oota-llvm.git
8 years ago[libFuzzer] experimental flag -drill (another search heuristic; Mike Aizatsky's idea)
Kostya Serebryany [Thu, 12 Nov 2015 01:02:01 +0000 (01:02 +0000)]
[libFuzzer] experimental flag -drill (another search heuristic; Mike Aizatsky's idea)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252838 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[TLS on Darwin] use a different mask for tls calls on x86-64.
Manman Ren [Thu, 12 Nov 2015 00:54:04 +0000 (00:54 +0000)]
[TLS on Darwin] use a different mask for tls calls on x86-64.

Calls involved in thread-local variable lookup save more registers
than normal calls.

rdar://problem/23073171

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252837 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodwarfdump: Use the right enum for the tu_index section
David Blaikie [Thu, 12 Nov 2015 00:44:35 +0000 (00:44 +0000)]
dwarfdump: Use the right enum for the tu_index section

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252835 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodwarfdump: add command line support for dumping the tu_index section
David Blaikie [Thu, 12 Nov 2015 00:42:49 +0000 (00:42 +0000)]
dwarfdump: add command line support for dumping the tu_index section

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252830 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoFix problems in coding style
Xinliang David Li [Thu, 12 Nov 2015 00:32:17 +0000 (00:32 +0000)]
Fix problems in coding style

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252829 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ARM] Enable shrink-wrapping by default.
Quentin Colombet [Wed, 11 Nov 2015 23:31:46 +0000 (23:31 +0000)]
[ARM] Enable shrink-wrapping by default.

Differential Revision: http://reviews.llvm.org/D14357

rdar://problem/21942589

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252825 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WinEH] Don't forward branches across empty EH pad BBs
Reid Kleckner [Wed, 11 Nov 2015 23:09:31 +0000 (23:09 +0000)]
[WinEH] Don't forward branches across empty EH pad BBs

For really simple SEH catchpads, we tried to forward the invoke unwind
edge across the empty block.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252822 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[LIR] General refactor to improve compile-time and simplify code.
Chad Rosier [Wed, 11 Nov 2015 23:00:59 +0000 (23:00 +0000)]
[LIR] General refactor to improve compile-time and simplify code.

First create a list of candidates, then transform.  This simplifies the code in
that you have don't have to worry that you may be using an invalidated
iterator.

Previously, each time we created a memset/memcpy we would reevaluate the entire
loop potentially resulting in lots of redundant work for large basic blocks.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252817 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoFix docs build break for revision r252798
Hemant Kulkarni [Wed, 11 Nov 2015 22:14:58 +0000 (22:14 +0000)]
Fix docs build break for revision r252798

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252812 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[IR] Add support for empty tokens
David Majnemer [Wed, 11 Nov 2015 21:57:16 +0000 (21:57 +0000)]
[IR] Add support for empty tokens

When working with tokens, it is often the case that one has instructions
which consume a token and produce a new token.  Currently, we have no
mechanism to represent an initial token state.

Instead, we can create a notional "empty token" by inventing a new
constant which captures the semantics we would like.  This new constant
is called ConstantTokenNone and is written textually as "token none".

Differential Revision: http://reviews.llvm.org/D14581

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252811 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[CMake] Fixing passthrough for variables starting with the sub-project name
Chris Bieneman [Wed, 11 Nov 2015 21:54:31 +0000 (21:54 +0000)]
[CMake] Fixing passthrough for variables starting with the sub-project name

This allows TEST_SUITE variables to be passed from the top-level CMake into the external project.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252810 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoIntroduce deoptimization operand bundles
Sanjoy Das [Wed, 11 Nov 2015 21:38:02 +0000 (21:38 +0000)]
Introduce deoptimization operand bundles

Summary:
This change introduces the notion of "deoptimization" operand bundles.
LLVM can recognize and optimize these in more precise ways than it can a
generic "unknown" operand bundles.

The current form of this special recognition / optimization is an enum
entry in LLVMContext, a LangRef blurb and a verifier rule.  Over time we
will teach LLVM to do more aggressive optimization around deoptimization
operand bundles, exploiting known facts about kinds of state
deoptimization operand bundles are allowed to track.

Reviewers: reames, majnemer, chandlerc, dexonsmith

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14551

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252806 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAdding qualifier for make_unique.
Mike Aizatsky [Wed, 11 Nov 2015 21:28:40 +0000 (21:28 +0000)]
Adding qualifier for make_unique.

Summary: MSVC gets confused.

Differential Revision: http://reviews.llvm.org/D14585

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252804 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agosancov tool in c++
Mike Aizatsky [Wed, 11 Nov 2015 20:58:20 +0000 (20:58 +0000)]
sancov tool in c++

First batch of sancov.py rewrite in c++.
Supports "-print" and "-coveredfns" commands.

Followup to http://reviews.llvm.org/D14356 in a better location.

Differential Revision: http://reviews.llvm.org/D14579

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252801 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoReport Windows error code in a fatal error after a system call.
Paul Robinson [Wed, 11 Nov 2015 20:49:32 +0000 (20:49 +0000)]
Report Windows error code in a fatal error after a system call.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252800 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agounique_ptrify the AllocValueProfData helper function introduced in r252783
David Blaikie [Wed, 11 Nov 2015 20:44:52 +0000 (20:44 +0000)]
unique_ptrify the AllocValueProfData helper function introduced in r252783

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252799 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[Symbolizer]: Add -pretty-print option
Hemant Kulkarni [Wed, 11 Nov 2015 20:41:43 +0000 (20:41 +0000)]
[Symbolizer]: Add -pretty-print option

Differential Revision: http://reviews.llvm.org/D13671

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252798 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoMove the enum attributes defined in Attributes.h to a table-gen file.
Akira Hatanaka [Wed, 11 Nov 2015 20:35:42 +0000 (20:35 +0000)]
Move the enum attributes defined in Attributes.h to a table-gen file.

