platform/upstream/llvm.git
5 years ago[NFC] Ensure deterministic order of dead exit blocks
Max Kazantsev [Thu, 22 Nov 2018 12:33:41 +0000 (12:33 +0000)]
[NFC] Ensure deterministic order of dead exit blocks

llvm-svn: 347457

5 years ago[AArch64] Fix SelectionDAG infinite loop for v1i64 SCALAR_TO_VECTOR
John Brawn [Thu, 22 Nov 2018 11:45:23 +0000 (11:45 +0000)]
[AArch64] Fix SelectionDAG infinite loop for v1i64 SCALAR_TO_VECTOR

A consequence of r347274 is that SCALAR_TO_VECTOR can be converted into
BUILD_VECTOR by SimplifyDemandedBits, but LowerBUILD_VECTOR can turn
BUILD_VECTOR into SCALAR_TO_VECTOR so we get an infinite loop.

Fix this by making LowerBUILD_VECTOR not do this transformation for those
vectors that would get transformed back, i.e. BUILD_VECTOR of a single-element
constant vector. Doing that means we get a DUP, which we then need to recognise
in ISel as a copy.

llvm-svn: 347456

5 years ago[ELF] - Make SymbolTable::addDefined return Defined.
George Rimar [Thu, 22 Nov 2018 11:40:08 +0000 (11:40 +0000)]
[ELF] - Make SymbolTable::addDefined return Defined.

Now it returns Symbol. This should be NFC that
avoids doing cast at the caller's sides.

Differential revision: https://reviews.llvm.org/D54627

llvm-svn: 347455

5 years ago[NFC] Simplify code by using standard exit blocks collection
Max Kazantsev [Thu, 22 Nov 2018 10:48:30 +0000 (10:48 +0000)]
[NFC] Simplify code by using standard exit blocks collection

llvm-svn: 347454

5 years ago[ASTMatchers] Re-generate ast matchers doc after rL346455.
Clement Courbet [Thu, 22 Nov 2018 10:44:36 +0000 (10:44 +0000)]
[ASTMatchers] Re-generate ast matchers doc after rL346455.

llvm-svn: 347453

5 years ago[TI removal] Leverage the fact that TerminatorInst is gone to create
Chandler Carruth [Thu, 22 Nov 2018 10:31:35 +0000 (10:31 +0000)]
[TI removal] Leverage the fact that TerminatorInst is gone to create
a normal base class that provides all common "call" functionality.

This merges two complex CRTP mixins for the common "call" logic and
common operand bundle logic into a single, normal base class of
`CallInst` and `InvokeInst`. Going forward, users can typically
`dyn_cast<CallBase>` and use the resulting API. No more need for the
`CallSite` wrapper. I'm planning to migrate current usage of the wrapper
to directly use the base class and then it can be removed, but those are
simpler and much more incremental steps. The big change is to introduce
this abstraction into the type system.

I've tried to do some basic simplifications of the APIs that I couldn't
really help but touch as part of this:
- I've tried to organize the attribute API and bundle API into groups to
  make understanding the API of `CallBase` easier. Without this,
  I wasn't able to navigate the API sanely for all of the ways I needed
  to modify it.
- I've added what seem like more clear and consistent APIs for getting
  at the called operand. These ended up being especially useful to
  consolidate the *numerous* duplicated code paths trying to do this.
- I've largely reworked the organization and implementation of the APIs
  for computing the argument operands as they needed to change to work
  with the new subclass approach.

To minimize any cost associated with this abstraction, I've moved the
operand layout in memory to store the called operand last. This makes
its position relative to the end of the operand array the same,
regardless of the subclass. It should make it much cheaper to reference
from the `CallBase` abstraction, and this is likely one of the most
frequent things to query.

We do still pay one abstraction penalty here: we have to branch to
determine whether there are 0 or 2 extra operands when computing the end
of the argument operand sequence. However, that seems both rare and
should optimize well. I've implemented this in a way specifically
designed to allow it to optimize fairly well. If this shows up in
profiles, we can add overrides of the relevant methods to the subclasses
that bypass this penalty. It seems very unlikely that this will be an
issue as the code was *already* dealing with an ever present abstraction
of whether or not there are operand bundles, so this isn't the first
branch to go into the computation.

I've tried to remove as much of the obvious vestigial API surface of the
old CRTP implementation as I could, but I suspect there is further
cleanup that should now be possible, especially around the operand
bundle APIs. I'm leaving all of that for future work in this patch as
enough things are changing here as-is.

