platform/upstream/llvm.git
6 years ago[OPENMP, NVPTX] Reworked castToType() function, NFC.
Alexey Bataev [Mon, 19 Mar 2018 17:53:56 +0000 (17:53 +0000)]
[OPENMP, NVPTX] Reworked castToType() function, NFC.

Reworked function castToType to use more frontend functionality rather
than the backend.

llvm-svn: 327873

6 years ago[X86] Add MOV16ri*/MOV32ri*/MOV64ri* to scheduler models to match MOV8ri. Correct...
Craig Topper [Mon, 19 Mar 2018 17:46:59 +0000 (17:46 +0000)]
[X86] Add MOV16ri*/MOV32ri*/MOV64ri* to scheduler models to match MOV8ri. Correct SchedRW and itinerary for MOV32ri64.

llvm-svn: 327872

6 years ago[ELF] Add basic support for PPC LE
Zaara Syeda [Mon, 19 Mar 2018 17:40:14 +0000 (17:40 +0000)]
[ELF] Add basic support for PPC LE

This patch adds changes to start supporting the Power 64-Bit ELF V2 ABI.
This includes:
- Changing the ElfSym::GlobalOffsetTable to be named .TOC.
- Creating a GotHeader so the first entry in the .got is .TOC.
- Setting the e_flags to be 1 for ELF V1 and 2 for ELF V2

Differential Revision: https://reviews.llvm.org/D44483

llvm-svn: 327871

6 years ago[ObjC] Allow declaring __weak pointer fields in C structs in ARC.
Akira Hatanaka [Mon, 19 Mar 2018 17:38:40 +0000 (17:38 +0000)]
[ObjC] Allow declaring __weak pointer fields in C structs in ARC.

This patch uses the infrastructure added in r326307 for enabling
non-trivial fields to be declared in C structs to allow __weak fields in
C structs in ARC.

This recommits r327206, which was reverted because it caused
module-enabled builders to fail. I discovered that the
CXXRecordDecl::CanPassInRegisters flag wasn't being set correctly in
some cases after I moved it to RecordDecl.

Thanks to Eric Liu for helping me investigate the bug.

rdar://problem/33599681

https://reviews.llvm.org/D44095

llvm-svn: 327870

6 years ago[X86] Remove sse41 specific code from lowering v16i8 multiply
Craig Topper [Mon, 19 Mar 2018 17:31:41 +0000 (17:31 +0000)]
[X86] Remove sse41 specific code from lowering v16i8 multiply

With the SRAs removed from the SSE2 code in D44267, then there doesn't appear to be any advantage to the sse41 code. The punpcklbw instruction and pmovsx seem to have the same latency and throughput on most CPUs. And the SSE41 code requires moving the upper 64-bits into the lower 64-bit before the sign extend can be done. The unpckhbw in sse2 code can do better than that.

llvm-svn: 327869

6 years ago[OPENMP] Fix build with MSVC, NFC.
Alexey Bataev [Mon, 19 Mar 2018 17:18:13 +0000 (17:18 +0000)]
[OPENMP] Fix build with MSVC, NFC.

llvm-svn: 327868

6 years ago[OPENMP, NVPTX] Emit correct thread id.
Alexey Bataev [Mon, 19 Mar 2018 17:04:07 +0000 (17:04 +0000)]
[OPENMP, NVPTX] Emit correct thread id.

We emitted fake thread id for the outined function in NVPTX codegen.
Patch adds emission of the real thread id.

llvm-svn: 327867

6 years ago[X86] Make the multiply and divide itineraries more consistent.
Craig Topper [Mon, 19 Mar 2018 16:38:33 +0000 (16:38 +0000)]
[X86] Make the multiply and divide itineraries more consistent.

Sometimes we used the same itinerary for MEM and REG forms, but that seems inconsistent with our usual usage.

We also used the MUL8 itinerary for MULX32/64 which was also weird.

The test changes are because we were using IIC_IMUL32_RR and IIC_IMUL64_RR instead of IIC_IMUL32_REG/IIC_IMUL64_REG for the 32 and 64 bit multiplies that produce double width result.

llvm-svn: 327866

6 years agoXcode: Include DWARFUnit.{h,cpp} in the build
Vedant Kumar [Mon, 19 Mar 2018 16:24:58 +0000 (16:24 +0000)]
Xcode: Include DWARFUnit.{h,cpp} in the build

This should address a bot failure due to r327809.

llvm-svn: 327865

6 years agoRevert [MachineLICM] This reverts commit rL327856
Zaara Syeda [Mon, 19 Mar 2018 16:19:44 +0000 (16:19 +0000)]
Revert [MachineLICM] This reverts commit rL327856

Failing build bots. Revert the commit now.

llvm-svn: 327864

6 years ago[Driver] Avoid invalidated iterator in insertTargetAndModeArgs
Serge Pavlov [Mon, 19 Mar 2018 16:13:43 +0000 (16:13 +0000)]
[Driver] Avoid invalidated iterator in insertTargetAndModeArgs

Doing an .insert() can potentially invalidate iterators by reallocating the
vector's storage. When all the stars align just right, this causes segfaults
or glibc aborts.

Gentoo Linux bug (crashes while building Chromium): https://bugs.gentoo.org/650082.

Patch by Hector Martin!

Differential Revision: https://reviews.llvm.org/D44607

llvm-svn: 327863

6 years ago[CodeGen] Avoid handling DBG_VALUE in the LivePhysRegs (addUses,removeDefs,stepForward)
Matt Davis [Mon, 19 Mar 2018 16:06:40 +0000 (16:06 +0000)]
[CodeGen] Avoid handling DBG_VALUE in the LivePhysRegs (addUses,removeDefs,stepForward)

Summary:
This patch prevents DBG_VALUE instructions from influencing
LivePhysRegs::stepBackwards and stepForwards.  In at least one case,
specifically branch folding, the stepBackwards logic was having an
influence on code generation.  The result was that certain code
compiled with '-g -O2' would differ from that compiled with '-O2'
alone. It seems that the original logic, accounting for DBG_VALUE,
was influencing the placement of an IMPLICIT_DEF which had a later
impact on how blocks were processed in branch folding.

