +0000 (16:37 +0000)]
fix typo "varaible"

llvm-svn: 279639

8 years agofix typo "varaibles"
Nico Weber [Wed, 24 Aug 2016 16:36:41 +0000 (16:36 +0000)]
fix typo "varaibles"

llvm-svn: 279638

8 years ago[Hexagon] Remove the utilization of IMPLICIT_DEFs from expand-condsets
Krzysztof Parzyszek [Wed, 24 Aug 2016 16:36:37 +0000 (16:36 +0000)]
[Hexagon] Remove the utilization of IMPLICIT_DEFs from expand-condsets

This is no longer necessary, because since r279625 the subregister
liveness properly accounts for read-undefs.

llvm-svn: 279637

8 years agofix typo 'varaible' in assert
Nico Weber [Wed, 24 Aug 2016 16:34:54 +0000 (16:34 +0000)]
fix typo 'varaible' in assert

llvm-svn: 279636

8 years agoAdd target REQUIRES directives to offload bundler test.
Samuel Antao [Wed, 24 Aug 2016 15:47:06 +0000 (15:47 +0000)]
Add target REQUIRES directives to offload bundler test.

llvm-svn: 279635

8 years ago[Driver][OpenMP][CUDA] Add capability to bundle object files in sections of the host...
Samuel Antao [Wed, 24 Aug 2016 15:39:07 +0000 (15:39 +0000)]
[Driver][OpenMP][CUDA] Add capability to bundle object files in sections of the host binary format.

Summary:
This patch adds the capability to bundle object files in sections of the host binary using a designated naming convention for these sections. This patch uses the functionality of the object reader already in the LLVM library to read bundled files, and invokes clang with the incremental linking options to create bundle files.

Bundling files involves creating an IR file with the contents of the bundle assigned as initializers of globals binded to the designated sections. This way the bundling implementation is agnostic of the host object format.

The features added by this patch were requested in the RFC discussion in  http://lists.llvm.org/pipermail/cfe-dev/2016-February/047547.html.

Reviewers: echristo, tra, jlebar, hfinkel, ABataev, Hahnfeld

Subscribers: mkuron, whchung, cfe-commits, andreybokhanko, Hahnfeld, arpith-jacob, carlo.bertolli, mehdi_amini, caomhin

Differential Revision: https://reviews.llvm.org/D21851

llvm-svn: 279634

8 years agoGlobalISel: fix cmp test to be in SSA form
Tim Northover [Wed, 24 Aug 2016 15:37:51 +0000 (15:37 +0000)]
GlobalISel: fix cmp test to be in SSA form

llvm-svn: 279633

8 years agoclang-offload-bundler - offload files bundling/unbundling tool
Samuel Antao [Wed, 24 Aug 2016 15:21:05 +0000 (15:21 +0000)]
clang-offload-bundler - offload files bundling/unbundling tool

Summary:
One of the goals of programming models that support offloading (e.g. OpenMP) is to enable users to offload with little effort, by annotating the code with a few pragmas. I'd also like to save users the trouble of changing their existent applications' build system. So having the compiler always return a single file instead of one for the host and each target even if the user is doing separate compilation is desirable.

This diff proposes a tool named clang-offload-bundler (happy to change the name if required) that is used to bundle files associated with the same user source file but different targets, or to unbundle a file into separate files associated with different targets.

This tool supports the driver support for OpenMP under review in http://reviews.llvm.org/D9888. The tool is used there to enable separate compilation, so that the very first action on input files that are not source files is a "unbundling action" and the very last non-linking action is a "bundling action".

The format of the bundled files is currently very simple: text formats are concatenated with comments that have a magic string and target identifying triple in between, and binary formats have a header that contains the triple and the offset and size of the code for host and each target.

The goal is to improve this tool in the future to deal with archive files so that each individual file in the archive is properly dealt with. We see that archives are very commonly used in current applications to combine separate compilation results. So I'm convinced users would enjoy this feature.

This tool can be used like this:

`clang-offload-bundler -targets=triple1,triple2 -type=ii -inputs=a.triple1.ii,a.triple2.ii -outputs=a.ii`

or

`clang-offload-bundler -targets=triple1,triple2 -type=ii -outputs=a.triple1.ii,a.triple2.ii -inputs=a.ii -unbundle`

I implemented the tool under clang/tools. Please let me know if something like this should live somewhere else.

This patch is prerequisite for http://reviews.llvm.org/D9888.

Reviewers: hfinkel, rsmith, echristo, chandlerc, tra, jlebar, ABataev, Hahnfeld

Subscribers: whchung, caomhin, andreybokhanko, arpith-jacob, carlo.bertolli, mehdi_amini, guansong, Hahnfeld, cfe-commits

Differential Revision: https://reviews.llvm.org/D13909

llvm-svn: 279632

8 years ago[ThinLTO/gold] Add caching support to gold-plugin
Teresa Johnson [Wed, 24 Aug 2016 15:11:47 +0000 (15:11 +0000)]
[ThinLTO/gold] Add caching support to gold-plugin

Summary:
With support now in the new LTO API for caching (r279576), add
optional ThinLTO caching in the gold-plugin.

Reviewers: mehdi_amini

Subscribers: mehdi_amini, llvm-commits

Differential Revision: https://reviews.llvm.org/D23836

llvm-svn: 279631

8 years ago[X86][SSE] Regenerate scalar math load folding tests for 32 and 64 bit targets
Simon Pilgrim [Wed, 24 Aug 2016 15:07:11 +0000 (15:07 +0000)]
[X86][SSE] Regenerate scalar math load folding tests for 32 and 64 bit targets

llvm-svn: 279630

8 years agoAMDGPU : Add V_SAD_U32 instruction pattern.
Wei Ding [Wed, 24 Aug 2016 14:59:47 +0000 (14:59 +0000)]
AMDGPU : Add V_SAD_U32 instruction pattern.