One thing that made this harder for me to reason about and debug was the
pervasive use of unsigned values in subtraction and other arithmetic
computations. I had to debug more than one unintentional wrap. I've
switched a few of these to use `int` which seems substantially simpler,
but I've held back from doing this more broadly to avoid creating
confusing divergence within a single class's API.

I also worked to remove all of the magic numbers used to index into
operands, putting them behind named constants or putting them into
a single method with a comment and strictly using the method elsewhere.
This was necessary to be able to re-layout the operands as discussed
above.

Thanks to Ben for reviewing this (somewhat large and awkward) patch!

Differential Revision: https://reviews.llvm.org/D54788

llvm-svn: 347452

5 years agoUnbreak FreeBSD build.
David Carlier [Thu, 22 Nov 2018 10:31:01 +0000 (10:31 +0000)]
Unbreak FreeBSD build.
M    lib/sanitizer_common/sanitizer_platform_limits_posix.cc

llvm-svn: 347451

5 years ago[clangd] Respect task cancellation in TUScheduler.
Sam McCall [Thu, 22 Nov 2018 10:22:16 +0000 (10:22 +0000)]
[clangd] Respect task cancellation in TUScheduler.

Summary:
- Reads are never executed if canceled before ready-to run.
  In practice, we finalize cancelled reads eagerly and out-of-order.
- Cancelled reads don't prevent prior updates from being elided, as they don't
  actually depend on the result of the update.
- Updates are downgraded from WantDiagnostics::Yes to WantDiagnostics::Auto when
  cancelled, which allows them to be elided when all dependent reads are
  cancelled and there are subsequent writes. (e.g. when the queue is backed up
  with cancelled requests).

The queue operations aren't optimal (we scan the whole queue for cancelled
tasks every time the scheduler runs, and check cancellation twice in the end).
However I believe these costs are still trivial in practice (compared to any
AST operation) and the logic can be cleanly separated from the rest of the
scheduler.

Reviewers: ilya-biryukov

Subscribers: javed.absar, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits

Differential Revision: https://reviews.llvm.org/D54746

llvm-svn: 347450

5 years agoMove the llvm lit test dependencies to clang-tools-extra.
Haojian Wu [Thu, 22 Nov 2018 10:14:55 +0000 (10:14 +0000)]
Move the llvm lit test dependencies to clang-tools-extra.

Summary: Part of revert r343473

Reviewers: mgorny

Subscribers: cfe-commits

Differential Revision: https://reviews.llvm.org/D54798

llvm-svn: 347449

5 years agoRevert r343473 "Move llvm util dependencies from clang-tools-extra to add_lit_target."
Haojian Wu [Thu, 22 Nov 2018 10:14:24 +0000 (10:14 +0000)]
Revert r343473 "Move llvm util dependencies from clang-tools-extra to add_lit_target."

Summary:
It will cause test tools `FileCheck`, `count`, `not` being built blindly, these
dependencies should move back to clang-tools-extra.

Reviewers: mgorny

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D54797

llvm-svn: 347448

5 years ago[ARM GlobalISel] Add test for BFC. NFCI
Diana Picus [Thu, 22 Nov 2018 09:54:14 +0000 (09:54 +0000)]
[ARM GlobalISel] Add test for BFC. NFCI

r334871 has made it possible for TableGen'erated code to select BFC, but
it has not added a test for it on the ARM side. Add it now to make sure
we don't introduce regressions if we ever change anything about that
rule.

llvm-svn: 347447

5 years agoThe result of is.constant() is unsigned.
Bill Wendling [Thu, 22 Nov 2018 09:31:08 +0000 (09:31 +0000)]
The result of is.constant() is unsigned.

llvm-svn: 347446

5 years ago[SystemZTTIImpl] Give correct cost values for vector bswap intrinsics.
Jonas Paulsson [Thu, 22 Nov 2018 07:17:29 +0000 (07:17 +0000)]
[SystemZTTIImpl]  Give correct cost values for vector bswap intrinsics.

Implement getIntrinsicInstrCost() and return costs reflecting that bswap can
be done with a vperm per vector register.