Reviewers: kparzysz, MatzeB

Reviewed By: kparzysz

Subscribers: bjope, llvm-commits

Tags: #debug-info

Differential Revision: https://reviews.llvm.org/D43850

llvm-svn: 327862

6 years ago[clang-format] Remove empty lines before }[;] // comment
Krasimir Georgiev [Mon, 19 Mar 2018 15:33:40 +0000 (15:33 +0000)]
[clang-format] Remove empty lines before }[;] // comment

Summary:
This addresses bug 36766 and a FIXME in tests about empty lines before
`}[;] // comment` lines.

Subscribers: klimek, cfe-commits

Differential Revision: https://reviews.llvm.org/D44631

llvm-svn: 327861

6 years ago[CMake] Use libc++ and compiler-rt as default libraries in Fuchsia toolchain
Petr Hosek [Mon, 19 Mar 2018 15:19:19 +0000 (15:19 +0000)]
[CMake] Use libc++ and compiler-rt as default libraries in Fuchsia toolchain

Fuchsia already defaults to libc++ and compiler-rt, but we want to use
these as default runtimes even on the host platform.

Differential Revision: https://reviews.llvm.org/D39930

llvm-svn: 327860

6 years ago[demangler] Recopy the demangler from libcxxabi.
Erik Pilkington [Mon, 19 Mar 2018 15:18:23 +0000 (15:18 +0000)]
[demangler] Recopy the demangler from libcxxabi.

Some significant work has gone into libcxxabi's copy of this file:
  - Uses an AST to represent mangled names.
  - Support/bugfixes for many C++ features.
  - Uses LLVM coding style.

llvm-svn: 327859

6 years ago[InstCombine] canonicalize fcmp+select to fabs
Sanjay Patel [Mon, 19 Mar 2018 15:14:30 +0000 (15:14 +0000)]
[InstCombine] canonicalize fcmp+select to fabs

This is complicated by -0.0 and nan. This is based on the DAG patterns
as shown in D44091. I'm hoping that we can just remove those DAG folds
and always rely on IR canonicalization to handle the matching to fabs.

We would still need to delete the broken code from DAGCombiner to fix
PR36600:
https://bugs.llvm.org/show_bug.cgi?id=36600

Differential Revision: https://reviews.llvm.org/D44550

llvm-svn: 327858

6 years agoMove code generation test case to test/CodeGen/
Tobias Grosser [Mon, 19 Mar 2018 15:05:30 +0000 (15:05 +0000)]
Move code generation test case to test/CodeGen/

llvm-svn: 327857

6 years ago[MachineLICM] Add functions to MachineLICM to hoist invariant stores
Zaara Syeda [Mon, 19 Mar 2018 14:52:25 +0000 (14:52 +0000)]
[MachineLICM] Add functions to MachineLICM to hoist invariant stores

This patch adds functions to allow MachineLICM to hoist invariant stores.
Currently, MachineLICM does not hoist any store instructions, however
when storing the same value to a constant spot on the stack, the store
instruction should be considered invariant and be hoisted. The function
isInvariantStore iterates each operand of the store instruction and checks
that each register operand satisfies isCallerPreservedPhysReg. The store
may be fed by a copy, which is hoisted by isCopyFeedingInvariantStore.
This patch also adds the PowerPC changes needed to consider the stack
register as caller preserved.

Differential Revision: https://reviews.llvm.org/D40196

llvm-svn: 327856

6 years ago[X86] Generalize schedule classes to support multiple stages
Simon Pilgrim [Mon, 19 Mar 2018 14:46:07 +0000 (14:46 +0000)]
[X86] Generalize schedule classes to support multiple stages

Currently the WriteResPair style multi-classes take a single pipeline stage and latency, this patch generalizes this to make it easier to create complex schedules with ResourceCycles and NumMicroOps be overriden from their defaults.

This has already been done for the Jaguar scheduler to remove a number of custom schedule classes and adding it to the other x86 targets will make it much tidier as we add additional classes in the future to try and replace so many custom cases.

I've converted some instructions but a lot of the models need a bit of cleanup after the patch has been committed - memory latencies not being consistent, the class not actually being used when we could remove some/all customs, etc. I'd prefer to keep this as NFC as possible so later patches can be smaller and target specific.

Differential Revision: https://reviews.llvm.org/D44612

llvm-svn: 327855

6 years agorun-clang-tidy: forward clang-tidy exit status
Miklos Vajna [Mon, 19 Mar 2018 14:43:59 +0000 (14:43 +0000)]
run-clang-tidy: forward clang-tidy exit status

Exit with a non-zero value in case any of the underlying clang-tidy
invocations exit with a non-zero value.

This is useful in case WarningsAsErrors is enabled for some of the
checks: if any of those checks find something, the exit status now
reflects that.

Also add the ability to use run-clang-tidy.py via lit, and assert that
the exit code is not 0 when modernize-use-auto is triggered
intentionally.

Reviewers: alexfh, aaron.ballman

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D44366

llvm-svn: 327854

6 years ago[x86] put nops into the WriteNop class and customize for Jaguar
Sanjay Patel [Mon, 19 Mar 2018 14:26:50 +0000 (14:26 +0000)]
[x86] put nops into the WriteNop class and customize for Jaguar

1. Given that we already have a classification bucket with 'nop' in the name,
   that's where 'nop' belongs. Right now, it's only used for prefix bytes and 'pause'.
2. Make the latency of this class '1' for Jaguar to tell the scheduler (and presumably
   llvm-mca) how to model the resource requirements better even though a nop has no
   dependencies.

Differential Revision: https://reviews.llvm.org/D44608

llvm-svn: 327853

6 years agoUpdated a usage of createTemporaryFile that does not expect file to be created.
Ilya Biryukov [Mon, 19 Mar 2018 14:20:25 +0000 (14:20 +0000)]
Updated a usage of createTemporaryFile that does not expect file to be created.

Summary:
This fixes a usage of createTemporaryFile in clang repo after
a change in llvm repo.