Differential Revision: http://reviews.llvm.org/D23069

llvm-svn: 279629

8 years ago[llvm-cov] Add the project summary to each source file coverage report.
Ying Yi [Wed, 24 Aug 2016 14:27:23 +0000 (14:27 +0000)]
[llvm-cov] Add the project summary to each source file coverage report.

This patch includes the following changes:
- Included header "Code coverage report" and include the date that the report was created.
- Included title (as specified in a command line option, (i.e llvm-cov  -project-title="Simple Test")
- In the summary, list the elf files that the source code file has contributed to.
- Used column heading for "Line No.", "Count No.", Source".

Differential Revision: https://reviews.llvm.org/D23345

llvm-svn: 279628

8 years agoFix mismatched new/free in Process:RunPrivateStateThread
Pavel Labath [Wed, 24 Aug 2016 14:03:10 +0000 (14:03 +0000)]
Fix mismatched new/free in Process:RunPrivateStateThread

NFC

llvm-svn: 279627

8 years ago[InstCombine] add assert and explanatory comment for fold removed in r279568; NFC
Sanjay Patel [Wed, 24 Aug 2016 13:55:55 +0000 (13:55 +0000)]
[InstCombine] add assert and explanatory comment for fold removed in r279568; NFC

I deleted a fold from InstCombine at:
https://reviews.llvm.org/rL279568

because it (like any InstCombine to a constant?) should always happen in InstSimplify,
however, it's not obvious what the assumptions are in the remaining code.

Add a comment and assert to make it clearer.

Differential Revision: https://reviews.llvm.org/D23819

llvm-svn: 279626

8 years agoCreate subranges for new intervals resulting from live interval splitting
Krzysztof Parzyszek [Wed, 24 Aug 2016 13:37:55 +0000 (13:37 +0000)]
Create subranges for new intervals resulting from live interval splitting

The register allocator can split a live interval of a register into a set
of smaller intervals. After the allocation of registers is complete, the
rewriter will modify the IR to replace virtual registers with the corres-
ponding physical registers. At this stage, if a register corresponding
to a subregister of a virtual register is used, the rewriter will check
if that subregister is undefined, and if so, it will add the <undef> flag
to the machine operand. The function verifying liveness of the subregis-
ter would assume that it is undefined, unless any of the subranges of the
live interval proves otherwise.
The problem is that the live intervals created during splitting do not
have any subranges, even if the original parent interval did. This could
result in the <undef> flag placed on a register that is actually defined.

Differential Revision: http://reviews.llvm.org/D21189

llvm-svn: 279625

8 years ago[mips] Preparatory work for a generic scheduler
Simon Dardis [Wed, 24 Aug 2016 13:00:47 +0000 (13:00 +0000)]
[mips] Preparatory work for a generic scheduler

Extend instruction definitions from nearly all ISAs to include
appropriate instruction itineraries. Change MIPS16s gp prologue
generation to use real instructions instead of using a pseudo
instruction.

Reviewers: dsanders, vkalintiris

Differential Review: https://reviews.llvm.org/D23548

llvm-svn: 279623

8 years ago[X86][AVX2] Ensure on 32-bit targets that we broadcast f64 types not i64 (PR29101)
Simon Pilgrim [Wed, 24 Aug 2016 12:42:31 +0000 (12:42 +0000)]
[X86][AVX2] Ensure on 32-bit targets that we broadcast f64 types not i64 (PR29101)

llvm-svn: 279622

8 years ago[X86][F16C] Regenerated f16c tests
Simon Pilgrim [Wed, 24 Aug 2016 11:56:15 +0000 (11:56 +0000)]
[X86][F16C] Regenerated f16c tests

llvm-svn: 279621

8 years ago[Loop Vectorizer] Support predication of div/rem
Gil Rapaport [Wed, 24 Aug 2016 11:37:57 +0000 (11:37 +0000)]
[Loop Vectorizer] Support predication of div/rem

div/rem instructions in basic blocks that require predication currently prevent
vectorization. This patch extends the existing mechanism for predicating stores
to handle other instructions and leverages it to predicate divs and rems.

Differential Revision: https://reviews.llvm.org/D22918

llvm-svn: 279620

8 years ago[X86][SSE] Add support for 32-bit element vectors to X86ISD::VZEXT_LOAD
Simon Pilgrim [Wed, 24 Aug 2016 10:46:40 +0000 (10:46 +0000)]
[X86][SSE] Add support for 32-bit element vectors to X86ISD::VZEXT_LOAD

Consecutive load matching (EltsFromConsecutiveLoads) currently uses VZEXT_LOAD (load scalar into lowest element and zero uppers) for vXi64 / vXf64 vectors only.

For vXi32 / vXf32 vectors it instead creates a scalar load, SCALAR_TO_VECTOR and finally VZEXT_MOVL (zero upper vector elements), relying on tablegen patterns to match this into an equivalent of VZEXT_LOAD.

This patch adds the VZEXT_LOAD patterns for vXi32 / vXf32 vectors directly and updates EltsFromConsecutiveLoads to use this.

This has proven necessary to allow us to easily make VZEXT_MOVL a full member of the target shuffle set - without this change the call to combineShuffle (which is the main caller of EltsFromConsecutiveLoads) tended to recursively recreate VZEXT_MOVL nodes......

Differential Revision: https://reviews.llvm.org/D23673

llvm-svn: 279619