Review: Ulrich Weigand
https://reviews.llvm.org/D54789

llvm-svn: 347445

5 years ago[Driver] Support XRay on Fuchsia
Petr Hosek [Thu, 22 Nov 2018 02:36:47 +0000 (02:36 +0000)]
[Driver] Support XRay on Fuchsia

This enables support for XRay in Fuchsia Clang driver.

Differential Revision: https://reviews.llvm.org/D52160

llvm-svn: 347444

5 years ago[XRay] Support for Fuchsia
Petr Hosek [Thu, 22 Nov 2018 02:00:44 +0000 (02:00 +0000)]
[XRay] Support for Fuchsia

This extends XRay to support Fuchsia.

Differential Revision: https://reviews.llvm.org/D52162

llvm-svn: 347443

5 years agotsan: Update measurements in check_analyze.sh.
Peter Collingbourne [Thu, 22 Nov 2018 00:47:37 +0000 (00:47 +0000)]
tsan: Update measurements in check_analyze.sh.

These changed as a result of r347379. Unfortunately there was a
regression; filed PR39748 to track it.

Differential Revision: https://reviews.llvm.org/D54821

llvm-svn: 347442

5 years ago[llvm-size] Use empty() and range-based for loop. NFC
Fangrui Song [Thu, 22 Nov 2018 00:44:17 +0000 (00:44 +0000)]
[llvm-size] Use empty() and range-based for loop. NFC

llvm-svn: 347441

5 years ago[llvm-mca] Add test case (NFC)
Evandro Menezes [Thu, 22 Nov 2018 00:38:36 +0000 (00:38 +0000)]
[llvm-mca] Add test case (NFC)

Add test case that will serve as the base for D54820.

llvm-svn: 347440

5 years agotsan: Correct the name of an executable.
Peter Collingbourne [Thu, 22 Nov 2018 00:00:26 +0000 (00:00 +0000)]
tsan: Correct the name of an executable.

llvm-svn: 347439

5 years ago[x86] use FileCheck to verify output; NFC
Sanjay Patel [Wed, 21 Nov 2018 23:39:19 +0000 (23:39 +0000)]
[x86] use FileCheck to verify output; NFC

llvm-svn: 347438

5 years ago[llvm-mca] Add test case (NFC)
Evandro Menezes [Wed, 21 Nov 2018 23:36:40 +0000 (23:36 +0000)]
[llvm-mca] Add test case (NFC)

Fix previous commit r347434.

llvm-svn: 347437

5 years agoAdd a ubsan blacklist entry for libstdc++ 8.0.1.
Peter Collingbourne [Wed, 21 Nov 2018 23:04:39 +0000 (23:04 +0000)]
Add a ubsan blacklist entry for libstdc++ 8.0.1.

llvm-svn: 347436

5 years ago[libcxx] Improve error message when an invalid directory is provided as use_system_cx...
Louis Dionne [Wed, 21 Nov 2018 23:00:45 +0000 (23:00 +0000)]
[libcxx] Improve error message when an invalid directory is provided as use_system_cxx_lib

llvm-svn: 347435

5 years ago[llvm-mca] Add test case (NFC)
Evandro Menezes [Wed, 21 Nov 2018 22:57:46 +0000 (22:57 +0000)]
[llvm-mca] Add test case (NFC)

Add test case that will serve as the base for D54777.

llvm-svn: 347434

5 years agoRemoving test/MC/Mips/reloc-directive-label-offset.s temporarily
Vladimir Stefanovic [Wed, 21 Nov 2018 22:08:34 +0000 (22:08 +0000)]
Removing test/MC/Mips/reloc-directive-label-offset.s temporarily

This test is failing on llvm-clang-x86_64-expensive-checks-win builder.
Removing it until I get it fixed.

llvm-svn: 347433

5 years ago[PM] correcting return value for new-pass-manager version of Scalarizer
Fedor Sergeev [Wed, 21 Nov 2018 22:01:19 +0000 (22:01 +0000)]
[PM] correcting return value for new-pass-manager version of Scalarizer

Obvious mistake missed during D54695 review.

llvm-svn: 347432

5 years ago[mingw] Use unmangled name after the $ in the section name
Reid Kleckner [Wed, 21 Nov 2018 22:01:10 +0000 (22:01 +0000)]
[mingw] Use unmangled name after the $ in the section name

GCC does it this way, and we have to be consistent. This includes
stdcall and fastcall functions with suffixes. I confirmed that a
fastcall function named "foo" ends up in ".text$foo", not
".text$@foo@8".