This is a step towards consolidating some of the information regarding
attributes in a single place.

This patch moves the enum attributes in Attributes.h to the table-gen
file. Additionally, it adds definitions of target independent string
attributes that will be used in follow-up commits by the inliner to
check attribute compatibility.

rdar://problem/19836465

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252796 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAdd a libLTO diagnostic handler that supports lto_get_error_message API
Yunzhong Gao [Wed, 11 Nov 2015 19:59:08 +0000 (19:59 +0000)]
Add a libLTO diagnostic handler that supports lto_get_error_message API

This is a follow-up from the previous discussion on the thread:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151019/307763.html

The LibLTO lto_get_error_message() API reads error messages from a std::string
sLastErrorString. Instead of passing this string around as an argument, this
patch creates a diagnostic handler and then sends this handler to the
constructor of LTOCodeGenerator.

Differential Revision: http://reviews.llvm.org/D14313

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252791 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[DAGCombiner] Improve zextload optimization.
Geoff Berry [Wed, 11 Nov 2015 19:42:52 +0000 (19:42 +0000)]
[DAGCombiner] Improve zextload optimization.

Summary:
Don't fold
  (zext (and (load x), cst)) -> (and (zextload x), (zext cst))
if
  (and (load x) cst)
will match as a zextload already and has additional users.

For example, the following IR:

  %load = load i32, i32* %ptr, align 8
  %load16 = and i32 %load, 65535
  %load64 = zext i32 %load16 to i64
  store i32 %load16, i32* %dst1, align 4
  store i64 %load64, i64* %dst2, align 8

used to produce the following aarch64 code:

ldr w8, [x0]
and w9, w8, #0xffff
and x8, x8, #0xffff
str w9, [x1]
str x8, [x2]

but with this change produces the following aarch64 code:

ldrh w8, [x0]
str w8, [x1]
str x8, [x2]

Reviewers: resistor, mcrosier

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14340

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252789 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodwarfdump: DWP type unit index dumping skeleton
David Blaikie [Wed, 11 Nov 2015 19:40:49 +0000 (19:40 +0000)]
dwarfdump: DWP type unit index dumping skeleton

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252786 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoRefactoring and fix another instance of asan error
Xinliang David Li [Wed, 11 Nov 2015 19:31:53 +0000 (19:31 +0000)]
Refactoring and fix another instance of asan error

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252783 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoFormat my previous commit
David Blaikie [Wed, 11 Nov 2015 19:30:47 +0000 (19:30 +0000)]
Format my previous commit

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252782 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodwarfdump: First piece of support for DWP dumping
David Blaikie [Wed, 11 Nov 2015 19:28:21 +0000 (19:28 +0000)]
dwarfdump: First piece of support for DWP dumping

Just a tiny piece of index dumping - the header in this instance.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252781 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoTest commit after password reset
Hemant Kulkarni [Wed, 11 Nov 2015 19:24:08 +0000 (19:24 +0000)]
Test commit after password reset

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252779 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WinEH] Only generate UnwindHelp slot for MSVCXX
Joseph Tremoulet [Wed, 11 Nov 2015 19:21:09 +0000 (19:21 +0000)]
[WinEH] Only generate UnwindHelp slot for MSVCXX

Summary: Other personalities don't use this special frame slot.

Reviewers: majnemer, andrew.w.kaylor, rnk

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14580

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252778 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoSupport: Recognize Borland DWARF extensions.
Dawn Perchik [Wed, 11 Nov 2015 18:47:36 +0000 (18:47 +0000)]
Support: Recognize Borland DWARF extensions.

This patch adds DWARF values for the Delphi language and Borland C++
language extensions.

Reviewed by: dblaikie
Subscribers: llvm-commits, majnemer
Differential Revision: http://reviews.llvm.org/D14522

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252776 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAdd target preference for GatherAllAliases max depth
Matt Arsenault [Wed, 11 Nov 2015 18:44:33 +0000 (18:44 +0000)]
Add target preference for GatherAllAliases max depth

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252775 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoReverting r252760
Colin LeMahieu [Wed, 11 Nov 2015 18:11:06 +0000 (18:11 +0000)]
Reverting r252760

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252770 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoclang-format lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp
Dehao Chen [Wed, 11 Nov 2015 18:09:47 +0000 (18:09 +0000)]
clang-format lib/CodeGen/AsmPrinter/DwarfCompileUnit.cpp

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252769 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoEmit discriminator for inlined callsites.
Dehao Chen [Wed, 11 Nov 2015 18:08:18 +0000 (18:08 +0000)]
Emit discriminator for inlined callsites.

Summary: Inlined callsites need to be emitted in debug info so that sample profile can be annotated to the correct inlined instance.

Reviewers: dnovillo, dblaikie

Subscribers: dblaikie, llvm-commits

Differential Revision: http://reviews.llvm.org/D14511

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252768 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoSamplePGO - Fix PR 25482 - Do not rely on llvm.dbg.cu for discriminators
Diego Novillo [Wed, 11 Nov 2015 17:54:37 +0000 (17:54 +0000)]
SamplePGO - Fix PR 25482 - Do not rely on llvm.dbg.cu for discriminators

The discriminators pass relied on the presence of llvm.dbg.cu to decide
whether to add discriminators, but this fails in the case where debug
info is only enabled partially when -fprofile-sample-use is active.