Reviewers: klimek, bkramer, krasimir, espindola, ilya-biryukov

Reviewed By: ilya-biryukov

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D36828

llvm-svn: 327852

6 years agoChanged createTemporaryFile without FD to actually create a file.
Ilya Biryukov [Mon, 19 Mar 2018 14:19:58 +0000 (14:19 +0000)]
Changed createTemporaryFile without FD to actually create a file.

Summary:
This commit changes semantics of createUniqueFile and
createTemporaryFile variants that do not return file descriptors.
Previously they only checked if files exist, therefore being subject
to race conditions. Now they will create an empty file to avoid them.

Functions that do not create a file are now called
getPotentiallyUniqueTempFileName and getPotentiallyUniqueFileName.

Reviewers: klimek, bkramer, krasimir, JDevlieghere, espindola

Reviewed By: klimek

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D36827

llvm-svn: 327851

6 years agoTableGen: Explicitly forbid self-references to field members
Nicolai Haehnle [Mon, 19 Mar 2018 14:14:28 +0000 (14:14 +0000)]
TableGen: Explicitly forbid self-references to field members

Summary:
Otherwise, patterns like in the test case produce cryptic error
messages about fields being resolved incompletely.

Change-Id: I713c0191f00fe140ad698675803ab1f8823dc5bd

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44476

llvm-svn: 327850

6 years agoTableGen: Check the dynamic type of !cast<Rec>(string)
Nicolai Haehnle [Mon, 19 Mar 2018 14:14:20 +0000 (14:14 +0000)]
TableGen: Check the dynamic type of !cast<Rec>(string)

Summary:
The docs already claim that this happens, but so far it hasn't. As a
consequence, existing TableGen files get this wrong a lot, but luckily
the fixes are all reasonably straightforward.

To make this work with all the existing forms of self-references (since
the true type of a record is only built up over time), the lookup of
self-references in !cast is delayed until the final resolving step.

Change-Id: If5923a72a252ba2fbc81a889d59775df0ef31164

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, javed.absar, llvm-commits

Differential Revision: https://reviews.llvm.org/D44475

llvm-svn: 327849

6 years agoTableGen: Explicitly test some cases of self-references and !cast errors
Nicolai Haehnle [Mon, 19 Mar 2018 14:14:10 +0000 (14:14 +0000)]
TableGen: Explicitly test some cases of self-references and !cast errors

Summary:
These are cases of self-references that exist today in practice. Let's
add tests for them to avoid regressions.

The self-references in PPCInstrInfo.td can be expressed in a simpler
way. Allowing this type of self-reference while at the same time
consistently doing late-resolve even for self-references is problematic
because there are references to fields that aren't in any class. Since
there's no need for this type of self-reference anyway, let's just
remove it.

Change-Id: I914e0b3e1ae7adae33855fac409b536879bc3f62

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: nemanjai, wdng, kbarton, llvm-commits

Differential Revision: https://reviews.llvm.org/D44474

llvm-svn: 327848

6 years agoTableGen: Only fold when some operand made resolve progress
Nicolai Haehnle [Mon, 19 Mar 2018 14:14:04 +0000 (14:14 +0000)]
TableGen: Only fold when some operand made resolve progress

Summary:
Make sure that we always fold immediately, so there's no point in
attempting to re-fold when nothing changes.

Change-Id: I069e1989455b6f2ca8606152f6adc1a5e817f1c8

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44198

llvm-svn: 327847

6 years agoTableGen: Remove OpInit::Fold
Nicolai Haehnle [Mon, 19 Mar 2018 14:13:59 +0000 (14:13 +0000)]
TableGen: Remove OpInit::Fold

Summary:
Virtual dispatch is not actually used anywhere.

Change-Id: I9829c5c59920ea27fb9bc17f1442156a3bb09a65

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44197

llvm-svn: 327846

6 years agoTableGen: Move GenStrConcat to a helper function in BinOpInit
Nicolai Haehnle [Mon, 19 Mar 2018 14:13:54 +0000 (14:13 +0000)]
TableGen: Move GenStrConcat to a helper function in BinOpInit

Summary:
Make it accessible for more users.

Change-Id: Ib05f09ba14e7942ced5d2f24b205efa285e40cd5

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44196

llvm-svn: 327845

6 years agoTableGen: Remove the cast-from-string-to-variable-reference feature
Nicolai Haehnle [Mon, 19 Mar 2018 14:13:37 +0000 (14:13 +0000)]
TableGen: Remove the cast-from-string-to-variable-reference feature

Summary:
Cast-from-string for records isn't going away, but cast-from-string for
variables is a pretty dodgy feature to have, especially when referencing
template arguments. It's doubtful that this ever worked in a reliable
way, and nobody seems to be using it, so let's get rid of it and get
some related cleanups.

Change-Id: I395ac8a43fef4cf98e611f2f552300d21e99b66a

Reviewers: arsenm, craig.topper, tra, MartinO

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44195

llvm-svn: 327844

6 years agoAMDGPU/GlobalISel: RegBankSelect for basic int ops
Matt Arsenault [Mon, 19 Mar 2018 14:07:23 +0000 (14:07 +0000)]
AMDGPU/GlobalISel: RegBankSelect for basic int ops

llvm-svn: 327843

6 years agoAMDGPU: Don't leave dead illegal VGPR->SGPR copies
Matt Arsenault [Mon, 19 Mar 2018 14:07:15 +0000 (14:07 +0000)]
AMDGPU: Don't leave dead illegal VGPR->SGPR copies

Normally DCE kills these, but at -O0 these get left behind
leaving suspicious looking illegal copies.

Replace with IMPLICIT_DEF to avoid iterator issues.

llvm-svn: 327842

6 years ago[NFC] Fix minor typos in comments
Karl-Johan Karlsson [Mon, 19 Mar 2018 13:48:40 +0000 (13:48 +0000)]
[NFC] Fix minor typos in comments

llvm-svn: 327841

6 years ago[MergeICmps] Re-land 324317 "Enable the MergeICmps Pass by default."
Clement Courbet [Mon, 19 Mar 2018 13:37:04 +0000 (13:37 +0000)]
[MergeICmps] Re-land 324317 "Enable the MergeICmps Pass by default."