Based on a patch by Andrew Yohn!

Fixes PR39218.

Differential Revision: https://reviews.llvm.org/D54762

llvm-svn: 347431

5 years agoRevert "[Driver] Use --push/pop-state with Sanitizer link deps"
Petr Hosek [Wed, 21 Nov 2018 21:59:39 +0000 (21:59 +0000)]
Revert "[Driver] Use --push/pop-state with Sanitizer link deps"

This reverts commit r347413: older versions of ld.gold that are used
by Android don't support --push/pop-state which broke sanitizer bots.

llvm-svn: 347430

5 years agoCOFF: ICF: Include contents of referenced sections in initial partitioning hash....
Peter Collingbourne [Wed, 21 Nov 2018 21:29:35 +0000 (21:29 +0000)]
COFF: ICF: Include contents of referenced sections in initial partitioning hash. NFCI.

Previously we were taking over 13 minutes to link Firefox's xul.dll
on ARM64; this reduces link time to around 18s on my machine.

The root cause of the problem was that all of the input .pdata sections
had the same unrelocated section data and therefore the same hash,
which made segregation quadratic in the number of .pdata sections. The
reason why we weren't observing this on other architectures was that
ARM has a different .pdata format. On non-ARM the format is (start
address, end address, .xdata), which caused the size of the function
to appear in the unrelocated section data where the end address field
is. However, the ARM format omits the end address field.

Fixes PR39667.

Differential Revision: https://reviews.llvm.org/D54809

llvm-svn: 347429

5 years ago[PowerPC][NFC] Split PPCMCCodeEmitter into header and cpp file.
Stefan Pintilie [Wed, 21 Nov 2018 21:23:50 +0000 (21:23 +0000)]
[PowerPC][NFC] Split PPCMCCodeEmitter into header and cpp file.

This is further cleanup for PPCMCCodeEmitter. The class had been contained
within the cpp file alone. Now it has been split up between a header file and
a cpp file which allows other classes to make use of the functions in this class
if required.

llvm-svn: 347428

5 years ago[libcxx] Remove unused definition of aligned allocation macro on old OS X
Louis Dionne [Wed, 21 Nov 2018 21:22:08 +0000 (21:22 +0000)]
[libcxx] Remove unused definition of aligned allocation macro on old OS X

We don't support mac OS 10.6 and older anymore, so this macro can never
be defined. This bit of code had been added in D28931 as a fix for
PR31448, but it doesn't seem necessary anymore.

llvm-svn: 347427

5 years ago[Sanitizer] Adding setvbuf in supported platforms and other stream buffer functions
David Carlier [Wed, 21 Nov 2018 21:17:46 +0000 (21:17 +0000)]
[Sanitizer] Adding setvbuf in supported platforms and other stream buffer functions

- Enabling setvbuf interceptions for non NetBSD platforms.
- setbuf, setbuffer, setlinebuf as well.

Reviewers: vitalybuka, krytarowski

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D54779

llvm-svn: 347426

5 years ago[OPENMP][NVPTX]Emit default locations as constant with undefined mode.
Alexey Bataev [Wed, 21 Nov 2018 21:04:34 +0000 (21:04 +0000)]
[OPENMP][NVPTX]Emit default locations as constant with undefined mode.

For the NVPTX target default locations should be emitted as constants +
additional info must be emitted in the reserved_2 field of the ident_t
structure. The 1st bit controls the execution mode and the 2nd bit
controls use of the lightweight runtime. The combination of the bits for
Non-SPMD mode + lightweight runtime represents special undefined mode,
used outside of the target regions for orphaned directives or functions.
Should allow and additional optimization inside of the target regions.

llvm-svn: 347425

5 years ago[DAGCombiner] refactor select-of-FP-constants transform
Sanjay Patel [Wed, 21 Nov 2018 20:54:47 +0000 (20:54 +0000)]
[DAGCombiner] refactor select-of-FP-constants transform

This transform needs to be limited.

We are converting to a constant pool load very early, and we
are turning loads that are independent of the select condition
(and therefore speculatable) into a dependent non-speculatable
load.

We may also be transferring a condition code from an FP register
to integer to create that dependent load.

llvm-svn: 347424

5 years ago[libcxx] Fix incorrect iterator type in vector container test
Louis Dionne [Wed, 21 Nov 2018 20:53:07 +0000 (20:53 +0000)]
[libcxx] Fix incorrect iterator type in vector container test

The iterator types for different specializations of containers with the
same element type but different allocators are not required to be
convertible. This patch makes the test to take the iterator type from
the same container specialization as the created container.

