Commit Graph

330484 Commits

Author SHA1 Message Date
Craig Topper 1ce8a5b385 [X86] Only look up boolean reduction cost tables if the reduction is not pairwise.
Summary:
We don't pattern match pairwise shuffles in SelectionDAG. So we
should only return the optimized costs if its not a pairwise
shuffle.

I think SLP vectorizer gives priority to non pairwise shuffle if
the cost is the same. And the look up for reduction intrinsics
passes false for the pairwise flag. So this probably has no real
effect today.

Reviewers: RKSimon

Reviewed By: RKSimon

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69083
2019-10-26 16:41:19 -07:00
Roman Lebedev 9d77ad5754
[APInt] Introduce APIntOps::GetMostSignificantDifferentBit()
Summary:
Compare two values, and if they are different, return the position of the
most significant bit that is different in the values.

Needed for D69387.

Reviewers: nikic, spatel, sanjoy, RKSimon

Reviewed By: nikic

Subscribers: xbolva00, hiraditya, dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69439
2019-10-26 23:20:58 +03:00
David Zarzycki 11c920207a [X86] Prefer KORTEST on Knights Landing or later for memcmp()
PTEST and especially the MOVMSK instructions are slow on Knights Landing
or later. As a bonus, this patch increases instruction parallelism by
emitting:
    KORTEST(PCMPNEQ(a, b), PCMPNEQ(c, d)) == 0
Instead of:
    KORTEST(AND(PCMPEQ(a, b), PCMPEQ(c, d))) == ~0

https://reviews.llvm.org/D69157
2019-10-26 21:14:57 +03:00
David Zarzycki 0d0509384f [X86] NFC: expand inline memcmp test coverage
1) Adds SSE4.1 coverage.
2) Adds prefer-256-bit or not coverage.
3) Adds more power-of-two tests up to 512 bytes.
4) Adds power-of-two-minus-one tests to verify overlapping loads.
5) Adds power-of-two-plus-one-half tests (48, 96, 192, and 384).
6) Adds greater-than/less-than tests from 16 to 512 bytes.

https://reviews.llvm.org/D69222
2019-10-26 21:14:57 +03:00
marshall 82b82e39c4 Add all the issues to be voted upon in Belfast 2019-10-26 07:53:15 -07:00
Luboš Luňák 9231e44344 fix lldb build with -DLLVM_ENABLE_MODULES=On (missing #include)
The error message says to add the #include.
2019-10-26 15:34:45 +02:00
Georgii Rymar 073ab70b72 [ObjectYAML] - Do not use auto. NFC.
Using 'auto' when the type is not obvious is undesired.

(it is just a test commit actually)
2019-10-26 15:08:49 +03:00
georgerim 05a9e344ee [YAMLTraits] - Revert a change committed by a mistake in D68983
I've accidentally reverted one of my previous patches.
It was not catched by bots because (I guess) they do not
build in debug (we have a test case which triggers an assert
in MSVS when runs without this change).
More info: https://reviews.llvm.org/D68983#inline-624235

Reported by Jordan Rupprecht.
2019-10-26 14:52:12 +03:00
cdevadas e921ede540 [AMDGPU] Fix Vreg_1 PHI lowering in SILowerI1Copies.
There is a minor flaw in the implementation of function lowerPhis.
This function replaces values of regclass Vreg_1 (boolean values)
involved in PHIs into an SGPR. Currently it iterates over the MBBs
and performs an inplace lowering of PHIs and fails to lower any
incoming value that itself is another PHI of Vreg_1 regclass.
The failure occurs only when the MBB where the incoming PHI value
belongs is not visited/lowered yet.

To fix this problem, collect all Vreg_1 PHIs upfront and then
perform the lowering.