The reason llvm.dbg.cu is not present in these cases is to prevent
codegen from emitting debug info (as it is only used for the sample
profile pass).

This changes the discriminators pass to also emit discriminators even
when debug info is not being emitted.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252763 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agollvm-config.cpp: Prune an obsolete \param. [-Wdocumentation]
NAKAMURA Takumi [Wed, 11 Nov 2015 17:50:20 +0000 (17:50 +0000)]
llvm-config.cpp: Prune an obsolete \param. [-Wdocumentation]

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252762 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[Symbolizer]: Add -pretty-print option
Hemant Kulkarni [Wed, 11 Nov 2015 17:47:54 +0000 (17:47 +0000)]
[Symbolizer]: Add -pretty-print option

 Differential Revision: http://reviews.llvm.org/D13671

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252760 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[MIPS] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()
Sanjay Patel [Wed, 11 Nov 2015 17:24:56 +0000 (17:24 +0000)]
[MIPS] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()

MIPS32 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any MIPS32
implementation.

The net result of allowing this speculation for the regression tests in this patch is
that we get this code:

ctlz:
  jr  $ra
  clz  $2, $4

cttz:
  addiu  $1, $4, -1
  not  $2, $4
  and  $1, $2, $1
  clz  $1, $1
  addiu  $2, $zero, 32
  jr  $ra
  subu  $2, $2, $1

Instead of:

ctlz:
  beqz  $4, $BB0_2
  addiu  $2, $zero, 32
  clz  $2, $4
$BB0_2:
  jr  $ra
  nop

cttz:
  beqz  $4, $BB1_2
  addiu  $2, $zero, 32
  addiu  $1, $4, -1
  not  $2, $4
  and  $1, $2, $1
  clz  $1, $1
  addiu  $2, $zero, 32
  subu  $2, $2, $1
$BB1_2:
  jr  $ra
  nop

See D14469 for the larger motivation.

Differential Revision: http://reviews.llvm.org/D14500

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252755 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agotest/DebugInfo/ARM/prologue_end.ll references thumbv1, which is invalid.
Artyom Skrobov [Wed, 11 Nov 2015 17:22:18 +0000 (17:22 +0000)]
test/DebugInfo/ARM/prologue_end.ll references thumbv1, which is invalid.

The committer didn't respond at http://reviews.llvm.org/D14338, so we've got to fix this for them.

This test doesn't pass with thumbv6, so I suppose what they meant is thumbv7.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252754 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[mips] Move MC tests for the DSP ASE into the standard format.
Daniel Sanders [Wed, 11 Nov 2015 16:50:13 +0000 (16:50 +0000)]
[mips] Move MC tests for the DSP ASE into the standard format.

Summary:
Only DSPr2 is present because it appears we've never added DSPr1 tests.
We'll have to correct that in a later patch.

Reviewers: vkalintiris

Subscribers: llvm-commits, dsanders

Differential Revision: http://reviews.llvm.org/D14448

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252752 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoProperly fix unused variable in disable-assert builds.
Diego Novillo [Wed, 11 Nov 2015 16:39:22 +0000 (16:39 +0000)]
Properly fix unused variable in disable-assert builds.

I missed the side-effects of ParseBFI in my previous attempt (r252748).
Thanks dblaikie for the suggestion of adding a void use of the unused
variable instead.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252751 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[CMake] Make llvm_install_library_symlink respect LLVM_LIBDIR_SUFFIX.
Chris Bieneman [Wed, 11 Nov 2015 16:19:39 +0000 (16:19 +0000)]
[CMake] Make llvm_install_library_symlink respect LLVM_LIBDIR_SUFFIX.

This is required to support multilib install targets, and addresses a regression introduced in r252093.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252749 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoRemove unused variable in disable-assert builds. NFC.
Diego Novillo [Wed, 11 Nov 2015 16:14:52 +0000 (16:14 +0000)]
Remove unused variable in disable-assert builds.  NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252748 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[CMake] Add support for building the llvm test-suite as part of an LLVM build using...
Chris Bieneman [Wed, 11 Nov 2015 16:14:03 +0000 (16:14 +0000)]
[CMake] Add support for building the llvm test-suite as part of an LLVM build using clang and lld

Summary:
This patch adds a new CMake module for working with ExternalProjects. This wrapper for ExternalProject supports using just-built tools and can hook up dependencies properly so that projects get cleared out.

The example usage here is for the llvm test-suite. In this example, the test-suite is setup as dependent on clang and lld if they are in-tree. If the clang or lld binaries change the test-suite is re-configured, cleaned, and rebuilt.

This cleanup and abstraction wrapping ExternalProject can be extended and applied to other runtime libraries like compiler-rt and libcxx.

Reviewers: samsonov, jroelofs, rengolin, jmolloy

Subscribers: jmolloy, llvm-commits

Differential Revision: http://reviews.llvm.org/D14513

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252747 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoVisibly fail if attempting to encode register AH,BH,CH,DH in a REX-prefixed instruction.
Douglas Katzman [Wed, 11 Nov 2015 15:51:16 +0000 (15:51 +0000)]
Visibly fail if attempting to encode register AH,BH,CH,DH in a REX-prefixed instruction.

