platform/upstream/llvm.git
4 years ago[SelectionDAG] Reduce assumptions made about levels. NFC
Amaury Séchet [Mon, 2 Dec 2019 15:30:51 +0000 (16:30 +0100)]
[SelectionDAG] Reduce assumptions made about levels. NFC

4 years agoAdd AIX assembler support
stevewan [Mon, 2 Dec 2019 16:28:31 +0000 (11:28 -0500)]
Add AIX assembler support

Summary:
A skeleton of AIX toolchain and system linker support has been introduced in D68340, and this is a follow on patch to it.
This patch adds support to system assembler invocation to the AIX toolchain.

Reviewers: daltenty, hubert.reinterpretcast, jasonliu, Xiangling_L, dlj

Reviewed By: daltenty, hubert.reinterpretcast

Subscribers: wuzish, nemanjai, kbarton, jfb, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D69620

4 years ago[ARM,MVE] Add intrinsics to deal with predicates.
Simon Tatham [Mon, 2 Dec 2019 16:17:59 +0000 (16:17 +0000)]
[ARM,MVE] Add intrinsics to deal with predicates.

Summary:
This commit adds the `vpselq` intrinsics which take an MVE predicate
word and select lanes from two vectors; the `vctp` intrinsics which
create a tail predicate word suitable for processing the first m
elements of a vector (e.g. in the last iteration of a loop); and
`vpnot`, which simply complements a predicate word and is just
syntactic sugar for the `~` operator.

The `vctp` ACLE intrinsics are lowered to the IR intrinsics we've
already added (and which D70592 just reorganized). I've filled in the
missing isel rule for VCTP64, and added another set of rules to
generate the predicated forms.

I needed one small tweak in MveEmitter to allow the `unpromoted` type
modifier to apply to predicates as well as integers, so that `vpnot`
doesn't pointlessly convert its input integer to an `<n x i1>` before
complementing it.

Reviewers: ostannard, MarkMurrayARM, dmgreen

Reviewed By: dmgreen

Subscribers: kristof.beyls, hiraditya, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D70485

4 years ago[ARM,MVE] Rename and clean up VCTP IR intrinsics.
Simon Tatham [Mon, 2 Dec 2019 16:18:24 +0000 (16:18 +0000)]
[ARM,MVE] Rename and clean up VCTP IR intrinsics.

Summary:
D65884 added a set of Arm IR intrinsics for the MVE VCTP instruction,
to use in tail predication. But the 64-bit one doesn't work properly:
its predicate type is `<2 x i1>` / `v2i1`, which isn't a legal MVE
type (due to not having a full set of instructions that manipulate it
usefully). The test of `vctp64` in `basic-tail-pred.ll` goes through
`opt` fine, as the test expects, but if you then feed it to `llc` it
causes a type legality failure at isel time.

The usual workaround we've been using in the rest of the MVE
intrinsics family is to bodge `v2i1` into `v4i1`. So I've adjusted the
`vctp64` IR intrinsic to do that, and completely removed the code (and
test) that uses that intrinsic for 64-bit tail predication. That will
allow me to add isel rules (upcoming in D70485) that actually generate
the VCTP64 instruction.

Also renamed all four of these IR intrinsics so that they have `mve`
in the name, since its absence was confusing.

Reviewers: ostannard, MarkMurrayARM, dmgreen

Reviewed By: MarkMurrayARM

Subscribers: samparker, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70592

4 years ago[ARM,MVE] Add an InstCombine rule permitting VPNOT.
Simon Tatham [Mon, 2 Dec 2019 16:18:34 +0000 (16:18 +0000)]
[ARM,MVE] Add an InstCombine rule permitting VPNOT.

Summary:
If a user writing C code using the ACLE MVE intrinsics generates a
predicate and then complements it, then the resulting IR will use the
`pred_v2i` IR intrinsic to turn some `<n x i1>` vector into a 16-bit
integer; complement that integer; and convert back. This will generate
machine code that moves the predicate out of the `P0` register,
complements it in an integer GPR, and moves it back in again.

This InstCombine rule replaces `i2v(~v2i(x))` with a direct complement
of the original predicate vector, which we can already instruction-
select as the VPNOT instruction which complements P0 in place.

Reviewers: ostannard, MarkMurrayARM, dmgreen

Reviewed By: dmgreen

Subscribers: kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70484

4 years agoAdd contributing info to CONTRIBUTING.md and README.md
Florian Hahn [Mon, 2 Dec 2019 15:46:47 +0000 (15:46 +0000)]
Add contributing info to CONTRIBUTING.md and README.md

As discussed on llvm-dev [1], this patch adds a brief CONTRIBUTING.md to
the top-level of the repo, with a pointer to Contributing.html. This
should make it easier to discover the contributing information and also
be highlighted in the Github UI.

It also updates README.md to link to Contributing.html.

[1] http://lists.llvm.org/pipermail/llvm-dev/2019-November/137141.html

Reviewers: tonic, rnk, jhenderson, meikeb

Reviewed By: rnk, jhenderson, meikeb

Differential Revision: https://reviews.llvm.org/D70743

4 years ago[PatternMatch] Add support for matching intrinsics with 5 operands.
Florian Hahn [Mon, 2 Dec 2019 15:29:53 +0000 (15:29 +0000)]
[PatternMatch] Add support for matching intrinsics with 5 operands.

Summary: Also adds a test to the pattern matching unit tests.

Reviewers: spatel, craig.topper, RKSimon, majnemer, lebedev.ri

Reviewed By: spatel

Subscribers: merge_guards_bot, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70892

4 years ago[Attributor] Copy or port test cases related to Attributor to` Attributor` test folder
Hideto Ueno [Mon, 2 Dec 2019 13:40:09 +0000 (13:40 +0000)]
[Attributor] Copy or port test cases related to Attributor to` Attributor` test folder

Summary:
This patch moves the test cases related to Attributor to `Transforms/Attributor` folder.
We have used `Transforms/FunctionAttrs` as the primary folder for Attributor test but we need to change testing way now.

For the test cases which I think functionattrs doesn't infer anything something like (willreturn, nosync, value-simplify, h2s ..etc), I moved them with the command `git mv`.

For the test cases in which functoinattrs and attributor are tested, I copied the test to the folder and remove the check only used by functoinattrs.

Reviewers: jdoerfert, sstefan1

Reviewed By: jdoerfert

Subscribers: jfb, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70843

4 years agoAutogenerate test/Analysis/ValueTracking/non-negative-phi-bits.ll test
Roman Lebedev [Mon, 2 Dec 2019 15:28:17 +0000 (18:28 +0300)]
Autogenerate test/Analysis/ValueTracking/non-negative-phi-bits.ll test

Forgot to stage this change into 0f22e783a038b6983f0fe161eef6cf2add3a4156 commit.