Differential Revision: https://reviews.llvm.org/D69182
2019-10-26 14:37:45 +05:30
Shu-Chun Weng 5e30780855 Correct size_t format specifier
Differential Revision: https://reviews.llvm.org/D69455
2019-10-26 10:38:06 +02:00
Craig Topper a6a37e820c [X86][GISel] Fix typo in comment. NFC 2019-10-26 00:27:53 -07:00
Sergey Dmitriev edb1a1de1b Reland "[Clang][Bundler] Error reporting improvements"
- Changed FileHandler read/write methods to return llvm::Error
- Using unified way of reporting errors
- Removed trailing '.' from the error messages

Differential Revision: https://reviews.llvm.org/D67031
2019-10-25 19:00:06 -07:00
Lang Hames 4141bb50a8 [ORC] Avoid SymbolStringPtr copies in DynamicLibrarySearchGenerator predicate.
Pass SymbolStringPtr by const-ref to avoid copies (which require atomic
ref-count operations).
2019-10-25 18:19:25 -07:00
Sergey Dmitriev 93b29d3882 Revert "[Clang][Bundler] Error reporting improvements"
This reverts commit dd501045cd.
2019-10-25 17:57:55 -07:00
John McCall baf91d02da [NFC] Add a tablegen node for the root of the AST node hierarchies.
This is useful for the property databases we want to add for abstract
serialization, since root classes can have interesting properties.
2019-10-25 16:39:21 -07:00
John McCall 27e2c8faec Add Record::getValueAsOptionalDef().
Using `?` as an optional marker is very useful in Clang's AST-node
emitters because otherwise we need a separate class just to encode
the presence or absence of a base node reference.
2019-10-25 16:39:21 -07:00
Sergey Dmitriev dd501045cd [Clang][Bundler] Error reporting improvements
- Changed FileHandler read/write methods to return llvm::Error
- Using unified way of reporting errors
- Removed trailing '.' from the error messages

Differential Revision: https://reviews.llvm.org/D67031
2019-10-25 16:29:57 -07:00
Sanjay Patel 4c47617627 [SDAG] fold extract_vector_elt with undef index
This makes the DAG behavior consistent with IR's extractelement after:
rGb32e4664a715

https://bugs.llvm.org/show_bug.cgi?id=42689

I've tried to maintain test intent for WebAssembly.
The AMDGPU test is trying to test for crashing or other bad behavior,
but I'm not sure if that's possible after this change.
2019-10-25 19:27:26 -04:00
Julian Lettner f3ad8ae7b7 [lit] Move sharding logic into separate function 2019-10-25 16:23:52 -07:00
Matt Morehouse a88591cff4 [libFuzzer] Enable extra counters for Fuchsia. 2019-10-25 16:12:59 -07:00
Adrian McCarthy 5a3c657f3e Fix after 738af7a624
Default implementation of a new virtual method wasn't returning a value.
2019-10-25 15:57:52 -07:00
Reid Kleckner 6c89392592 Use __builtin_strlen in constexpr StringRef ctor with MSVC
MSVC supports it. Fixes the major MSVC compile time regression
introduced in r369961. Now
clang/lib/StaticAnalyzer/Frontend/CheckerRegistry.cpp compiles in 18s
instead of 7+ minutes.

Fixes PR43369
2019-10-25 15:53:54 -07:00
Daniel Sanders feab0334f5 [globalisel] Restructure the GlobalISel documentation
There's a couple minor deletions amongst this but 99% of it is just moving
the documentation around to prepare the way for more meaningful changes.
2019-10-25 15:51:09 -07:00
Jonas Devlieghere 10b5cd8ed5 [LLDB] Fix inline variable only used in assertion. (NFC)
This prevents unused variable warning/error in -DNDEBUG builds. The
variable was introduced in 5934cd11ea.

Patch by: Shu-Chun Weng

Differential revision: https://reviews.llvm.org/D69451
2019-10-25 15:47:17 -07:00
Michael Spencer 8da20560ab [clang][DependencyScanning] 80-col. 2019-10-25 15:43:57 -07:00
Pavel Labath 7c603a41e2 lldb/minidump: Refactor memory region computation code
The goal of this refactor is to enable ProcessMinidump to take into
account the loaded modules and their sections when computing the
permissions of various ranges of memory, as discussed in D66638.