Reviewed as https://reviews.llvm.org/D54806.
Thanks to Andrey Maksimov for the patch.

llvm-svn: 347423

5 years ago[PowerPC][NFC] Minor Code Cleaup for PPCMCCodeEmitter.
Stefan Pintilie [Wed, 21 Nov 2018 20:47:59 +0000 (20:47 +0000)]
[PowerPC][NFC] Minor Code Cleaup for PPCMCCodeEmitter.

llvm-svn: 347422

5 years ago[libcxx] Mark strstreams tests as being supported on all OS X versions
Louis Dionne [Wed, 21 Nov 2018 20:47:00 +0000 (20:47 +0000)]
[libcxx] Mark strstreams tests as being supported on all OS X versions

I wasn't able to reproduce the issue referred to by the comment using
the libc++'s shipped with mac OS X 10.7 and 10.8, so I assume this may
have been fixed in a function that is now shipped in the headers. In
that case, the tests will pass no matter what dylib we're using.

In the worst case, some test bots will start failing and I'll understand
why I was wrong, and I can create an actual lit feature for it. Note
that I could just leave this test alone, but this change is on the path
towards eradicating vendor-specific availability markup from the test
suite.

llvm-svn: 347421

5 years ago[LLVM] Allow modulemap installation
Eric Fiselier [Wed, 21 Nov 2018 20:46:50 +0000 (20:46 +0000)]
[LLVM] Allow modulemap installation

Summary:
Currently we can't install the modulemaps provided by LLVM, since they are not structured to support headers generated as part of the build (ex. `llvm/IR/Attributes.gen`).
This patch restructures the module maps in order to support installation.

Modules containing generated headers are defined in the new `module.extern.modulemap` file, and are referenced from the main `module.modulemap` using `extern module`. There are two versions of the `module.extern.modulemap` file; one used when building and another, `module.install.modulemap`, which is re-named during installation.

Users can opt-into module map installation using `-DLLVM_INSTALL_MODULEMAPS=ON`.  The default value is `OFF` due to llvm.org/PR31905.

Reviewers: rsmith, mehdi_amini, bruno, EricWF

Reviewed By: EricWF

Subscribers: tschuett, chapuni, mgorny, llvm-commits

Differential Revision: https://reviews.llvm.org/D53510

llvm-svn: 347420

5 years agoUpdate call to EvaluateAsInt() to the new syntax.
Bill Wendling [Wed, 21 Nov 2018 20:45:00 +0000 (20:45 +0000)]
Update call to EvaluateAsInt() to the new syntax.

llvm-svn: 347419

5 years agoUpdate call to EvaluateAsInt() to the new syntax.
Bill Wendling [Wed, 21 Nov 2018 20:44:38 +0000 (20:44 +0000)]
Update call to EvaluateAsInt() to the new syntax.

llvm-svn: 347418

5 years agoRe-Reinstate 347294 with a fix for the failures.
Bill Wendling [Wed, 21 Nov 2018 20:44:18 +0000 (20:44 +0000)]
Re-Reinstate 347294 with a fix for the failures.

Don't try to emit a scalar expression for a non-scalar argument to
__builtin_constant_p().

Third time's a charm!

llvm-svn: 347417

5 years agoFix missing includes in test header
Eric Fiselier [Wed, 21 Nov 2018 20:39:24 +0000 (20:39 +0000)]
Fix missing includes in test header

llvm-svn: 347416

5 years ago[compiler-rt][UBSan] silence_unsigned_overflow: do *NOT* ignore *fatal* unsigned...
Roman Lebedev [Wed, 21 Nov 2018 20:35:43 +0000 (20:35 +0000)]
[compiler-rt][UBSan] silence_unsigned_overflow: do *NOT* ignore *fatal* unsigned overflows

Summary:
D48660 / rL335762 added a `silence_unsigned_overflow` env flag for [[ https://github.com/google/oss-fuzz/pull/1717 | oss-fuzz needs ]],
that allows to silence the reports from unsigned overflows.
It makes sense, it is there because `-fsanitize=integer` sanitizer is not enabled on oss-fuzz,
so this allows to still use it as an interestingness signal, without getting the actual reports.