Now that PR36557 is fixed.

llvm-svn: 327840

6 years ago[ARM] Support for v4f16 and v8f16 vectors
Sjoerd Meijer [Mon, 19 Mar 2018 13:35:25 +0000 (13:35 +0000)]
[ARM] Support for v4f16 and v8f16 vectors

This is the groundwork for adding the Armv8.2-A FP16 vector intrinsics, which
uses v4f16 and v8f16 vector operands and return values. All the moving parts
are tested with two intrinsics, a 1-operand v8f16 and a 2-operand v4f16
intrinsic. In a follow-up patch the rest of the intrinsics and tests will be
added.

Differential Revision: https://reviews.llvm.org/D44538

llvm-svn: 327839

6 years agoStylish change. NFC
Xin Tong [Mon, 19 Mar 2018 13:35:23 +0000 (13:35 +0000)]
Stylish change. NFC

llvm-svn: 327838

6 years ago[llvm-mca] Add pipeline stall events.
Andrea Di Biagio [Mon, 19 Mar 2018 13:23:07 +0000 (13:23 +0000)]
[llvm-mca] Add pipeline stall events.

This patch introduces a new class named HWStallEvent (see HWEventListener.h),
and updates the event listener interface. A HWStallEvent represents a pipeline
stall caused by the lack of hardware resources. Similarly to HWInstructionEvent,
the event type is an unsigned, and the exact meaning depends on the subtarget.
At the moment, HWStallEvent supports a few generic dispatch events.

The main goals of this patch is to remove the logic that counts dispatch stalls
from the DispatchUnit to the BackendStatistics view.

Previously, DispatchUnit was responsible for counting and classifying dispatch
stall events. With this patch, we delegate the task of counting and classifying
stall events to the listeners (i.e. in our case, it is view
"BackendStatistics"). So, the DispatchUnit doesn't have to do extra
(unnecessary) bookkeeping.

This patch also helps futher simplifying the Backend interface. Now class
BackendStatistics no longer has to query the Backend interface to obtain the
number of dispatch stalls. As a consequence, we can get rid of all the
'getNumXXX()' methods from class Backend.
The long term goal is to remove all the remaining dependencies between the
Backend and the BackendStatistics interface.

Differential Revision: https://reviews.llvm.org/D44621

llvm-svn: 327837

6 years ago[ARM] Pass half or i16 types for NEON intrinsics
Sjoerd Meijer [Mon, 19 Mar 2018 13:22:49 +0000 (13:22 +0000)]
[ARM] Pass half or i16 types for NEON intrinsics

For generating NEON intrinsics, this determines the NEON data type, and whether
it should be a half type or an i16 type. I.e., we always pass a half type for
AArch64, this hasn't changed, but now also for ARM but only when FullFP16 is
enabled, and i16 otherwise.

This is intended to be non-functional change, but together with the backend
work in D44538 which adds support for f16 vectors, this enables adding the
AArch32 FP16 (vector) intrinsics.

Differential Revision: https://reviews.llvm.org/D44561

llvm-svn: 327836

6 years agobuild_llvm_package.bat: Drop LLDB from the package.
Hans Wennborg [Mon, 19 Mar 2018 13:05:37 +0000 (13:05 +0000)]
build_llvm_package.bat: Drop LLDB from the package.

I don't think anyone ever got this to work, what with getting exactly
the right Python dependency and so on. Removing it simplifies the
script, removes a number of hairy dependencies, and cuts ~30 MB off the
installer size.

llvm-svn: 327835

6 years ago[SystemZ] Bugfix of CC liveness in emitMemMemWrapper (CLC).
Jonas Paulsson [Mon, 19 Mar 2018 13:05:22 +0000 (13:05 +0000)]
[SystemZ]  Bugfix of CC liveness in emitMemMemWrapper (CLC).

If DoneMBB becomes empty it must have CC added to its live-in list, since it
will fall-through into EndMBB. This happens when the CLC loop does the
complete range.

Review: Ulrich Weigand
llvm-svn: 327834

6 years ago[clang-tidy] New check bugprone-unused-return-value
Alexander Kornienko [Mon, 19 Mar 2018 13:02:32 +0000 (13:02 +0000)]
[clang-tidy] New check bugprone-unused-return-value

Summary:
Detects function calls where the return value is unused.

Checked functions can be configured.

Reviewers: alexfh, aaron.ballman, ilya-biryukov, hokein

Reviewed By: alexfh, aaron.ballman

Subscribers: hintonda, JonasToth, Eugene.Zelenko, mgorny, xazax.hun, cfe-commits

Tags: #clang-tools-extra

Patch by Kalle Huttunen!

Differential Revision: https://reviews.llvm.org/D41655

llvm-svn: 327833

6 years agoHexagonISelLowering.cpp: fix 'enum in bool context' warning
Hans Wennborg [Mon, 19 Mar 2018 12:55:58 +0000 (12:55 +0000)]
HexagonISelLowering.cpp: fix 'enum in bool context' warning

llvm-svn: 327832

6 years ago[RISCV] Peephole optimisation for load/store of global values or constant addresses
Alex Bradbury [Mon, 19 Mar 2018 11:54:28 +0000 (11:54 +0000)]
[RISCV] Peephole optimisation for load/store of global values or constant addresses

(load (add base, off), 0) -> (load base, off)
(store val, (add base, off)) -> (store val, base, off)

This is similar to an equivalent peephole optimisation in PPCISelDAGToDAG.

llvm-svn: 327831

6 years ago[MSan] fix the types of RegSaveAreaPtrPtr and OverflowArgAreaPtrPtr
Alexander Potapenko [Mon, 19 Mar 2018 10:08:04 +0000 (10:08 +0000)]
[MSan] fix the types of RegSaveAreaPtrPtr and OverflowArgAreaPtrPtr

Despite their names, RegSaveAreaPtrPtr and OverflowArgAreaPtrPtr
used to be i8* instead of i8**.

This is important, because these pointers are dereferenced twice
(first in CreateLoad(), then in getShadowOriginPtr()), but for some
reason MSan allowed this - most certainly because it was possible
to optimize getShadowOriginPtr() away at compile time.