Differential Revision: http://reviews.llvm.org/D13316
Fixes PR25003

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252743 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ARM] Combine BFIs together
James Molloy [Wed, 11 Nov 2015 15:40:40 +0000 (15:40 +0000)]
[ARM] Combine BFIs together

If we have a chain of BFIs, we may be able to combine several together into one merged BFI. We can do this if the "from" bits from one BFI OR'd with the "from" bits from the other BFI form a contiguous range, and the same with the "to" bits.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252740 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[SLP] Enable -slp-vectorize-hor by default.
Charlie Turner [Wed, 11 Nov 2015 15:03:46 +0000 (15:03 +0000)]
[SLP] Enable -slp-vectorize-hor by default.

Measurements primarily on AArch64 have shown this feature does not
significantly effect compile-time. The are no significant perf changes in LNT,
but for AArch64 at least, there are wins in third party benchmarks.

As discussed on llvm-dev, we're going to try turning this on by default and see
how other targets react to the change.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252733 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoSilencing a signed vs unsigned type mismatch warning.
Aaron Ballman [Wed, 11 Nov 2015 14:57:28 +0000 (14:57 +0000)]
Silencing a signed vs unsigned type mismatch warning.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252732 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoSilencing nine warnings for "enumeral and non-enumeral type in conditional expression...
Aaron Ballman [Wed, 11 Nov 2015 13:44:06 +0000 (13:44 +0000)]
Silencing nine warnings for "enumeral and non-enumeral type in conditional expression"; NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252728 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[X86] Replace LEAs with INC/DEC when profitable
Michael Kuperstein [Wed, 11 Nov 2015 11:44:31 +0000 (11:44 +0000)]
[X86] Replace LEAs with INC/DEC when profitable

If possible and profitable, replace lea %reg, 1(%reg) and lea %reg, -1(%reg) with inc %reg and dec %reg respectively.

Patch by: anton.nadolsky@intel.com
Differential Revision: http://reviews.llvm.org/D14059

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252722 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ASan] Enable optional ASan recovery.
Yury Gribov [Wed, 11 Nov 2015 10:36:49 +0000 (10:36 +0000)]
[ASan] Enable optional ASan recovery.

Differential Revision: http://reviews.llvm.org/D14242

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252719 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[X86] Fix feature flags on some MMX register instructions that really were introduced...
Craig Topper [Wed, 11 Nov 2015 07:29:25 +0000 (07:29 +0000)]
[X86] Fix feature flags on some MMX register instructions that really were introduced with SSE or SSE2.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252709 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[X86] Remove redundant MMX isel patterns.
Craig Topper [Wed, 11 Nov 2015 07:29:22 +0000 (07:29 +0000)]
[X86] Remove redundant MMX isel patterns.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252708 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[libFuzzer] better links
Kostya Serebryany [Wed, 11 Nov 2015 05:25:24 +0000 (05:25 +0000)]
[libFuzzer] better links

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252705 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[libFuzzer] more trophies
Kostya Serebryany [Wed, 11 Nov 2015 05:20:55 +0000 (05:20 +0000)]
[libFuzzer] more trophies

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252704 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agolit: Show all output with --show-all, even in combination with --succinct
Matthias Braun [Wed, 11 Nov 2015 03:03:54 +0000 (03:03 +0000)]
lit: Show all output with --show-all, even in combination with --succinct

I missed an earlier exit for the --succinct case when I introduced the
-a option.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252698 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoADT: Avoid relying on UB in ilist_node::getNextNode()
Duncan P. N. Exon Smith [Wed, 11 Nov 2015 02:26:42 +0000 (02:26 +0000)]
ADT: Avoid relying on UB in ilist_node::getNextNode()

Re-implement `ilist_node::getNextNode()` and `getPrevNode()` without
relying on the sentinel having a "next" pointer.  Instead, get access to
the owning list and compare against the `begin()` and `end()` iterators.

This only works when the node *can* get access to the owning list.  The
new support is in `ilist_node_with_parent<>`, and any class `Ty`
inheriting from `ilist_node<NodeTy>` that wants `getNextNode()` and/or
`getPrevNode()` should inherit from
`ilist_node_with_parent<NodeTy, ParentTy>` instead.  The requirements:

  - `NodeTy` must have a `getParent()` function that returns the parent.
  - `ParentTy` must have a `getSublistAccess()` static that, given a(n
    ignored) `NodeTy*` (to determine which list), returns a member field
    pointer to the appropriate `ilist<>`.

This isn't the cleanest way to get access to the owning list, but it
leverages the API already used in the IR hierarchy (see, e.g.,
`Instruction::getSublistAccess()`).

If anyone feels like ripping out the calls to `getNextNode()` and
`getPrevNode()` and replacing with direct iterator logic, they can also
remove the access function, etc., but as an incremental step, I'm
maintaining the API where it's currently used in tree.

If these requirements are *not* met, call sites with access to the ilist
can call `iplist<NodeTy>::getNextNode(NodeTy*)` directly, as in
ilistTest.cpp.

Why rewrite this?

The old code was broken, calling `getNext()` on a sentinel that possibly
didn't have a "next" pointer at all!  The new code avoids that
particular flavour of UB (see the commit message for r252538 for more
details about the "lucky" memory layout that made this function so
interesting).

There's still some UB here: the end iterator gets downcast to `NodeTy*`,
even when it's a sentinel (which is typically
`ilist_half_node<NodeTy*>`).  I'll tackle that in follow-up commits.
See this llvm-dev thread for more details:
http://lists.llvm.org/pipermail/llvm-dev/2015-October/091115.html

What's the danger?