4 years ago[clangd] Try harder to find a plausible `clang` as argv0, particularly on Mac.
Sam McCall [Fri, 29 Nov 2019 18:37:48 +0000 (19:37 +0100)]
[clangd] Try harder to find a plausible `clang` as argv0, particularly on Mac.

Summary:
Fixes https://github.com/clangd/clangd/issues/211
Fixes https://github.com/clangd/clangd/issues/178

No tests - this is hard to test, and basically impossible to verify what we want
(this produces compile commands that work on a real mac with recent toolchain)

(Need someone on mac to verify it actually fixes these!)

Reviewers: kbobyrev, ilya-biryukov

Subscribers: MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70863

4 years ago[llvm-exegesis] Fix 44b9942898c7.
Clement Courbet [Mon, 2 Dec 2019 13:58:41 +0000 (14:58 +0100)]
[llvm-exegesis] Fix 44b9942898c7.

Summary:
Add missing stack release instructions in
loadImplicitRegAndFinalize.

Reviewers: pengfei, gchatelet

Subscribers: tschuett, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70903

4 years ago[InstCombine] Revert rL341831: relax one-use check in foldICmpAddConstant() (PR44100)
Roman Lebedev [Mon, 2 Dec 2019 14:34:55 +0000 (17:34 +0300)]
[InstCombine] Revert rL341831: relax one-use check in foldICmpAddConstant() (PR44100)

rL341831 moved one-use check higher up, restricting a few folds
that produced a single instruction from two instructions to the case
where the inner instruction would go away.

Original commit message:
> InstCombine: move hasOneUse check to the top of foldICmpAddConstant
>
> There were two combines not covered by the check before now,
> neither of which actually differed from normal in the benefit analysis.
>
> The most recent seems to be because it was just added at the top of the
> function (naturally). The older is from way back in 2008 (r46687)
> when we just didn't put those checks in so routinely, and has been
> diligently maintained since.

From the commit message alone, there doesn't seem to be a
deeper motivation, deeper problem that was trying to solve,
other than 'fixing the wrong one-use check'.

As i have briefly discusses in IRC with Tim, the original motivation
can no longer be recovered, too much time has passed.

However i believe that the original fold was doing the right thing,
we should be performing such a transformation even if the inner `add`
will not go away - that will still unchain the comparison from `add`,
it will no longer need to wait for `add` to compute.

Doing so doesn't seem to break any particular idioms,
as least as far as i can see.

References https://bugs.llvm.org/show_bug.cgi?id=44100

4 years ago[PowerPC] Fix crash in peephole optimization
Nemanja Ivanovic [Mon, 2 Dec 2019 14:32:59 +0000 (08:32 -0600)]
[PowerPC] Fix crash in peephole optimization

When converting reg+reg shifts to reg+imm rotates, we neglect to consider the
CodeGenOnly versions of the 32-bit shift mnemonics. This means we produce a
rotate with missing operands which causes a crash.

Committing this fix without review since it is non-controversial that the list
of mnemonics to consider should include the 64-bit aliases for the exact
mnemonics.

Fixes PR44183.

4 years ago[ARM][AArch64] Complex addition Neon intrinsics for Armv8.3-A
Victor Campos [Mon, 2 Dec 2019 12:13:04 +0000 (12:13 +0000)]
[ARM][AArch64] Complex addition Neon intrinsics for Armv8.3-A

Summary:
Add support for vcadd_* family of intrinsics. This set of intrinsics is
available in Armv8.3-A.

The fp16 versions require the FP16 extension, which has been available
(opt-in) since Armv8.2-A.

Reviewers: t.p.northover

Reviewed By: t.p.northover

Subscribers: t.p.northover, kristof.beyls, hiraditya, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D70862

4 years ago[InstCombine] fold copysign with constant sign argument to (fneg+)fabs
Sanjay Patel [Mon, 2 Dec 2019 14:21:59 +0000 (09:21 -0500)]
[InstCombine] fold copysign with constant sign argument to (fneg+)fabs

If the sign of the sign argument is known (this could be extended to use ValueTracking),
then we can use fneg+fabs to clear/set the sign bit of the magnitude argument.
http://llvm.org/docs/LangRef.html#llvm-copysign-intrinsic

This transform is already done in DAGCombiner, but we can do it sooner in IR as
suggested in PR44153:
https://bugs.llvm.org/show_bug.cgi?id=44153

We have effectively no analysis for copysign in IR, so we are taking the unusual step
of increasing the number of IR instructions for the negative constant case.

Differential Revision: https://reviews.llvm.org/D70792

4 years ago[OpenCL] Fix address space for implicit conversion (PR43145)
Sven van Haastregt [Mon, 2 Dec 2019 14:20:15 +0000 (14:20 +0000)]
[OpenCL] Fix address space for implicit conversion (PR43145)

Clang was creating a DerivedToBase ImplicitCastExpr that was also
casting between address spaces as part of the second step in the
standard conversion sequence.  Defer the address space conversion to
the third step in the sequence instead, such that we get a separate
ImplicitCastExpr for the address space conversion.

Differential Revision: https://reviews.llvm.org/D70605

4 years ago[lldb][NFC] Don't calculate member indices in DWARFASTParserClang::ParseChildMembers
Raphael Isemann [Mon, 2 Dec 2019 13:34:51 +0000 (14:34 +0100)]
[lldb][NFC] Don't calculate member indices in DWARFASTParserClang::ParseChildMembers

We keep counting members and then don't do anything with the computed result.

4 years ago[lldb][NFC] Use raw_ostream instead of Stream in Baton::GetDescription
Raphael Isemann [Sat, 30 Nov 2019 14:30:08 +0000 (15:30 +0100)]
[lldb][NFC] Use raw_ostream instead of Stream in Baton::GetDescription

Removing raw_ostream here is getting us closer to removing LLDB's Stream
class.

4 years ago[X86] Add initialization of FPCW in llvm-exegesis
Wang, Pengfei [Mon, 2 Dec 2019 11:39:16 +0000 (19:39 +0800)]
[X86] Add initialization of FPCW in llvm-exegesis

Summary: This is a following up to D70874. It adds the initialization of FPCW in llvm-exegesis.