This patch moves some of the responsibility for computing the ranges
from MinidumpParser into ProcessMinidump. MinidumpParser still does the
parsing, but ProcessMinidump becomes responsible for answering the
actual queries about memory ranges. This will enable it (in a follow-up
patch) to augment the information obtained from the parser with data
obtained from actual object files.

The changes in the actual code are fairly straight-forward and just
involve moving code around. MinidumpParser::GetMemoryRegions is renamed
to BuildMemoryRegions to emphasize that it does no caching. The only new
thing is the additional bool flag returned from this function. This
indicates whether the returned regions describe all memory mapped into
the target process. Data obtained from /proc/maps and the MemoryInfoList
stream is considered to be exhaustive. Data obtained from Memory(64)List
is not. This will be used to determine whether we need to augment the
data or not.

This reshuffle means that it is no longer possible/easy to test some of
this code via unit tests, as constructing a ProcessMinidump instance is
hard. Instead, I update the unit tests to only test the parsing of the
actual data, and test the answering of queries through a lit test using
the "memory region" command. The patch also includes some tweaks to the
MemoryRegion class to make the unit tests easier to write.

Reviewers: amccarth, clayborg

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D69035
2019-10-25 22:33:32 +00:00
Joel E. Denny 27fdf8a29d [lit] Don't fail when printing test output with special chars
This addresses a UnicodeEncodeError when using Python 3.6.5 in Windows
10.

Reviewed By: rnk

Differential Revision: https://reviews.llvm.org/D69207
2019-10-25 18:13:13 -04:00
Pavel Labath 73a7a55c0e lldb/COFF: Create a separate "section" for the file header
In an attempt to ensure that every part of the module's memory image is
accounted for, D56537 created a special "container section" spanning the
entire image. While that seemed reasonable at the time (and it still
mostly does), it did create a problem of what to put as the "file size"
of the section, because the image is not continuous on disk, as we
generally assume (which is why I put zero there). Additionally, this
arrangement makes it unclear what kind of permissions should be assigned
to that section (which is what my next patch does).

To get around these, this patch partially reverts D56537, and goes back
to top-level sections. Instead, what I do is create a new "section" for
the object file header, which is also being loaded into memory, though
its not considered to be a section in the strictest sense. This makes it
possible to correctly assign file size section, and we can later assign
permissions to it as well.

Reviewers: amccarth, mstorsjo

Subscribers: lldb-commits

Differential Revision: https://reviews.llvm.org/D69100
2019-10-25 22:11:53 +00:00
Stanislav Mekhanoshin 4c0251da14 [AMDGPU] Enable SGPR copy folding
That used to fail in the last testcase function because after
%0:sreg_64.sub0 was folded into %3:sreg_32_xm0_xexec COPY, it
was further folded into S_STORE_DWORD_IMM. Its legal effective
subreg class is SReg_32 while instruction expects more restricted
SReg_32_XM0_EXEC. However, SIInstrInfo::isLegalRegOperand()
passed the legality check and it was caught in the verifier.

Borrowed code from the verifier to check for RC legality.

Differential Revision: https://reviews.llvm.org/D69445
2019-10-25 15:08:30 -07:00
Daniel Sanders 27887bc1e7 [globalisel] Fix typo in 'Add LLVMDev 2019 talks and links for the 2017 talks' 2019-10-25 15:01:14 -07:00
Daniel Sanders 7913126a08 [globalisel] Add LLVMDev 2019 talks and links for the 2017 talks 2019-10-25 14:53:58 -07:00
Yonghong Song a27c998c00 [BPF] fix a CO-RE issue with -mattr=+alu32
Ilya Leoshkevich (<iii@linux.ibm.com>) reported an issue that
with -mattr=+alu32 CO-RE has a segfault in BPF MISimplifyPatchable
pass.