Differential revision: https://reviews.llvm.org/D44520

llvm-svn: 327830

6 years ago[MSan] Don't create zero offsets in getShadowPtrForArgument(). NFC
Alexander Potapenko [Mon, 19 Mar 2018 10:03:47 +0000 (10:03 +0000)]
[MSan] Don't create zero offsets in getShadowPtrForArgument(). NFC

For MSan instrumentation with MS.ParamTLS and MS.ParamOriginTLS being
TLS variables, the CreateAdd() with ArgOffset==0 is a no-op, because
the compiler is able to fold the addition of 0.

But for KMSAN, which receives ParamTLS and ParamOriginTLS from a call
to the runtime library, this introduces a stray instruction which
complicates reading/testing the IR.

Differential revision: https://reviews.llvm.org/D44514

llvm-svn: 327829

6 years ago[MSan] Introduce insertWarningFn(). NFC
Alexander Potapenko [Mon, 19 Mar 2018 09:59:44 +0000 (09:59 +0000)]
[MSan] Introduce insertWarningFn(). NFC

This is a step towards the upcoming KMSAN implementation patch.
KMSAN is going to use a different warning function,
__msan_warning_32(uptr origin), so we'd better create the warning
calls in one place.

Differential Revision: https://reviews.llvm.org/D44513

llvm-svn: 327828

6 years ago[ARM] Fix warnings about missing parentheses in ARMAsmParser
Mikhail Maltsev [Mon, 19 Mar 2018 09:48:58 +0000 (09:48 +0000)]
[ARM] Fix warnings about missing parentheses in ARMAsmParser

llvm-svn: 327827

6 years ago[SCEV] Factor out isKnownViaInduction. NFC.
Serguei Katkov [Mon, 19 Mar 2018 08:32:09 +0000 (08:32 +0000)]
[SCEV] Factor out isKnownViaInduction. NFC.

This just extracts the isKnownViaInduction from isKnownPredicate.

Reviewers: sanjoy, mkazantsev, reames
Reviewed By: mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44554

llvm-svn: 327824

6 years ago[ELF] Recommit 327248 with Arm using the .got for _GLOBAL_OFFSET_TABLE_
Peter Smith [Mon, 19 Mar 2018 06:52:51 +0000 (06:52 +0000)]
[ELF] Recommit 327248 with Arm using the .got for _GLOBAL_OFFSET_TABLE_

This is the same as 327248 except Arm defining _GLOBAL_OFFSET_TABLE_ to
be the base of the .got section as some existing code is relying upon it.

For most Targets the _GLOBAL_OFFSET_TABLE_ symbol is expected to be at
the start of the .got.plt section so that _GLOBAL_OFFSET_TABLE_[0] =
reserved value that is by convention the address of the dynamic section.
Previously we had defined _GLOBAL_OFFSET_TABLE_ as either the start or end
of the .got section with the intention that the .got.plt section would
follow the .got. However this does not always hold with the current
default section ordering so _GLOBAL_OFFSET_TABLE_[0] may not be consistent
with the reserved first entry of the .got.plt.

X86, X86_64 and AArch64 will use the .got.plt. Arm, Mips and Power use .got

Fixes PR36555

Differential Revision: https://reviews.llvm.org/D44259

llvm-svn: 327823

6 years ago[SCEV] Re-land: Fix isKnownPredicate
Serguei Katkov [Mon, 19 Mar 2018 06:35:30 +0000 (06:35 +0000)]
[SCEV] Re-land: Fix isKnownPredicate

This is re-land of https://reviews.llvm.org/rL327362 with a fix
and regression test.

The crash was due to it is possible that for found MDL loop,
LHS or RHS may contain an invariant unknown SCEV which
does not dominate the MDL. Please see regression
test for an example.

Reviewers: sanjoy, mkazantsev, reames
Reviewed By: mkazantsev
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D44553

llvm-svn: 327822

6 years ago[X86] Merge XADD8rr regular expression with XADD16rr/XADD32rr/XADD64rr in a couple...
Craig Topper [Mon, 19 Mar 2018 04:21:42 +0000 (04:21 +0000)]
[X86] Merge XADD8rr regular expression with XADD16rr/XADD32rr/XADD64rr in a couple scheduler models.

llvm-svn: 327821

6 years ago[X86] Add ADD16i16/ADD32i32/ADD64i32 and similar to the scheduler models to match...
Craig Topper [Mon, 19 Mar 2018 04:21:40 +0000 (04:21 +0000)]
[X86] Add ADD16i16/ADD32i32/ADD64i32 and similar to the scheduler models to match ADD8i8.

Also move ADC8i8 and SBB8i8 in the Sandy Bridge model to the same class as ADC8ri and SBB8ri. That seems more accurate since its the 8i8 is just the register forced to AL instead of coming from modrm.

llvm-svn: 327820

6 years ago[X6] Remove two unused InstrItinClass
Craig Topper [Mon, 19 Mar 2018 02:07:32 +0000 (02:07 +0000)]
[X6] Remove two unused InstrItinClass

llvm-svn: 327819

6 years agoremainder: Port from amd builtins
Jan Vesely [Mon, 19 Mar 2018 01:01:10 +0000 (01:01 +0000)]
remainder: Port from amd builtins

Mostly ported from amd_builtins, uses only denormal path for fp32.
Passes CTS on carrizo and turks

Reviewer: Aaron Watry <awatry@gmail.com>
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 327818

6 years ago[X86] Use IIC_CMOV64_RR/RM on 64-bit cmov instructions.
Craig Topper [Mon, 19 Mar 2018 00:56:12 +0000 (00:56 +0000)]
[X86] Use IIC_CMOV64_RR/RM on 64-bit cmov instructions.

llvm-svn: 327817

6 years ago[X86] Merge 32 and 64-bit RORX/SHLX/SARX/SHRX into single regular expressions in...
Craig Topper [Mon, 19 Mar 2018 00:56:11 +0000 (00:56 +0000)]
[X86] Merge 32 and 64-bit RORX/SHLX/SARX/SHRX into single regular expressions in scheduler models.

llvm-svn: 327816

6 years ago[X86] Merge 8-bit instructions into instregex with 16/32/64 instructions in the sched...
Craig Topper [Mon, 19 Mar 2018 00:56:09 +0000 (00:56 +0000)]
[X86] Merge 8-bit instructions into instregex with 16/32/64 instructions in the scheduler models as much as possible. NFCI

This reduces the total number of generated scheduler classes from 5404 to 5316.

llvm-svn: 327815

6 years ago[AVR] Lower i128 divisions to runtime library calls
Dylan McKay [Mon, 19 Mar 2018 00:55:50 +0000 (00:55 +0000)]
[AVR] Lower i128 divisions to runtime library calls

This patch adds i128 division support by instruction LLVM to lower
128-bit divisions to the __udivmodti4 and __divmodti4 rtlib functions.