Reviewers: craig.topper, RKSimon, courbet, gchatelet

Subscribers: tschuett, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70891

4 years ago[llvm-readobj/llvm-readelf] - Simplify the code that dumps versions.
Georgii Rymar [Thu, 28 Nov 2019 11:12:09 +0000 (14:12 +0300)]
[llvm-readobj/llvm-readelf] - Simplify the code that dumps versions.

After changes introduced in D70495 and D70826 its now possible
to significantly simplify the code we have.

This also fixes an issue: previous code assumed that version strings
should always be read from the dynamic string table. While it is
normally true, the string table should be taken from the corresponding
sh_link field.

Differential revision: https://reviews.llvm.org/D70855

4 years agoAMDGPU: Fixed indeterminate map iteration in SIPeepholeSDWA
Tim Renouf [Wed, 27 Nov 2019 15:16:59 +0000 (15:16 +0000)]
AMDGPU: Fixed indeterminate map iteration in SIPeepholeSDWA

Differential Revision: https://reviews.llvm.org/D70783

Change-Id: Ic26f915a4acb4c00ecefa9d09d7c24cec370ed06

4 years ago[lldb][NFC] Make Stream's IndentLevel an unsigned integers.
Raphael Isemann [Mon, 2 Dec 2019 11:44:55 +0000 (12:44 +0100)]
[lldb][NFC] Make Stream's IndentLevel an unsigned integers.

We expect it to be always positive values and LLVM/Clang's IndentLevel
values are already unsigned integers, so we should do the same.

4 years ago[ARM][MVE][Intrinsics] Add VMINQ/VMAXQ/VMINNMQ/VMAXNMQ intrinsics.
Mark Murray [Thu, 28 Nov 2019 16:38:01 +0000 (16:38 +0000)]
[ARM][MVE][Intrinsics] Add VMINQ/VMAXQ/VMINNMQ/VMAXNMQ intrinsics.

Summary: Add VMINQ/VMAXQ/VMINNMQ/VMAXNMQ intrinsics and their predicated versions. Add unit tests.

Subscribers: kristof.beyls, hiraditya, dmgreen, cfe-commits, llvm-commits

Tags: #clang, #llvm

Differential Revision: https://reviews.llvm.org/D70829

4 years ago[update_cc_test_checks.py] Handle extern "C" and namespaces
Alex Richardson [Mon, 2 Dec 2019 10:53:57 +0000 (10:53 +0000)]
[update_cc_test_checks.py] Handle extern "C" and namespaces

Summary:
My change to use the clang AST JSON dump did not handle functions declared
inside scopes other than the root TranslationUnitDecl. After this change
update_cc_test_checks.py also works for C++ test cases that use extern "C"
and namespaces.

Reviewers: MaskRay

Reviewed By: MaskRay

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70389

4 years ago[UpdateTestChecks] Share the code to parse RUN: lines between all scripts
Alex Richardson [Mon, 2 Dec 2019 10:50:23 +0000 (10:50 +0000)]
[UpdateTestChecks] Share the code to parse RUN: lines between all scripts

Summary:
This commit also introduces a common.debug() function to avoid many
`if args.verbose:` statements. Depends on D70428.

Reviewers: xbolva00, MaskRay, jdoerfert

Reviewed By: MaskRay

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70432

4 years ago[lldb][NFC] Add 'breakpoint command list' test
Raphael Isemann [Mon, 2 Dec 2019 10:08:10 +0000 (11:08 +0100)]
[lldb][NFC] Add 'breakpoint command list' test

The command has zero test coverage and I'll have to touch the
code formatting the output commands, so let's start by adding a
test for it.

4 years ago[ARM] Remove VHADD patterns
David Green [Mon, 2 Dec 2019 10:29:01 +0000 (10:29 +0000)]
[ARM] Remove VHADD patterns

These instructions do not work quite like I expected them to. They
perform the addition and then shift in a higher precision integer, so do
not match up with the patterns that we added.

For example with s8s, adding 100 and 100 should wrap leaving the shift
to work on a negative number. VHADD will instead do the arithmetic in
higher precision, giving 100 overall. The vhadd gives a "better" result,
but not one that matches up with the input.

I am just removing the patterns here. We might be able to re-add them in
the future by checking for wrap flags or changing bitwidths. But for the
moment just remove them to remove the problem cases.

4 years ago[X86] Add initialization of MXCSR in llvm-exegesis
Wang, Pengfei [Sun, 1 Dec 2019 05:35:53 +0000 (13:35 +0800)]
[X86] Add initialization of MXCSR in llvm-exegesis

Summary: This patch is used to initialize the new added register MXCSR.

Reviewers: craig.topper, RKSimon

Subscribers: tschuett, courbet, llvm-commits, LiuChen3

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70874

4 years ago[InstCombine] Fix big-endian miscompile of (bitcast (zext/trunc (bitcast)))
Bjorn Pettersson [Thu, 28 Nov 2019 22:18:28 +0000 (23:18 +0100)]
[InstCombine] Fix big-endian miscompile of (bitcast (zext/trunc (bitcast)))

Summary:
optimizeVectorResize is rewriting patterns like:
  %1 = bitcast vector %src to integer
  %2 = trunc/zext %1
  %dst = bitcast %2 to vector

Since bitcasting between integer an vector types gives
different integer values depending on endianness, we need
to take endianness into account. As it happens the old
implementation only produced the correct result for little
endian targets.

Fixes: https://bugs.llvm.org/show_bug.cgi?id=44178

Reviewers: spatel, lattner, lebedev.ri

Reviewed By: spatel, lebedev.ri

Subscribers: lebedev.ri, hiraditya, uabelho, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70844

4 years ago[llvm-readelf/llvm-readobj] - Check the version of SHT_GNU_verneed section entries.
Georgii Rymar [Fri, 29 Nov 2019 08:57:13 +0000 (11:57 +0300)]
[llvm-readelf/llvm-readobj] - Check the version of SHT_GNU_verneed section entries.

It is a follow-up for D70826 and it is similar to D70810.

SHT_GNU_verneed contains the following fields:
`vn_version`: Version of structure. This value is currently set to 1, and will be reset
if the versioning implementation is incompatibly altered.
(https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/symversion.html)

We should check it for correctness.

Differential revision: https://reviews.llvm.org/D70842

4 years ago[ORC] Add a runAsMain utility function to ExecutionUtils.
Lang Hames [Mon, 2 Dec 2019 09:45:49 +0000 (01:45 -0800)]
[ORC] Add a runAsMain utility function to ExecutionUtils.

The runAsMain function takes a pointer to a function with a standard C main
signature, int(*)(int, char*[]), and invokes it using the given arguments and
program name. The arguments are copied into writable temporary storage as
required by the C and C++ specifications, so runAsMain safe to use when calling
main functions that modify their arguments in-place.