The pattern will be transformed by MISimplifyPatchable
pass looks like below:
  r5 = ld_imm64 @"b:0:0$0:0"
  r2 = ldw r5, 0
  ... r2 ... // use r2
The pass will remove the intermediate 'ldw' instruction
and replacing all r2 with r5 likes below:
  r5 = ld_imm64 @"b:0:0$0:0"
  ... r5 ... // use r5
Later, the ld_imm64 insn will be replaced with
  r5 = <patched immediate>
for field relocation purpose.

With -mattr=+alu32, the input code may become
  r5 = ld_imm64 @"b:0:0$0:0"
  w2 = ldw32 r5, 0
  ... w2 ... // use w2
Replacing "w2" with "r5" is incorrect and will
trigger compiler internal errors.

To fix the problem, if the register class of ldw* dest
register is sub_32, we just replace the original ldw*
register with:
  w2 = w5
Directly replacing all uses of w2 with in-place
constructed w5 for the use operand seems not working in all cases.

The latest kernel will have -mattr=+alu32 on by default,
so added this flag to all CORE tests.
Tested with latest kernel bpf-next branch as well with this patch.

Differential Revision: https://reviews.llvm.org/D69438
2019-10-25 14:27:25 -07:00
Sanjay Patel e070cf8119 [x86] add tests for extractelement with undef index (PR42689); NFC 2019-10-25 17:22:37 -04:00
David Tenty 0e4d415311 Update version number in llvm python bindings
Summary:
The version number has come out of sync with what is in CMakeLists.txt,
causing loading the bindings to fail.

Reviewers: AustinWells, abhina.sree

Reviewed By: AustinWells

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69436
2019-10-25 17:14:11 -04:00
Jim Ingham 738af7a624 Add the ability to pass extra args to a Python breakpoint callback.
For example, it is pretty easy to write a breakpoint command that implements "stop when my caller is Foo", and
    it is pretty easy to write a breakpoint command that implements "stop when my caller is Bar". But there's no
    way to write a generic "stop when my caller is..." function, and then specify the caller when you add the
    command to a breakpoint.

    With this patch, you can pass this data in a SBStructuredData dictionary. That will get stored in
    the PythonCommandBaton for the breakpoint, and passed to the implementation function (if it has the right
    signature) when the breakpoint is hit. Then in lldb, you can say:

    (lldb) break com add -F caller_is -k caller_name -v Foo

    More generally this will allow us to write reusable Python breakpoint commands.

    Differential Revision: https://reviews.llvm.org/D68671
2019-10-25 14:05:07 -07:00
Jian Cai a6b0219fc4 Revert "[ARM] Uses "Sun Style" syntax for section switching"
This reverts commit 03de2f84fc.
2019-10-25 14:03:07 -07:00
Stanislav Mekhanoshin c7dcacf16a [AMDGPU] Fixed asan failure in SIFoldOperands
Both tryFoldOMod() and tryFoldClamp() remove original instruction,
so the check MI.modifiesRegister() may use a deleted MI.

Differential Revision: https://reviews.llvm.org/D69448
2019-10-25 13:59:56 -07:00
Matt Arsenault 1a276d1e8c GlobalISel: Implement widenScalar for G_INSERT_VECTOR_ELT 2019-10-25 13:55:07 -07:00
Guillaume Chatelet e8a0a0904b [Alignment][NFC] Convert AllocaInst to MaybeAlign
Summary:
This is patch is part of a series to introduce an Alignment type.
See this thread for context: http://lists.llvm.org/pipermail/llvm-dev/2019-July/133851.html
See this patch for the introduction of the type: https://reviews.llvm.org/D64790