This also adds test for 64-bit division and 128-bit division.

Patch by Peter Nimmervoll.

llvm-svn: 327814

6 years ago[Mips] Remove duplicate lines from MipsScheduleP5600.td and enable FullInstRWOverlapC...
Craig Topper [Sun, 18 Mar 2018 22:16:54 +0000 (22:16 +0000)]
[Mips] Remove duplicate lines from MipsScheduleP5600.td and enable FullInstRWOverlapCheck.

This fixes the errors found by the new check added in r327808.

llvm-svn: 327813

6 years ago[AArch64] Fix a few InstRWs in the A53 scheduler model and enable FullInstRWOverlapCheck.
Craig Topper [Sun, 18 Mar 2018 22:16:53 +0000 (22:16 +0000)]
[AArch64] Fix a few InstRWs in the A53 scheduler model and enable FullInstRWOverlapCheck.

This fixes the errors found by the new check added in r327808.

llvm-svn: 327812

6 years ago[SelectionDAG] Don't default the SelectionDAG* parameter to SDValue::dump to nullptr...
Craig Topper [Sun, 18 Mar 2018 21:28:11 +0000 (21:28 +0000)]
[SelectionDAG] Don't default the SelectionDAG* parameter to SDValue::dump to nullptr. Use two different signatures instead.

This matches what we do in SDNode.

This should allow SDValue::dump to be used in the debugger without getting an error if you don't pass an argument.

llvm-svn: 327811

6 years agoMove the codebase to use: DWARFCompileUnit -> DWARFUnit
Jan Kratochvil [Sun, 18 Mar 2018 20:11:02 +0000 (20:11 +0000)]
Move the codebase to use: DWARFCompileUnit -> DWARFUnit

Now the codebase can use the DWARFUnit superclass. It will make it later
seamlessly work also with DWARFPartialUnit for DWZ.

This patch is only a search-and-replace easily undone, nothing interesting
in it.

Differential revision: https://reviews.llvm.org/D42892

llvm-svn: 327810

6 years agoDWARFUnit split out of DWARFCompileUnit
Jan Kratochvil [Sun, 18 Mar 2018 20:09:02 +0000 (20:09 +0000)]
DWARFUnit split out of DWARFCompileUnit

DW_TAG_partial_unit for DWZ can be then presented by DWARFPartialUnit also
inherited from DWARFUnit.

Differential revision: https://reviews.llvm.org/D40466

llvm-svn: 327809

6 years ago[TableGen] When trying to reuse a scheduler class for instructions from an InstRW...
Craig Topper [Sun, 18 Mar 2018 19:56:15 +0000 (19:56 +0000)]
[TableGen] When trying to reuse a scheduler class for instructions from an InstRW, make sure we haven't already seen another InstRW containing this instruction on this CPU.

This is similar to the check later when we remap some of the instructions from one class to a new one. But if we reuse the class we don't get to do that check.

So many CPUs have violations of this check that I had to add a flag to the SchedMachineModel to allow it to be disabled. Hopefully we can get those cleaned up quickly and remove this flag.

A lot of the violations are due to overlapping regular expressions, but that's not the only kind of issue it found.

llvm-svn: 327808

6 years ago[X86][Btver2] Fix crc32 schedule costs
Simon Pilgrim [Sun, 18 Mar 2018 19:54:42 +0000 (19:54 +0000)]
[X86][Btver2] Fix crc32 schedule costs

The default is currently FAdd for some reason

llvm-svn: 327807

6 years agoUpdated C++2a status with changes from Jacksonville WG21 meeting
Marshall Clow [Sun, 18 Mar 2018 19:29:21 +0000 (19:29 +0000)]
Updated C++2a status with changes from Jacksonville WG21 meeting

llvm-svn: 327806

6 years ago[X86][Btver2] Add crc32 resource tests
Simon Pilgrim [Sun, 18 Mar 2018 18:55:34 +0000 (18:55 +0000)]
[X86][Btver2] Add crc32 resource tests

llvm-svn: 327805

6 years ago[X86][Btver2] FADD/FHADD ymm instructions are double pumped on the JFPA functional...
Simon Pilgrim [Sun, 18 Mar 2018 18:45:57 +0000 (18:45 +0000)]
[X86][Btver2] FADD/FHADD ymm instructions are double pumped on the JFPA functional pipe

llvm-svn: 327804

6 years ago[X86][Btver2] Float bitwise ymm instructions are double pumped on the JFPX (JFPA...
Simon Pilgrim [Sun, 18 Mar 2018 17:10:12 +0000 (17:10 +0000)]
[X86][Btver2] Float bitwise ymm instructions are double pumped on the JFPX (JFPA/JFPM) functional pipes

llvm-svn: 327803

6 years agoResolve unused variable 'VR' warning in RetainCountChecker.cpp
Bjorn Pettersson [Sun, 18 Mar 2018 16:07:20 +0000 (16:07 +0000)]
Resolve unused variable 'VR' warning in RetainCountChecker.cpp