This patch also uses the new runAsMain function to replace hand-rolled versions
in lli, llvm-jitlink, and the SpeculativeJIT example.

4 years ago[ExecutionEngine] Add a jitTargetAddressToFunction utility function.
Lang Hames [Mon, 2 Dec 2019 09:40:54 +0000 (01:40 -0800)]
[ExecutionEngine] Add a jitTargetAddressToFunction utility function.

jitTargetAddressToFunction takes a JITTargetAddress and returns a pointer of
the given function pointer type suitable for calling to invoke the function
at the target address.

jitTargetAddressToFunction currently behaves the same as
jitTargetAddressToPointer, but in the near future will be updated to perform
pointer signing on architectures that require it (e.g. arm64e). For this
reason it should always be preferred when generating callable pointers for
JIT'd functions.

4 years ago[Orc] Add setters for target options and features to JITTargetMachineBuilder.
Lang Hames [Sun, 1 Dec 2019 23:51:37 +0000 (15:51 -0800)]
[Orc] Add setters for target options and features to JITTargetMachineBuilder.

Also remove redundant feature initialization steps from the detectHost method.

4 years ago[clangd] Fix a regression issue in local rename.
Haojian Wu [Fri, 29 Nov 2019 13:58:44 +0000 (14:58 +0100)]
[clangd] Fix a regression issue in local rename.

Summary:
The regression is that we can't rename symbols in annonymous
namespaces.

Reviewers: ilya-biryukov

Subscribers: MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70853

4 years ago[llvm-readobj/llvm-readelf] - Reimplement dumping of the SHT_GNU_verneed section.
Georgii Rymar [Thu, 28 Nov 2019 09:19:50 +0000 (12:19 +0300)]
[llvm-readobj/llvm-readelf] - Reimplement dumping of the SHT_GNU_verneed section.

This is similar to D70495, but for SHT_GNU_verneed section.
It solves the same problems: different implementations, lack of error reporting
and no test coverage.

DIfferential revision: https://reviews.llvm.org/D70826

4 years ago[NFC] Precommit test showing SROA loses `!tbaa.struct` metadata
Anton Afanasyev [Mon, 2 Dec 2019 07:07:55 +0000 (10:07 +0300)]
[NFC] Precommit test showing SROA loses `!tbaa.struct` metadata

This issue impacts llvm.org/pr42022

4 years ago[Clang-Tidy] Quick fix for bug in bugprone-macro-parentheses 43804
Adam Balogh [Fri, 29 Nov 2019 12:30:26 +0000 (13:30 +0100)]
[Clang-Tidy] Quick fix for bug in bugprone-macro-parentheses 43804

Applying parentheses for statement leads to compilation error. Bug
[[ 43804 | https://bugs.llvm.org/show_bug.cgi?id=43804 ]] is a
compilation error suggested by a wrong fix of this checker. This
patch is a quick fix for this issue.

Differential Revision: https://reviews.llvm.org/D70850

4 years agoFix broken comment phrasing and indentation
Matt Arsenault [Thu, 21 Nov 2019 05:54:44 +0000 (11:24 +0530)]
Fix broken comment phrasing and indentation

4 years agoAMDGPU/GlobalISel: Add AGPR bank and RegBankSelect mfma intrinsics
Austin Kerbow [Wed, 27 Nov 2019 23:07:56 +0000 (15:07 -0800)]
AMDGPU/GlobalISel: Add AGPR bank and RegBankSelect mfma intrinsics

Differential Revision: https://reviews.llvm.org/D70871

4 years ago[SCEV] Make SCEV verification available from command line with new PM
Daniil Suchkov [Tue, 19 Nov 2019 07:16:39 +0000 (14:16 +0700)]
[SCEV] Make SCEV verification available from command line with new PM

New pass manager doesn't use verifyAnalysis, so currently there is no
way to call SCEV verification from command line when new PM is used.
This patch adds a pass that allows you to do that.

Reviewers: reames, fhahn, sanjoy.google, nikic

Reviewed By: fhahn

Subscribers: hiraditya, javed.absar, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70423

4 years agoRevert "[Examples] Add IRTransformations directory to examples."
Florian Hahn [Sun, 1 Dec 2019 22:19:05 +0000 (22:19 +0000)]
Revert "[Examples] Add IRTransformations directory to examples."

This breaks LLVMExports.cmake in some build configurations.

PR44197

This reverts commits ceb72d07b004af9c428c4a3c73a98ea97d49a713
                     7d0b1d77b3d4d47df477519fd1bf099b3df6f899.

4 years agoRevert "[clang][modules] Add support for merging lifetime-extended temporaries"
Tyker [Sun, 1 Dec 2019 21:38:31 +0000 (22:38 +0100)]
Revert "[clang][modules] Add support for merging lifetime-extended temporaries"

This reverts commit a3cbe1a202df6ec8e23bd55e14db254e4bc33021.

4 years ago[clang][modules] Add support for merging lifetime-extended temporaries
Tyker [Sat, 30 Nov 2019 15:42:33 +0000 (16:42 +0100)]
[clang][modules] Add support for merging lifetime-extended temporaries

Summary: Add support for merging lifetime-extended temporaries

Reviewers: rsmith

Reviewed By: rsmith

Subscribers: xbolva00, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70190

4 years ago[X86][InstCombine] Move non-X86 specific instcombine test from test/CodeGen/X86/...
Craig Topper [Sun, 1 Dec 2019 05:53:28 +0000 (21:53 -0800)]
[X86][InstCombine] Move non-X86 specific instcombine test from test/CodeGen/X86/ to test/Transforms/InstCombine/

4 years ago[X86][InstCombine] Move instcombine test from test/CodeGen/X86 to test/Transforms...
Craig Topper [Sun, 1 Dec 2019 05:47:41 +0000 (21:47 -0800)]
[X86][InstCombine] Move instcombine test from test/CodeGen/X86 to test/Transforms/InstCombine/ and replace grep with FileCheck

4 years ago[libunwind] Emit dependent libraries only when detected by CMake
Michał Górny [Sat, 30 Nov 2019 14:13:56 +0000 (15:13 +0100)]
[libunwind] Emit dependent libraries only when detected by CMake

996e62eef75 added Linux-specific dependent libraries to libunwind
sources.  As a result, building libunwind with modern LLD on *BSD
started failing due to trying to link libdl.  Instead, add those
libraries only if they were detected by CMake.