There might be some code that relies on `getNextNode()` or
`getPrevNode()` *never* returning `nullptr` -- i.e., that relies on them
being broken when the sentinel is an `ilist_half_node<NodeTy>`.  I tried
to root out those cases with the audits I did leading up to r252380, but
it's possible I missed one or two.  I hope not.

(If (1) you have out-of-tree code, (2) you've reverted r252380
temporarily, and (3) you get some weird crashes with this commit, then I
recommend un-reverting r252380 and auditing the compile errors looking
for "strange" implicit conversions.)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252694 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoSort the enums in Attributes.h in case insensitive alphabetical order.
Akira Hatanaka [Wed, 11 Nov 2015 02:11:46 +0000 (02:11 +0000)]
Sort the enums in Attributes.h in case insensitive alphabetical order.

Sort the enums in preparation for moving the attributes to a table-gen
file.

rdar://problem/19836465

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252692 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WebAssembly] Support non-legal argument and return types.
Dan Gohman [Wed, 11 Nov 2015 01:33:02 +0000 (01:33 +0000)]
[WebAssembly] Support non-legal argument and return types.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252687 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[MC] Use LShr for constant evaluation of ">>" on non-arm64 darwin.
Ahmed Bougacha [Wed, 11 Nov 2015 00:51:36 +0000 (00:51 +0000)]
[MC] Use LShr for constant evaluation of ">>" on non-arm64 darwin.

Follow-up to r235963: this matches other assemblers and is less
unexpected (e.g. PR23227).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252681 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoMachineInstr: addRegisterDefReadUndef() => setRegisterDefReadUndef()
Matthias Braun [Wed, 11 Nov 2015 00:41:58 +0000 (00:41 +0000)]
MachineInstr: addRegisterDefReadUndef() => setRegisterDefReadUndef()

This way we can not only add but also remove read undef flags.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252678 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAMDGPU: Print more fields in comments
Matt Arsenault [Wed, 11 Nov 2015 00:27:46 +0000 (00:27 +0000)]
AMDGPU: Print more fields in comments

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252677 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ValueTracking] Remove untested / unreachable code, NFC
Sanjoy Das [Wed, 11 Nov 2015 00:16:41 +0000 (00:16 +0000)]
[ValueTracking] Remove untested / unreachable code, NFC

Right now isTruePredicate is only ever called with Pred == ICMP_SLE or
ICMP_ULE, and the ICMP_SLT and ICMP_ULT cases are dead.  This change
removes the untested dead code so that the function is not misleading.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252676 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAMDGPU: Remove dead code
Matt Arsenault [Wed, 11 Nov 2015 00:01:36 +0000 (00:01 +0000)]
AMDGPU: Remove dead code

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252675 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAMDGPU: Set isAllocatable = 0 on VS_32/VS_64
Matt Arsenault [Wed, 11 Nov 2015 00:01:32 +0000 (00:01 +0000)]
AMDGPU: Set isAllocatable = 0 on VS_32/VS_64

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252674 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ValueTracking] Teach isImpliedCondition a new bitwise trick
Sanjoy Das [Tue, 10 Nov 2015 23:56:20 +0000 (23:56 +0000)]
[ValueTracking] Teach isImpliedCondition a new bitwise trick

Summary:
This change teaches isImpliedCondition to prove things like

  (A | 15) < L  ==>  (A | 14) < L

if the low 4 bits of A are known to be zero.

Depends on D14391

Reviewers: majnemer, reames, hfinkel

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14392

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252673 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ValueTracking] Use m_APInt instead of m_ConstantInt, NFC
Sanjoy Das [Tue, 10 Nov 2015 23:56:15 +0000 (23:56 +0000)]
[ValueTracking] Use m_APInt instead of m_ConstantInt, NFC

This change would add functionality if isImpliedCondition worked on
vector types; but since it bail out on vector predicates this change is
an NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252672 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoTableGen: Emit LaneMask for register classes without subregisters as ~0u
Matthias Braun [Tue, 10 Nov 2015 23:23:05 +0000 (23:23 +0000)]
TableGen: Emit LaneMask for register classes without subregisters as ~0u

This makes it slightly easier to handle classes with and without
subregister uniformly.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252671 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WinEH] Insert the MBB for EH_RESTORE after the catchret
Reid Kleckner [Tue, 10 Nov 2015 23:22:20 +0000 (23:22 +0000)]
[WinEH] Insert the MBB for EH_RESTORE after the catchret

Inserting it before the target block could be bad, we might already have
a fallthrough edge to it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252670 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[cmake] move SONAME handling to llvm_add_library
Andrew Wilkins [Tue, 10 Nov 2015 23:19:21 +0000 (23:19 +0000)]
[cmake] move SONAME handling to llvm_add_library

Summary:
Move handling of the SONAME option from add_llvm_library
to llvm_add_library, so that it can be used in sub-projects.
In particular, this makes it possible to have consistently
named shared libraries for LLVM, Clang and LLDB.

Also, base the SONAME and symlinks on the output name
by extracting the OUTPUT_NAME property, rather than assuming
it is the same as the target name.

Reviewers: beanz

Subscribers: llvm-commits

Differential Revision: http://reviews.llvm.org/D14539

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252669 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodon't repeat function/class/variable names in comments; NFC
Sanjay Patel [Tue, 10 Nov 2015 22:59:11 +0000 (22:59 +0000)]
don't repeat function/class/variable names in comments; NFC

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252666 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[doc] Compile CUDA with LLVM
Jingyue Wu [Tue, 10 Nov 2015 22:35:47 +0000 (22:35 +0000)]
[doc] Compile CUDA with LLVM

Summary:
This patch adds documentation on compiling CUDA with LLVM as requested by many
engineers and researchers. It includes not only user guides but also some
internals (mostly optimizations) so that early adopters can start hacking and
contributing.