Getting rid of
  error: unused variable 'VR' [-Werror,-Wunused-variable]
warning/error at
  lib/StaticAnalyzer/Checkers/RetainCountChecker.cpp:1933

llvm-svn: 327802

6 years ago[X86][Btver2] F16C instructions are performed on the JSTC functional pipe
Simon Pilgrim [Sun, 18 Mar 2018 15:59:51 +0000 (15:59 +0000)]
[X86][Btver2] F16C instructions are performed on the JSTC functional pipe

llvm-svn: 327801

6 years ago[LICM] Salvage DI from dying Instructions
Anastasis Grammenos [Sun, 18 Mar 2018 15:59:19 +0000 (15:59 +0000)]
[LICM] Salvage DI from dying Instructions

LICM deletes trivially dead instructions which it won't attempt to sink.
Attempt to salvage debug values which reference these instructions.

llvm-svn: 327800

6 years ago[InstCombine] peek through unsigned FP casts for zero-equality compares (PR36682)
Roman Lebedev [Sun, 18 Mar 2018 15:53:02 +0000 (15:53 +0000)]
[InstCombine] peek through unsigned FP casts for zero-equality compares (PR36682)

Summary:
This pattern came up in PR36682 / D44390
https://bugs.llvm.org/show_bug.cgi?id=36682
https://reviews.llvm.org/D44390
https://godbolt.org/g/oKvT5H

See also D44416

Reviewers: spatel, majnemer, efriedma, arsenm

Reviewed By: spatel

Subscribers: wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D44424

llvm-svn: 327799

6 years ago[llvm-mca] Allow the definition of multiple register files.
Andrea Di Biagio [Sun, 18 Mar 2018 15:33:27 +0000 (15:33 +0000)]
[llvm-mca] Allow the definition of multiple register files.

This is a refactoring in preparation for other two changes that will allow
scheduling models to define multiple register files. This is the first step
towards fixing PR36662.

class RegisterFile (in Dispatch.h) now can emulate multiple register files.
Internally, it tracks the number of available physical registers in each
register file (described by class RegisterFileInfo).

Each register file is associated to a list of MCRegisterClass indices. Knowing
the register class indices allows to map physical registers to register files.

The long term goal is to allow processor models to optionally specify how many
register files are implemented via tablegen.

Differential Revision: https://reviews.llvm.org/D44488

llvm-svn: 327798

6 years ago[InstCombine] add nnan requirement for sqrt(x) * sqrt(y) -> sqrt(x*y)
Sanjay Patel [Sun, 18 Mar 2018 14:32:54 +0000 (14:32 +0000)]
[InstCombine] add nnan requirement for sqrt(x) * sqrt(y) -> sqrt(x*y)

This is similar to D43765.

llvm-svn: 327797

6 years ago[InstSimplify] loosen FMF for sqrt(X) * sqrt(X) --> X
Sanjay Patel [Sun, 18 Mar 2018 14:12:25 +0000 (14:12 +0000)]
[InstSimplify] loosen FMF for sqrt(X) * sqrt(X) --> X

As shown in the code comment, we don't need all of 'fast',
but we do need reassoc + nsz + nnan.

Differential Revision: https://reviews.llvm.org/D43765

llvm-svn: 327796

6 years ago[X86][Btver2] Strip default latency/resource values. NFCI.
Simon Pilgrim [Sun, 18 Mar 2018 13:16:11 +0000 (13:16 +0000)]
[X86][Btver2] Strip default latency/resource values. NFCI.

llvm-svn: 327795

6 years ago[X86][Btver2] SSE4A EXTRQ/INSERTQ instructions are performed on the JVALU0/JVALU1...
Simon Pilgrim [Sun, 18 Mar 2018 13:05:09 +0000 (13:05 +0000)]
[X86][Btver2] SSE4A EXTRQ/INSERTQ instructions are performed on the JVALU0/JVALU1 functional pipes

llvm-svn: 327794

6 years ago[X86][Btver2] Modelled float bitwise instructions as being performed on the float...
Simon Pilgrim [Sun, 18 Mar 2018 12:37:35 +0000 (12:37 +0000)]
[X86][Btver2] Modelled float bitwise instructions as being performed on the float cluster (FPA/FPM) not the integer.

llvm-svn: 327793

6 years ago[dsymutil] Fix add_llvm_tool_symlink
Jonas Devlieghere [Sun, 18 Mar 2018 12:27:05 +0000 (12:27 +0000)]
[dsymutil] Fix add_llvm_tool_symlink

Update the arguments to add_llvm_tool_symlink to symlink llvm-dsymutil
to dsymutil.

llvm-svn: 327792

6 years ago[X86][Btver2] Correctly distinguish between scheduling pipe and functional unit for...
Simon Pilgrim [Sun, 18 Mar 2018 12:09:17 +0000 (12:09 +0000)]
[X86][Btver2] Correctly distinguish between scheduling pipe and functional unit for JWriteResFpuPair defs

Jaguar's FPU has 2 scheduler pipes (JFPU0/JFPU1) which forward to multiple functional sub-units each. We need to model that an micro-op will both consume the scheduler pipe and a functional unit.

This patch just handles the ops defined through JWriteResFpuPair, I'll go through the custom cases later.

llvm-svn: 327791

6 years ago[dsymutil] Rename llvm-dsymutil -> dsymutil
Jonas Devlieghere [Sun, 18 Mar 2018 11:38:41 +0000 (11:38 +0000)]
[dsymutil] Rename llvm-dsymutil -> dsymutil

Now that almost all functionality of Apple's dsymutil has been
upstreamed, the open source variant can be used as a drop in
replacement. Hence we feel it's no longer necessary to have the llvm
prefix.