While technically we could create a long list of systems that need -ldl
and -lpthread, maintaining a duplicate list makes little sense when
CMake needs to detect it for non-LLD systems anyway.  Remove existing
system exceptions since they should be covered by the CMake check
anyway.

Remove -D_LIBUNWIND_HAS_COMMENT_LIB_PRAGMA since it is no longer
explicitly needed, if we make the library-specific defines dependent
on presence of this pragma support.

Differential Revision: https://reviews.llvm.org/D70868

4 years agoremove UB from test by making GV alignment explicit
Nuno Lopes [Sun, 1 Dec 2019 15:16:31 +0000 (15:16 +0000)]
remove UB from test by making GV alignment explicit

4 years agoRevert "[clang][modules] Add support for merging lifetime-extended temporaries"
Tyker [Sun, 1 Dec 2019 10:58:14 +0000 (11:58 +0100)]
Revert "[clang][modules] Add support for merging lifetime-extended temporaries"

This reverts commit 85c74384778909789389b9012a75cfcca7964a28.

4 years ago[clang][modules] Add support for merging lifetime-extended temporaries
Tyker [Sat, 30 Nov 2019 15:42:33 +0000 (16:42 +0100)]
[clang][modules] Add support for merging lifetime-extended temporaries

Summary: Add support for merging lifetime-extended temporaries

Reviewers: rsmith

Reviewed By: rsmith

Subscribers: xbolva00, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70190

4 years ago[Format] Add format check for coroutine keywords with negative numbers
Brian Gesiak [Sat, 30 Nov 2019 20:36:35 +0000 (15:36 -0500)]
[Format] Add format check for coroutine keywords with negative numbers

Summary:
As a followup to D69144, this diff fixes the coroutine keyword spacing
for co_yield / co_returning negative numbers.

Reviewers: modocache, sammccall, Quuxplusone

Reviewed By: modocache

Subscribers: cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D69180

Patch by Jonathan Thomas (jonathoma)!

4 years ago[X86] Add floating point execution domain to comi/ucomi/cvtss2si/cvtsd2si/cvttss2si...
Craig Topper [Sat, 30 Nov 2019 19:12:07 +0000 (11:12 -0800)]
[X86] Add floating point execution domain to comi/ucomi/cvtss2si/cvtsd2si/cvttss2si/cvttsd2si/cvtsi2ss/cvtsi2sd instructions.

4 years ago[InstCombine] Expand usub_sat patterns to handle constants
David Green [Sat, 30 Nov 2019 16:39:29 +0000 (16:39 +0000)]
[InstCombine] Expand usub_sat patterns to handle constants

The constants come through as add %x, -C, not a sub as would be
expected. They need some extra matchers to canonicalise them towards
usub_sat.

Differential Revision: https://reviews.llvm.org/D69514

4 years ago[InstCombine] Adjust usub_sat fold one use checks
David Green [Sat, 30 Nov 2019 14:20:55 +0000 (14:20 +0000)]
[InstCombine] Adjust usub_sat fold one use checks

This adjusts the one use checks in the the usub_sat fold code to not
increase instruction count, but otherwise do the fold. Reviewed as a
part of D69514.

4 years ago[InstCombine] More usub_sat tests. NFC.
David Green [Sat, 30 Nov 2019 14:14:55 +0000 (14:14 +0000)]
[InstCombine] More usub_sat tests. NFC.

4 years agoRevert "[clang][modules] Add support for merging lifetime-extended temporaries"
Tyker [Sat, 30 Nov 2019 16:52:26 +0000 (17:52 +0100)]
Revert "[clang][modules] Add support for merging lifetime-extended temporaries"

This reverts commit 3c7f6b439699a9cbbc0ac8d288cc70aff357446b.

4 years ago[clang][modules] Add support for merging lifetime-extended temporaries
Tyker [Sat, 30 Nov 2019 15:42:33 +0000 (16:42 +0100)]
[clang][modules] Add support for merging lifetime-extended temporaries

Summary: Add support for merging lifetime-extended temporaries

Reviewers: rsmith

Reviewed By: rsmith

Subscribers: xbolva00, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70190

4 years agoRevert 651f07908a1 "[AArch64] Don't combine callee-save and local stack adjustment...
Hans Wennborg [Sat, 30 Nov 2019 13:20:11 +0000 (14:20 +0100)]
Revert 651f07908a1 "[AArch64] Don't combine callee-save and local stack adjustment when optimizing for size"

This caused asserts (and perhaps also miscompiles) while building for Windows
on AArch64. See the discussion on D68530 for details and reproducer.

Reverting until this can be investigated and fixed.

> For arm64, D18619 introduced the ability to combine bumping the stack pointer
> upfront in case it needs to be bumped for both the callee-save area as well as
> the local stack area.
>
> That diff already remarks that "This change can cause an increase in
> instructions", but argues that even when that happens, it should be still be a
> performance benefit because the number of micro-ops is reduced.
>
> We have observed that this code-size increase can be significant in practice.
> This diff disables combining stack bumping for methods that are marked as
> optimize-for-size.
>
> Example of a prologue with the behavior before this diff (combining stack bumping when possible):
>   sub        sp, sp, #0x40
>   stp        d9, d8, [sp, #0x10]
>   stp        x20, x19, [sp, #0x20]
>   stp        x29, x30, [sp, #0x30]
>   add        x29, sp, #0x30
>   [... compute x8 somehow ...]
>   stp        x0, x8, [sp]
>
> And after this  diff, if the method is marked as optimize-for-size:
>   stp        d9, d8, [sp, #-0x30]!
>   stp        x20, x19, [sp, #0x10]
>   stp        x29, x30, [sp, #0x20]
>   add        x29, sp, #0x20
>   [... compute x8 somehow ...]
>   stp        x0, x8, [sp, #-0x10]!
>
> Note that without combining the stack bump there are two auto-decrements,
> nicely folded into the stp instructions, whereas otherwise there is a single
> sub sp, ... instruction, but not folded.
>
> Patch by Nikolai Tillmann!
>
> Differential Revision: https://reviews.llvm.org/D68530

4 years agoUpdated the OCaml/bitwriter.ml test for OCaml 4.06+
Dmitri Gribenko [Sat, 30 Nov 2019 12:31:16 +0000 (13:31 +0100)]
Updated the OCaml/bitwriter.ml test for OCaml 4.06+

Since OCaml 4.02 (released in 2014), strings and bytes are different
types, but up until OCaml 4.06, the compiler defaulted to a
compatibility mode "unsafe-string". OCaml 4.06 flips the default to
"safe-string", breaking the test.