Reviewers: courbet

Reviewed By: courbet

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69301
2019-10-25 22:41:34 +02:00
Jian Cai 03de2f84fc [ARM] Uses "Sun Style" syntax for section switching
Summary:
Support "Sun Style" syntax for section switching ("#alloc,#write" etc).
https://bugs.llvm.org/show_bug.cgi?id=43759

Reviewers: peter.smith, eli.friedman, kristof.beyls, t.p.northover

Reviewed By: peter.smith

Subscribers: MaskRay, llozano, manojgupta, nickdesaulniers, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69296
2019-10-25 13:27:35 -07:00
Matt Arsenault 171cf5302f AMDGPU/GlobalISel: Handle flat/global G_ATOMIC_CMPXCHG
Custom lower this to a target instruction with the merge operands. I
think it might be better to directly select this and emit a
REG_SEQUENCE, but this would be more work since it would require
splitting the tablegen patterns for these cases from the other
atomics.
2019-10-25 13:11:09 -07:00
Changpeng Fang 1ce552f3ef AMDGPU: Fix the broken dominator tree when creating waterfall loop for resource descriptor
Summary:
  In loadSRsrcFromVGPR, if MBB is the same as Succ, Remiander is not the immediate dominator of Succ.

Reviewer:
  arsenm

Differential Revision:
  https://reviews.llvm.org/D69358
2019-10-25 13:08:04 -07:00
Daniel Sanders a6e1de4afc [gicombiner] Add parse failure tests for defs/match 2019-10-25 12:56:49 -07:00
Amy Huang 64c1f6602a Revert "Add an instruction marker field to the ExtraInfo in MachineInstrs."
Reverting commit b85b4e5a6f due to some
buildbot failures/ out of memory errors.
2019-10-25 12:41:34 -07:00
Teresa Johnson cc0b9647b7 [LLD][ThinLTO] Handle GUID collision in import global processing
Summary:
If there are a GUID collision between two globals checking the
summarylist from the import index to make assumption can be dangerous.

Do not assume that a GlobalValue that has a GlobalVarSummary
actually is a GlobalVariable as it can be another GlobalValue with
the same GUID that the summary is connected to.

Patch by Joel Klinghed (the_jk@opera.com)

Reviewers: evgeny777, tejohnson

Reviewed By: tejohnson

Subscribers: tejohnson, dblaikie, MaskRay, mehdi_amini, inglorion, hiraditya, steven_wu, dexonsmith, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D67322
2019-10-25 12:36:01 -07:00
Guillaume Chatelet a4783ef58d [Alignment][NFC] getMemoryOpCost uses MaybeAlign
Summary:
This is patch is part of a series to introduce an Alignment type.
See this thread for context: http://lists.llvm.org/pipermail/llvm-dev/2019-July/133851.html
See this patch for the introduction of the type: https://reviews.llvm.org/D64790

Reviewers: courbet

Subscribers: nemanjai, hiraditya, kbarton, MaskRay, jsji, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69307
2019-10-25 21:26:59 +02:00
Sanjay Patel 3c7c371793 [CVP] add test for poison propagation bug (PR43802); NFC 2019-10-25 15:01:57 -04:00
Saleem Abdulrasool 2724d9e129 build: remove `LLVM_CXX_STD` extension point
This extension point is not needed. Provide the equivalent option
through `CMAKE_CXX_STANDARD` which mirrors the previous extension point. Rely on
CMake to provide the check for the compiler instead.
2019-10-25 11:51:47 -07:00
Alexander Shaposhnikov 074af2daf5 [llvm-objcopy][MachO] Add support for min os version load commands
Add support for min os version load commands.

Test plan: make check-all

Differential revision: https://reviews.llvm.org/D69419
2019-10-25 11:42:29 -07:00
Stanislav Mekhanoshin d4303b3861 [AMDGPU] Fold AGPR reg_sequence initializers
Differential Revision: https://reviews.llvm.org/D69413
2019-10-25 11:39:02 -07:00