Quite a few researchers who contacted us haven't used LLVM before, which is
unsurprising as it hasn't been long since LLVM picked up CUDA. So I added a
short summary to help these folks get started with LLVM.

I expect this document to evolve substantially down the road. The user guides
will be much simplified after the Clang integration is done. However, the
internals should continue growing to include for example performance debugging
and key areas to improve.

Reviewers: chandlerc, meheff, broune, tra

Subscribers: silvas, jingyue, llvm-commits, eliben

Differential Revision: http://reviews.llvm.org/D14370

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252660 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[libFuzzer] add UninstrumentedTest.cpp (missing from a previous commit)
Kostya Serebryany [Tue, 10 Nov 2015 22:02:56 +0000 (22:02 +0000)]
[libFuzzer] add UninstrumentedTest.cpp (missing from a previous commit)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252658 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WebAssembly] Remove special cases for things that are no longer special. NFC.
Dan Gohman [Tue, 10 Nov 2015 21:48:21 +0000 (21:48 +0000)]
[WebAssembly] Remove special cases for things that are no longer special. NFC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252656 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAdd PPCMIPeephole.cpp to CMakeLists.txt
Bill Schmidt [Tue, 10 Nov 2015 21:43:45 +0000 (21:43 +0000)]
Add PPCMIPeephole.cpp to CMakeLists.txt

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252654 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[WebAssembly] Support for floating point min and max.
Dan Gohman [Tue, 10 Nov 2015 21:40:21 +0000 (21:40 +0000)]
[WebAssembly] Support for floating point min and max.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252653 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[CMake] Autoconf builds libLTO with -fPIC, CMake should be able to as well.
Chris Bieneman [Tue, 10 Nov 2015 21:38:58 +0000 (21:38 +0000)]
[CMake] Autoconf builds libLTO with -fPIC, CMake should be able to as well.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252652 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[PowerPC] Add an MI SSA peephole pass.
Bill Schmidt [Tue, 10 Nov 2015 21:38:26 +0000 (21:38 +0000)]
[PowerPC] Add an MI SSA peephole pass.

This patch adds a pass for doing PowerPC peephole optimizations at the
MI level while the code is still in SSA form.  This allows for easy
modifications to the instructions while depending on a subsequent pass
of DCE.  Both passes are very fast due to the characteristics of SSA.

At this time, the only peepholes added are for cleaning up various
redundancies involving the XXPERMDI instruction.  However, I would
expect this will be a useful place to add more peepholes for
inefficiencies generated during instruction selection.  The pass is
placed after VSX swap optimization, as it is best to let that pass
remove unnecessary swaps before performing any remaining clean-ups.

The utility of these clean-ups are demonstrated by changes to four
existing test cases, all of which now have tighter expected code
generation.  I've also added Eric Schweiz's bugpoint-reduced test from
PR25157, for which we now generate tight code.  One other test started
failing for me, and I've fixed it
(test/Transforms/PlaceSafepoints/finite-loops.ll) as well; this is not
related to my changes, and I'm not sure why it works before and not
after.  The problem is that the CHECK-NOT: of "statepoint" from test1
fails because of the "statepoint" in test2, and so forth.  Adding a
CHECK-LABEL in between keeps the different occurrences of that string
properly scoped.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252651 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agodsymutil: Prune module forward decl DIEs if a uniquable definition was
Adrian Prantl [Tue, 10 Nov 2015 21:31:05 +0000 (21:31 +0000)]
dsymutil: Prune module forward decl DIEs if a uniquable definition was
already emitted and fix a latent bug in DIECloner where the DW_CHILDREN_yes
flag is set based on the number of children in the input DIE rather than
the number of children that are actually being cloned.

rdar://problem/23439845

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252649 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoEnsure ModuleLinker materializes complete comdat groups
Teresa Johnson [Tue, 10 Nov 2015 21:09:06 +0000 (21:09 +0000)]
Ensure ModuleLinker materializes complete comdat groups

Summary:
The module linker lazy links some "discardable if unused" global
values (e.g. linkonce), materializing and linking them only
if they are referenced in the module. If a comdat group contains a
linkonce member that is not referenced, however, it would not be
materialized and linked, leading to an incomplete comdat group.

If there are other object files not part of the same LTO link that also
define and use that comdat group, the linker may select the incomplete
group leading to link time unsats.

To solve this, whenever a global value body is linked, make sure we
materialize any other members of the same comdat group that are not yet
materialized. This ensures they are in the lazy link list and get linked
as well.

Added new test and adjusted old test to remove parts that didn't
make sense with fix.

Reviewers: rafael

Subscribers: dexonsmith, davidxl, llvm-commits

Differential Revision: http://reviews.llvm.org/D14516

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252647 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[IR] Make {Call,Invoke}::cloneImpl aware of operand bundles
Sanjoy Das [Tue, 10 Nov 2015 20:13:21 +0000 (20:13 +0000)]
[IR] Make {Call,Invoke}::cloneImpl aware of operand bundles

This was an omission in the patch that landed initial support for
operand bundles.  So far we haven't hit this, but we will once the
inliner is able to inline calls to functions that contain calls with
operand bundles.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252645 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[OperandBundles] Identify operand bundles with both their names and IDs
Sanjoy Das [Tue, 10 Nov 2015 20:13:15 +0000 (20:13 +0000)]
[OperandBundles] Identify operand bundles with both their names and IDs

No code uses this functionality yet.  This change just exposes
information / structure that was already present.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252644 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoless indent; NFCI
Sanjay Patel [Tue, 10 Nov 2015 20:09:02 +0000 (20:09 +0000)]
less indent; NFCI

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252643 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ARM] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()
Sanjay Patel [Tue, 10 Nov 2015 19:24:31 +0000 (19:24 +0000)]
[ARM] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()

ARM V6T2 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any ARM V6T2
implementation.