This change should be compatible with OCaml 4.02+, but is only truly
necessary for OCaml 4.06+.

For more information, see:

https://caml.inria.fr/pub/docs/manual-ocaml/libref/String.html
https://ocaml.org/releases/4.02.html

4 years agoFix a typo.
Hans Wennborg [Sat, 30 Nov 2019 12:23:49 +0000 (13:23 +0100)]
Fix a typo.

4 years agoRevert "[clangd] Rethink how SelectionTree deals with macros and #includes."
Sam McCall [Fri, 29 Nov 2019 18:59:02 +0000 (19:59 +0100)]
Revert "[clangd] Rethink how SelectionTree deals with macros and #includes."

This reverts commit 19daa21f841ad45290c923689ee3d25198651a4c.

It causes a bunch of failures on a bot that I've been unable to
reproduce so far:
http://45.33.8.238/mac/3308/step_7.txt

4 years ago[PowerPC][AIX] Add support for lowering int/float/double formal arguments.
Sean Fertile [Fri, 29 Nov 2019 17:44:56 +0000 (12:44 -0500)]
[PowerPC][AIX] Add support for lowering int/float/double formal arguments.

This patch adds LowerFormalArguments_AIX, support is added for lowering
int, float, and double formal arguments into general purpose and
floating point registers only.

The aix calling convention testcase have been redone to test for caller
and callee functionality in the same lit test.

Patch by Zarko Todorovski!

Differential Revision: https://reviews.llvm.org/D69578

4 years agoRevert "[ARM] Allocatable Global Register Variables for ARM"
Carey Williams [Fri, 29 Nov 2019 17:01:05 +0000 (17:01 +0000)]
Revert "[ARM] Allocatable Global Register Variables for ARM"

This reverts commit 2d739f98d8a53e38bf9faa88cdb6b0c2a363fb77.

4 years agoRevert "[NFC] Fix test reserve_global_reg.ll after 2d739f9"
Carey Williams [Fri, 29 Nov 2019 17:00:55 +0000 (17:00 +0000)]
Revert "[NFC] Fix test reserve_global_reg.ll after 2d739f9"

This reverts commit aea7578fade2563cb5ea60548914667b515c457a.

4 years agoOn Windows, fix fuse-ld.c test when lld is provided explictly in -DCMAKE_LINKER
Alexandre Ganea [Fri, 29 Nov 2019 16:28:49 +0000 (11:28 -0500)]
On Windows, fix fuse-ld.c test when lld is provided explictly in -DCMAKE_LINKER

4 years ago[CIndex] Fix annotate-deep-statements test when using a Debug build
Alexandre Ganea [Fri, 29 Nov 2019 15:52:13 +0000 (10:52 -0500)]
[CIndex] Fix annotate-deep-statements test when using a Debug build

Differential Revision: https://reviews.llvm.org/D70149

4 years ago[clangd] Rethink how SelectionTree deals with macros and #includes.
Sam McCall [Wed, 20 Nov 2019 22:25:17 +0000 (23:25 +0100)]
[clangd] Rethink how SelectionTree deals with macros and #includes.

Summary:
The exclusive-claim model is successful at resolving conflicts over tokens
between parent/child or siblings. However claims at the spelled-token
level do the wrong thing for macro expansions, where siblings can be
equally associated with the macro invocation.
Moreover, any model that only uses the endpoints in a range can fail when
a macro invocation occurs inside the node.

To address this, we use the existing TokenBuffer in more depth.
Claims are expressed in terms of expanded tokens, so there is no need to worry
about macros, includes etc.

Once we know which expanded tokens were claimed, they are mapped onto
spelled tokens for hit-testing.
This mapping is fairly flexible, currently the handling of macros is
pretty simple (map macro args onto spellings, other macro expansions onto the
macro name token).
This mapping is in principle token-by-token for correctness (though
there's some batching for performance).

The aggregation of the selection enum is now more principled as we need to be
able to aggregate several hit-test results together.

For simplicity i removed the ability to determine selectedness of TUDecl.
(That was originally implemented in 90a5bf92ff97b1, but doesn't seem to be very
important or worth the complexity any longer).

The expandedTokens(SourceLocation) helper could be added locally, but seems to
make sense on TokenBuffer.

Fixes https://github.com/clangd/clangd/issues/202
Fixes https://github.com/clangd/clangd/issues/126

Reviewers: hokein

Subscribers: MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits, ilya-biryukov

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70512

4 years ago[LLDB] [ARM] Use r11 as frame pointer on Windows on ARM
Martin Storsjö [Tue, 15 Oct 2019 21:08:28 +0000 (00:08 +0300)]
[LLDB] [ARM] Use r11 as frame pointer on Windows on ARM

Extend EmulateMOVRdRm to identify "mov r11, sp" in thumb mode as
setting the frame pointer, if r11 is the frame pointer register.

Differential Revision: https://reviews.llvm.org/D70797

4 years ago[lldb][NFC] Remove unused ClangASTContext::GetBasicType(ConstString)
Raphael Isemann [Fri, 29 Nov 2019 13:08:01 +0000 (14:08 +0100)]
[lldb][NFC] Remove unused ClangASTContext::GetBasicType(ConstString)

4 years ago[lldb][NFC] Remove ClangASTContext::GetBuiltinTypeForEncodingAndBitSize overload
Raphael Isemann [Fri, 29 Nov 2019 12:43:23 +0000 (13:43 +0100)]
[lldb][NFC] Remove ClangASTContext::GetBuiltinTypeForEncodingAndBitSize overload

4 years ago[lldb][NFC] Explicitly ask for a ClangASTContext in ClangASTSource
Raphael Isemann [Fri, 29 Nov 2019 12:02:41 +0000 (13:02 +0100)]
[lldb][NFC] Explicitly ask for a ClangASTContext in ClangASTSource

ClangASTSource currently takes a clang::ASTContext and keeps that
around, but a lot of LLDB's functionality for doing operations
on a clang::ASTContext is in its ClangASTContext twin class. We
currently constantly recompute the respective ClangASTContext
from the clang::ASTContext while we instead could just pass and
store a ClangASTContext in the ClangASTSource. This also allows
us to get rid of a bunch of unreachable error checking for cases
where recomputation fails for some reason.

4 years ago[InstCombine] Run the cast.ll test a twice, now also testing little endian. NFC
Bjorn Pettersson [Fri, 29 Nov 2019 12:24:13 +0000 (13:24 +0100)]
[InstCombine] Run the cast.ll test a twice, now also testing little endian. NFC

Some tests in test/Transforms/InstCombine/cast.ll depend on
endianness. Added a second run line to run the tests with both
big and little endian. In the past we only compiled for big
endian, and then it was hard to see if any big endian bugfixes
would impact the little endian result etc.