Differential revision: https://reviews.llvm.org/D44527

llvm-svn: 327790

6 years ago[X86][Btver2] Merge equivalent VBLENDVY + VPERMILY schedule groups
Simon Pilgrim [Sun, 18 Mar 2018 10:22:35 +0000 (10:22 +0000)]
[X86][Btver2] Merge equivalent VBLENDVY + VPERMILY schedule groups

Thanks to Craig Topper for noticing this.

llvm-svn: 327789

6 years ago[X86][Btver2] Add llvm-mca tests to show pipe resource usage of most vector instructions
Simon Pilgrim [Sun, 18 Mar 2018 09:32:38 +0000 (09:32 +0000)]
[X86][Btver2] Add llvm-mca tests to show pipe resource usage of most vector instructions

Hopefully these tests can be easily reused should any other subtarget get in depth llvm-mca coverage (we can either copy the tests or move them into a common dir and run it with multiple prefixes).

llvm-svn: 327788

6 years ago[X86] Fix a bunch of overlapping regular expressions in the scheduler models.
Craig Topper [Sun, 18 Mar 2018 08:38:06 +0000 (08:38 +0000)]
[X86] Fix a bunch of overlapping regular expressions in the scheduler models.

llvm-svn: 327787

6 years ago[X86] Fix a couple typos in the Zen scheduler model.
Craig Topper [Sun, 18 Mar 2018 08:38:04 +0000 (08:38 +0000)]
[X86] Fix a couple typos in the Zen scheduler model.

llvm-svn: 327786

6 years ago[TableGen] Remove unnecessary uses of make_range.
Craig Topper [Sun, 18 Mar 2018 08:38:03 +0000 (08:38 +0000)]
[TableGen] Remove unnecessary uses of make_range.

llvm-svn: 327785

6 years ago[TableGen] Move some variables into for loop declaration. NFC
Craig Topper [Sun, 18 Mar 2018 08:38:02 +0000 (08:38 +0000)]
[TableGen] Move some variables into for loop declaration. NFC

They aren't needed after the loop.

llvm-svn: 327784

6 years ago[X86] Remove MMX_MASKMOVQ64 and VMASKMOVDQU from scheduler models.
Craig Topper [Sun, 18 Mar 2018 03:24:42 +0000 (03:24 +0000)]
[X86] Remove MMX_MASKMOVQ64 and VMASKMOVDQU from scheduler models.

The information was so wildly inaccurate and incomplete its better to just remove it.

MMX_MASKMOVQ64 showed up twice in several scheduler models. In Haswell and Broadwell they were on adjacent lines. On Skylake the copies had different information.

MMX_MASKMOVQ and MASKMOVDQU were completely missing.

MMX_MASKMOVQ64 was listed on Haswell/Broadwell as 1 cycle on port 1 despite it being a store instruction.

Filed PR36780 to track fixing this right.

llvm-svn: 327783

6 years ago[C++17] Allow an empty expression in an if init statement
Zhihao Yuan [Sat, 17 Mar 2018 21:42:10 +0000 (21:42 +0000)]
[C++17] Allow an empty expression in an if init statement

Summary:
This fixes [PR35381](https://llvm.org/pr35381) and an additional bug where clang didn't warn about the C++17 extension when having an expression in the init statement.

Thanks Nicolas Lesser for contributing the patch.

Reviewers: rsmith

Reviewed By: rsmith

Subscribers: erik.pilkington, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D40445

llvm-svn: 327782

6 years agoImplement DR2229, which prohibits unnamed bit-fields from having qualifiers in C++.
Aaron Ballman [Sat, 17 Mar 2018 21:08:40 +0000 (21:08 +0000)]
Implement DR2229, which prohibits unnamed bit-fields from having qualifiers in C++.

llvm-svn: 327781

6 years agoFix codegen for structured binding binding in conditions
Zhihao Yuan [Sat, 17 Mar 2018 21:01:27 +0000 (21:01 +0000)]
Fix codegen for structured binding binding in conditions

Summary:
The codegen for conditions assumes that a normal variable declaration is used in a condition, but this is not the case when a structured binding is used.

This fixes [PR36747](http://llvm.org/pr36747).

Thanks Nicolas Lesser for contributing the patch.

Reviewers: lichray, rsmith

Reviewed By: lichray

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D44534

llvm-svn: 327780

6 years ago[AArch64] Skip an unnecessary getCopyToReg in DYNAMIC_STACKALLOC
Martin Storsjo [Sat, 17 Mar 2018 20:08:48 +0000 (20:08 +0000)]
[AArch64] Skip an unnecessary getCopyToReg in DYNAMIC_STACKALLOC

Differential Revision: https://reviews.llvm.org/D44586

llvm-svn: 327779

6 years agoRevert "[DAG, X86] Revert r327197 "Revert r327170, r327171, r327172""
Nirav Dave [Sat, 17 Mar 2018 19:24:54 +0000 (19:24 +0000)]
Revert "[DAG, X86] Revert r327197 "Revert r327170, r327171, r327172""

as it times out building test-suite on PPC.

llvm-svn: 327778

6 years ago[DAG, X86] Revert r327197 "Revert r327170, r327171, r327172"
Nirav Dave [Sat, 17 Mar 2018 17:42:10 +0000 (17:42 +0000)]
[DAG, X86] Revert r327197 "Revert r327170, r327171, r327172"

Reland ISel cycle checking improvements after simplifying and reducing
node id invariant traversal.

llvm-svn: 327777

6 years agoFix some user facing typos
Sylvestre Ledru [Sat, 17 Mar 2018 17:30:08 +0000 (17:30 +0000)]
Fix some user facing typos

llvm-svn: 327776

6 years ago[Driver] Fix the descriptions for -Tdata and -Ttext options
Aaron Smith [Sat, 17 Mar 2018 15:24:35 +0000 (15:24 +0000)]
[Driver] Fix the descriptions for -Tdata and -Ttext options

Reviewers: llvm-commits

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D44581

llvm-svn: 327775

6 years agoAMDGPU/GlobalISel: Cleanup constant legality
Matt Arsenault [Sat, 17 Mar 2018 15:17:48 +0000 (15:17 +0000)]
AMDGPU/GlobalISel: Cleanup constant legality

llvm-svn: 327774

6 years agoAMDGPU/GlobalISel: Basic G_GEP legality
Matt Arsenault [Sat, 17 Mar 2018 15:17:45 +0000 (15:17 +0000)]
AMDGPU/GlobalISel: Basic G_GEP legality

llvm-svn: 327773

6 years agoAMDGPU/GlobalISel: Basic legality for load/store
Matt Arsenault [Sat, 17 Mar 2018 15:17:41 +0000 (15:17 +0000)]
AMDGPU/GlobalISel: Basic legality for load/store

llvm-svn: 327772