The net result of allowing this speculation for the regression tests in this patch is
that we get this code:

ctlz:
  clz  r0, r0
  bx  lr
cttz:
  rbit  r0, r0
  clz  r0, r0
  bx  lr

Instead of:

ctlz:
  cmp  r0, #0
  moveq  r0, #32
  clzne  r0, r0
  bx  lr
cttz:
  cmp   r0, #0
  moveq  r0, #32
  rbitne  r0, r0
  clzne  r0, r0
  bx  lr

This will help solve a general speculation/despeculation problem noted in PR24818:
https://llvm.org/bugs/show_bug.cgi?id=24818

Differential Revision: http://reviews.llvm.org/D14469

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252639 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agollvm-lto: trivial spelling changes to distinguish custom diagnostic handler and
Yunzhong Gao [Tue, 10 Nov 2015 18:52:48 +0000 (18:52 +0000)]
llvm-lto: trivial spelling changes to distinguish custom diagnostic handler and
default diagnostic handler.

Differential Revision: http://reviews.llvm.org/D14520

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252633 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoLegalizeDAG: Implement promote for scalar_to_vector
Matt Arsenault [Tue, 10 Nov 2015 18:48:11 +0000 (18:48 +0000)]
LegalizeDAG: Implement promote for scalar_to_vector

This allows avoiding the default Expand behavior which
introduces stack usage. Bitcast the scalar and replace
the missing elements with undef.

This is covered by existing tests and used by a future
commit which makes 64-bit vectors legal types on AMDGPU.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252632 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoLegalizeDAG: Implement promote for insert_vector_elt
Matt Arsenault [Tue, 10 Nov 2015 18:48:08 +0000 (18:48 +0000)]
LegalizeDAG: Implement promote for insert_vector_elt

This is covered by existing tests and used by a future
commit which makes 64-bit vectors legal types on AMDGPU.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252631 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoLegalizeDAG: Implement promote for extract_vector_elt
Matt Arsenault [Tue, 10 Nov 2015 18:48:04 +0000 (18:48 +0000)]
LegalizeDAG: Implement promote for extract_vector_elt

This is for AMDGPU to implement v2i64 extract as extract of
half of a v4i32.

This is covered by existing tests and used by a future
commit which makes 64-bit vectors legal types on AMDGPU.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252630 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ValueTracking] Recognize that and(x, add (x, -1)) clears the low bit
Philip Reames [Tue, 10 Nov 2015 18:46:14 +0000 (18:46 +0000)]
[ValueTracking] Recognize that and(x, add (x, -1)) clears the low bit

This is a cleaned up version of a patch by John Regehr with permission. Originally found via the souper tool.

If we add an odd number to x, then bitwise-and the result with x, we know that the low bit of the result must be zero. Either it was zero in x originally, or the add cleared it in the temporary value. As a result, one of the two values anded together must have the bit cleared.

Differential Revision: http://reviews.llvm.org/D14315

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252629 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoAddLLVM: squelch CMP0007 by not adding empty elements to list
Ramkumar Ramachandra [Tue, 10 Nov 2015 18:26:34 +0000 (18:26 +0000)]
AddLLVM: squelch CMP0007 by not adding empty elements to list

When configuring various llvm projects that use AddLLVM.cmake, this warning is
emitted many times, flooding the screen:

Policy CMP0007 is not set: list command no longer ignores empty elements.

The fix is removing an extra semicolon.

Differential Revision: http://reviews.llvm.org/D14339

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252628 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ThinLTO] Update comment per change in WeakAny handling (NFC)
Teresa Johnson [Tue, 10 Nov 2015 18:26:31 +0000 (18:26 +0000)]
[ThinLTO] Update comment per change in WeakAny handling (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252627 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[ThinLTO] WeakAny fixes/cleanup
Teresa Johnson [Tue, 10 Nov 2015 18:20:11 +0000 (18:20 +0000)]
[ThinLTO] WeakAny fixes/cleanup

Ensure WeakAny variables are imported as ExternalWeak declarations. To
handle WeakAny more consistently and fix this issue:

1) Update helper doImportAsDefinition to properly flag WeakAny variables
   and aliases as not importing defintions.

   Update callers of doImportAsDefinition to remove now redundant checks for
   WeakAny aliases, or ignore aliases, as appropriate.

2) Add any !doImportAsDefinition GVs to DoNotLinkFromSource set during
   linking of the GV prototype, where we usually add GVs to the
   DoNotLinkFromSource set for other reasons.

   Remove now unnecessary adding of WeakAny aliases to
   DoNotLinkFromSource set from copyGlobalAliasProto.

   Remove now unnecessary guard against linking non-imported function
   bodies from ModuleLinker::run.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252626 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[AArch64] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()
Sanjay Patel [Tue, 10 Nov 2015 18:11:37 +0000 (18:11 +0000)]
[AArch64] add overrides for isCheapToSpeculateCttz() and isCheapToSpeculateCtlz()

AArch64 has instructions for efficient count-leading/trailing-zeros, so this should be
considered a cheap operation (and therefore fair game for speculation) for any AArch64
implementation.