4 years ago[lldb][NFC] Early exit in ClangASTContext::CreateInstance
Raphael Isemann [Fri, 29 Nov 2019 11:40:19 +0000 (12:40 +0100)]
[lldb][NFC] Early exit in ClangASTContext::CreateInstance

4 years ago[lldb] Fix windows build for 38870af
Pavel Labath [Fri, 29 Nov 2019 11:48:25 +0000 (12:48 +0100)]
[lldb] Fix windows build for 38870af

4 years ago[lldb][NFC] Simplify regex_chars in CommandCompletions
Raphael Isemann [Fri, 29 Nov 2019 11:26:33 +0000 (12:26 +0100)]
[lldb][NFC] Simplify regex_chars in CommandCompletions

4 years ago[lldb][NFC] Remove dead logging code from DWARFASTParserClang::CompleteRecordType
Raphael Isemann [Fri, 29 Nov 2019 11:05:47 +0000 (12:05 +0100)]
[lldb][NFC] Remove dead logging code from DWARFASTParserClang::CompleteRecordType

This code is behind a `if (log)` that is always a nullptr as the initializer
was commented out. One could uncomment the initializer code, but then this logging
code just leads to a deadlock as it tries to aquire the module lock.
This removes the logging code until I get this working again.

4 years ago[clangd] Fix 407ac2e, which was broken and committed too soon
Sam McCall [Fri, 29 Nov 2019 11:05:00 +0000 (12:05 +0100)]
[clangd] Fix 407ac2e, which was broken and committed too soon

4 years ago[lldb] Remove FileSpec->CompileUnit inheritance
Pavel Labath [Thu, 28 Nov 2019 15:22:44 +0000 (16:22 +0100)]
[lldb] Remove FileSpec->CompileUnit inheritance

Summary:
CompileUnit is a complicated class. Having it be implicitly convertible
to a FileSpec makes reasoning about it even harder.

This patch replaces the inheritance by a simple member and an accessor
function. This avoid the need for casting in places where one needed to
force a CompileUnit to be treated as a FileSpec, and does not add much
verbosity elsewhere.

It also fixes a bug where we were wrongly comparing CompileUnit& and a
CompileUnit*, which compiled due to a combination of this inheritance
and the FileSpec*->FileSpec implicit constructor.

Reviewers: teemperor, JDevlieghere, jdoerfert

Subscribers: lldb-commits

Tags: #lldb

Differential Revision: https://reviews.llvm.org/D70827

4 years ago[ARM] Fix instruction selection for ARMISD::CMOV with f16 type
Victor Campos [Tue, 19 Nov 2019 09:55:16 +0000 (09:55 +0000)]
[ARM] Fix instruction selection for ARMISD::CMOV with f16 type

Summary:
In the cases where the CMOV (f16) SDNode is used with condition codes
LT, LE, VC or NE, it is successfully selected into a VSEL instruction.

In the remaining cases, however, instruction selection fails since VSEL
does not support other condition codes.

This patch handles such cases by using the single-precision version of
the VMOV instruction.

Reviewers: ostannard, dmgreen

Reviewed By: dmgreen

Subscribers: kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70667

4 years ago[Syntax] Build SimpleDeclaration node that groups multiple declarators
Ilya Biryukov [Fri, 29 Nov 2019 10:32:58 +0000 (11:32 +0100)]
[Syntax] Build SimpleDeclaration node that groups multiple declarators

Summary:
Also remove the temporary TopLevelDeclaration node and add
UnknownDeclaration to represent other unknown nodes.

See the follow-up change for building more top-level declarations.
Adding declarators is also pretty involved and will be done in another
follow-up patch.

Reviewers: gribozavr2

Reviewed By: gribozavr2

Subscribers: merge_guards_bot, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70787

4 years ago[lldb][NFC] Fix header guard comment in ThreadSafeDenseMap.h
Raphael Isemann [Fri, 29 Nov 2019 10:34:18 +0000 (11:34 +0100)]
[lldb][NFC] Fix header guard comment in ThreadSafeDenseMap.h

4 years ago[AST] Remove unused and undefined `TypeLoc::IgnoreMacroDefinitions` function. NFC
Ilya Biryukov [Fri, 29 Nov 2019 10:10:01 +0000 (11:10 +0100)]
[AST] Remove unused and undefined `TypeLoc::IgnoreMacroDefinitions` function. NFC

Looks like an accidental leftover from the older version of the code.

4 years ago[clangd] Log cc1 args at verbose level.
Sam McCall [Thu, 28 Nov 2019 18:22:50 +0000 (19:22 +0100)]
[clangd] Log cc1 args at verbose level.

Summary: This will help debugging driver issues.

Reviewers: kbobyrev

Subscribers: ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70832

4 years ago[clangd] Correct the file path in Edit::replacements when generating the rename edit.
Haojian Wu [Thu, 28 Nov 2019 15:48:49 +0000 (16:48 +0100)]
[clangd] Correct the file path in Edit::replacements when generating the rename edit.

Summary:
The file path was set to the file content previously, and it isn't
covered by normal clangd & unittest code path (as we only uses the
offset, length, replacement text).

Reviewers: ilya-biryukov

Subscribers: MaskRay, jkorous, arphaman, kadircet, usaxena95, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D70828

4 years ago[Syntax] Add a comment explaining the pointer keys in std::map<Token*, ...>. NFC
Ilya Biryukov [Fri, 29 Nov 2019 08:48:00 +0000 (09:48 +0100)]
[Syntax] Add a comment explaining the pointer keys in std::map<Token*, ...>. NFC

4 years ago[Syntax] Remove unused parameter from `TreeBuilder::markChildToken`. NFC
Ilya Biryukov [Fri, 29 Nov 2019 08:44:25 +0000 (09:44 +0100)]
[Syntax] Remove unused parameter from `TreeBuilder::markChildToken`. NFC

4 years ago[llvm-readelf][test] - Update comment in elf-verdef-invalid.test. NFC.
Georgii Rymar [Fri, 29 Nov 2019 08:38:27 +0000 (11:38 +0300)]
[llvm-readelf][test] - Update comment in elf-verdef-invalid.test. NFC.

It was suggested to change it during review of D70810,
but I've forgotten to update it before commit.

4 years ago[llvm-readelf/llvm-readobj] - Check version of SHT_GNU_verdef section entries when...
Georgii Rymar [Thu, 28 Nov 2019 10:20:25 +0000 (13:20 +0300)]
[llvm-readelf/llvm-readobj] - Check version of SHT_GNU_verdef section entries when dumping.

Elfxx_Verdef contains the following field:

vd_version
Version revision. This field shall be set to 1.
(https://refspecs.linuxfoundation.org/LSB_5.0.0/LSB-Core-generic/LSB-Core-generic/symversion.html)

Our code should check the struct version for correctness. This patch does that.
(This will help to simplify or eliminate ELFDumper<ELFT>::LoadVersionDefs() which
has it's own logic to parse version definitions for no reason. It checks the
struct version currently).

Differential revision: https://reviews.llvm.org/D70810

4 years ago[libcxx] Add -Wno-deprecated-copy to the test config
David Zarzycki [Fri, 29 Nov 2019 07:22:56 +0000 (09:22 +0200)]
[libcxx] Add -Wno-deprecated-copy to the test config

4 years ago[yaml2obj] - Add a way to describe content of the SHT_GNU_verneed section with "Content".
Georgii Rymar [Thu, 28 Nov 2019 13:27:01 +0000 (16:27 +0300)]
[yaml2obj] - Add a way to describe content of the SHT_GNU_verneed section with "Content".

There is no way to set raw content for SHT_GNU_verneed section.
This patch implements it.

Differential revision: https://reviews.llvm.org/D70816

4 years ago[Attributor] Deduce dereferenceable based on accessed bytes map
Hideto Ueno [Fri, 29 Nov 2019 06:55:58 +0000 (06:55 +0000)]
[Attributor] Deduce dereferenceable based on accessed bytes map

Summary:
This patch introduces the deduction based on load/store instructions whose pointer operand is a non-inbounds GEP instruction.
For example if we have,
```
void f(int *u){
 u[0] = 0;
 u[1] = 1;
 u[2] = 2;
}
```
then u must be dereferenceable(12).

This patch is inspired by D64258

Reviewers: jdoerfert, spatel, hfinkel, RKSimon, sstefan1, xbolva00, dtemirbulatov

Reviewed By: jdoerfert

Subscribers: jfb, lebedev.ri, xbolva00, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70714

4 years ago[Attributor] Remove dereferenceable_or_null when nonull is present
Hideto Ueno [Fri, 29 Nov 2019 06:45:07 +0000 (06:45 +0000)]
[Attributor] Remove dereferenceable_or_null when nonull is present

Summary: This patch prevents the simultaneous presence of `dereferenceable` and `dereferenceable_or_null` attribute

Reviewers: jdoerfert, sstefan1

Reviewed By: jdoerfert

Subscribers: lebedev.ri, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D70789

4 years ago[PassInstrumentation] Remove excess newline for the new pass manager
Fangrui Song [Fri, 29 Nov 2019 00:42:27 +0000 (16:42 -0800)]
[PassInstrumentation] Remove excess newline for the new pass manager

This also removes excess newline for the legacy pass manager when -filter-print-funcs is specified.

4 years agogn build: Add a toggle for building against the commandline tools SDK on macOS
Nico Weber [Fri, 29 Nov 2019 00:30:21 +0000 (19:30 -0500)]
gn build: Add a toggle for building against the commandline tools SDK on macOS

4 years ago[LegalizeTypes] Add strict FP support to SoftenFloatRes_FP_ROUND. Fix mistake in...
Craig Topper [Thu, 28 Nov 2019 23:28:27 +0000 (15:28 -0800)]
[LegalizeTypes] Add strict FP support to SoftenFloatRes_FP_ROUND. Fix mistake in SoftenFloatRes_FP_EXTEND.

These will be needed for ARM fp-instrinsics.ll which is currently
XFAILed.

One of the getOperand calls in SoftenFloatRes_FP_EXTEND was not
taking strict FP into account. It only affected the call
to setTypeListBeforeSoften which only has an effect on some targets.

4 years ago[LegalizeTypes] In SoftenFloatRes_FNEG, always generate integer arithmetic, never...
Craig Topper [Thu, 28 Nov 2019 19:36:55 +0000 (11:36 -0800)]
[LegalizeTypes] In SoftenFloatRes_FNEG, always generate integer arithmetic, never fall back to using fsub.

We would previously fallback if the type wasn't f32/f64/f128. But
I don't think any of the other floating point types ever go through
the softening code anyway. So this code is dead.

4 years ago[LegalizeTypes] Use SoftenFloatRes_Unary in SoftenFloatRes_FCBRT to reduce code.
Craig Topper [Thu, 28 Nov 2019 18:40:50 +0000 (10:40 -0800)]
[LegalizeTypes] Use SoftenFloatRes_Unary in SoftenFloatRes_FCBRT to reduce code.

We don't have a STRICT_CBRT ISD opcode, but we can still
use SoftenFloatRes_Unary to simplify some code.

4 years ago[DAGCombiner] Peek through vector concats when trying to combine shuffles.
Amaury Séchet [Fri, 22 Nov 2019 22:39:18 +0000 (23:39 +0100)]
[DAGCombiner] Peek through vector concats when trying to combine shuffles.

Summary: This combine showed up as needed when exploring the regression when processing the DAG in topological order.

Reviewers: craig.topper, efriedma, RKSimon, lebedev.ri

Subscribers: llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68195

4 years ago[OpenMP] build offload plugins before testing them
Bryan Chan [Thu, 28 Nov 2019 07:46:18 +0000 (02:46 -0500)]
[OpenMP] build offload plugins before testing them

Summary:
"make check-all" or "make check-libomptarget" would attempt to run offloading
tests before the offload plugins are built. This patch corrects that by adding
dependencies to the libomptarget CMake rules.

Reviewers: jdoerfert

Subscribers: mgorny, guansong, openmp-commits

Tags: #openmp

Differential Revision: https://reviews.llvm.org/D70803

4 years ago[LegacyPassManager] Simplify FunctionPass::assignPassManager
Fangrui Song [Thu, 28 Nov 2019 22:00:12 +0000 (14:00 -0800)]
[LegacyPassManager] Simplify FunctionPass::assignPassManager

And make it clear the parameter PreferredType is unused for FunctionPass.

4 years ago[Clang] Bypass distro detection on non-Linux hosts
Alexandre Ganea [Thu, 28 Nov 2019 20:56:00 +0000 (15:56 -0500)]
[Clang] Bypass distro detection on non-Linux hosts

Skip distro detection when we're not running on Linux, or when the target triple is not Linux. This saves a few OS calls for each invocation of clang.exe.

Differential Revision: https://reviews.llvm.org/D70467