The net result of allowing this speculation for the regression tests in this
patch is that we get this code:

ctlz:
  clz  w0, w0
  ret

cttz:
  rbit  w8, w0
  clz  w0, w8
  ret

Instead of:

ctlz:
  cbz  w0, .LBB0_2
  clz  w0, w0
  ret
.LBB0_2:
  orr  w0, wzr, #0x20
  ret

cttz:
  cbz  w0, .LBB1_2
  rbit  w8, w0
  clz  w0, w8
  ret
.LBB1_2:
  orr  w0, wzr, #0x20
  ret

See D14469 for the larger motivation.

Differential Revision: http://reviews.llvm.org/D14505

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252625 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoRevert "Strip metadata when speculatively hoisting instructions"
Renato Golin [Tue, 10 Nov 2015 18:01:16 +0000 (18:01 +0000)]
Revert "Strip metadata when speculatively hoisting instructions"

This reverts commit r252604, as it broke all ARM and AArch64 buildbots, as
well as some x86, et al.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252623 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[X86] Do not try to custom-lower sitofp/fptosi in soft-float mode
Michael Kuperstein [Tue, 10 Nov 2015 17:37:49 +0000 (17:37 +0000)]
[X86] Do not try to custom-lower sitofp/fptosi in soft-float mode

Differential Revision: http://reviews.llvm.org/D14495

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252621 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoFix asan warning (NFC)
Xinliang David Li [Tue, 10 Nov 2015 17:11:33 +0000 (17:11 +0000)]
Fix asan warning (NFC)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252617 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoadd 'MustReduceDepth' as an objective/cost-metric for the MachineCombiner
Sanjay Patel [Tue, 10 Nov 2015 16:48:53 +0000 (16:48 +0000)]
add 'MustReduceDepth' as an objective/cost-metric for the MachineCombiner

This is one of the problems noted in PR25016:
https://llvm.org/bugs/show_bug.cgi?id=25016
and:
http://lists.llvm.org/pipermail/llvm-dev/2015-October/090998.html

The spilling problem is independent and not addressed by this patch.

The MachineCombiner was doing reassociations that don't improve or even worsen the critical path.
This is caused by inclusion of the "slack" factor when calculating the critical path of the original
code sequence. If we don't add that, then we have a more conservative cost comparison of the old code
sequence vs. a new sequence. The more liberal calculation must be preserved, however, for the AArch64
MULADD patterns because benchmark regressions were observed without that.

The two failing test cases now have identical asm that does what we want:
a + b + c + d ---> (a + b) + (c + d)

Differential Revision: http://reviews.llvm.org/D13417

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252616 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoReapply "[ARM] Combine CMOV into BFI where possible"
James Molloy [Tue, 10 Nov 2015 14:22:05 +0000 (14:22 +0000)]
Reapply "[ARM] Combine CMOV into BFI where possible"

Added fixes for stage2 failures: CMOV is not commutable; commuting the operands results in the condition being flipped! d'oh!

Original commit message:

If we have a CMOV, OR and AND combination such as:
  if (x & CN)
      y |= CM;

And:
  * CN is a single bit;
    * All bits covered by CM are known zero in y;

Then we can convert this to a sequence of BFI instructions. This will always be a win if CM is a single bit, will always be no worse than the TST & OR sequence if CM is two bits, and for thumb will be no worse if CM is three bits (due to the extra IT instruction).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252606 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoStrip metadata when speculatively hoisting instructions
Igor Laevsky [Tue, 10 Nov 2015 14:10:31 +0000 (14:10 +0000)]
Strip metadata when speculatively hoisting instructions

This is fix for PR24059.

When we are hoisting instruction above some condition it may turn out
that metadata on this instruction was control dependant on the condition.
This metadata becomes invalid and we need to drop it.

This patch should cover most obvious places of speculative execution (which
I have found by greping isSafeToSpeculativelyExecute). I think there are more
cases but at least this change covers the severe ones.

Differential Revision: http://reviews.llvm.org/D14398

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252604 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoUpdate test to use explicit triple
Oliver Stannard [Tue, 10 Nov 2015 14:09:08 +0000 (14:09 +0000)]
Update test to use explicit triple

This is needed for targets which do not support big-endian with the default
triple.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252603 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[PowerPC] Remove redundant code.
Tilmann Scheller [Tue, 10 Nov 2015 12:29:37 +0000 (12:29 +0000)]
[PowerPC] Remove redundant code.

The local variable Hi is never being read.

Issue identified by the Clang static analyzer.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252600 91177308-0d34-0410-b5e6-96231b3b80d8

8 years ago[AArch64] Fix halfword load merging for big-endian targets
Oliver Stannard [Tue, 10 Nov 2015 11:04:18 +0000 (11:04 +0000)]
[AArch64] Fix halfword load merging for big-endian targets

For big-endian targets, when we merge two halfword loads into a word load, the
order of the halfwords in the loaded value is reversed compared to
little-endian, so the load-store optimiser needs to swap the destination
registers.

This does not affect merging of two word loads, as we use ldp, which treats the
memory as two separate 32-bit words.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252597 91177308-0d34-0410-b5e6-96231b3b80d8

8 years agoInliner: Do zero-cost inlines even if above a negative threshold (PR24851)
Hans Wennborg [Tue, 10 Nov 2015 09:47:48 +0000 (09:47 +0000)]
Inliner: Do zero-cost inlines even if above a negative threshold (PR24851)

Differential Revision: http://reviews.llvm.org/D14499

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@252595 91177308-0d34-0410-b5e6-96231b3b80d8