However there is a slight problem here.
All types of unsigned overflows are ignored.
Even if `-fno-sanitize-recover=unsigned` was used (which means the program will die after the report)
there will still be no report, the program will just silently die.

At the moment there are just two projects on oss-fuzz that care:
* [[ https://github.com/google/oss-fuzz/blob/8eeffa627f937040aaf8ba1b7d93f43f77d74fb9/projects/llvm_libcxx/build.sh#L18-L20 | libc++ ]]
* [[ https://github.com/google/oss-fuzz/blob/8eeffa627f937040aaf8ba1b7d93f43f77d74fb9/projects/librawspeed/build.sh | RawSpeed ]] (me)

I suppose this could be overridden there ^, but i really don't think this is intended behavior in any case..

Reviewers: kcc, Dor1s, #sanitizers, filcab, vsk, kubamracek

Reviewed By: Dor1s

Subscribers: dberris, mclow.lists, llvm-commits

Tags: #sanitizers

Differential Revision: https://reviews.llvm.org/D54771

llvm-svn: 347415

5 years ago[InstCombine] Add tests for funnel shift with zero operand; NFC
Nikita Popov [Wed, 21 Nov 2018 20:34:11 +0000 (20:34 +0000)]
[InstCombine] Add tests for funnel shift with zero operand; NFC

These are additional baseline tests for D54778.

llvm-svn: 347414

5 years ago[Driver] Use --push/pop-state with Sanitizer link deps
Petr Hosek [Wed, 21 Nov 2018 20:33:12 +0000 (20:33 +0000)]
[Driver] Use --push/pop-state with Sanitizer link deps

Sanitizer runtime link deps handling passes --no-as-needed because of
PR15823, but it never undoes it and this flag may affect other libraries
that come later on the link line. To avoid this, wrap Sanitizer link
deps in --push/pop-state.

Differential Revision: https://reviews.llvm.org/D54805

llvm-svn: 347413

5 years ago[lit] Add pthread to the compilation of the tests on Linux
Stella Stamenova [Wed, 21 Nov 2018 20:16:06 +0000 (20:16 +0000)]
[lit] Add pthread to the compilation of the tests on Linux

Summary: Right now only some platforms add pthread to the compilation, however, at least one of the tests requires the pthread library on Linux as well. Since the library is available, this change adds it by default on Linux.

Reviewers: labath, zturner, asmith

Subscribers: stella.stamenova, jfb, lldb-commits

Differential Revision: https://reviews.llvm.org/D54808

llvm-svn: 347412

5 years ago[OPENMP] Refactor code for parsing omp declare target directive and its clauses ...
Kelvin Li [Wed, 21 Nov 2018 20:15:57 +0000 (20:15 +0000)]
[OPENMP] Refactor code for parsing omp declare target directive and its clauses (NFC)

This patch refactor the code for parsing omp declare target directive and
its clauses.

Patch by pjeeva01 (Jeeva P.)

Differential Revision: https://reviews.llvm.org/D54708

llvm-svn: 347411

5 years ago[DAGCombiner] reduce code duplication; NFC
Sanjay Patel [Wed, 21 Nov 2018 20:00:32 +0000 (20:00 +0000)]
[DAGCombiner] reduce code duplication; NFC

llvm-svn: 347410

5 years ago[OPENMP]Fix handling of the LCVs in loop-based directives.
Alexey Bataev [Wed, 21 Nov 2018 19:41:10 +0000 (19:41 +0000)]
[OPENMP]Fix handling of the LCVs in loop-based directives.

Loop-control variables with the default data-sharing attributes should
not be captured in the OpenMP region as they are private by default.
Also, default attributes should be emitted for such variables in the
inner OpenMP regions for the correct data sharing during codegen.

llvm-svn: 347409

5 years ago[OPENMP] remove redundant MapTypeModifierSpecified flag in ParseOpenMP.cpp (NFC)
Kelvin Li [Wed, 21 Nov 2018 19:38:53 +0000 (19:38 +0000)]
[OPENMP] remove redundant MapTypeModifierSpecified flag in ParseOpenMP.cpp (NFC)

Whether the map type modifier is specified or not, the flag
MapTypeModifierSpecified is always set to true.

Patch by Ahsan Saghir

Differential Revision: https://reviews.llvm.org/D54638

llvm-svn: 347408

5 years ago[MergeFuncs] Generate alias instead of thunk if possible
Nikita Popov [Wed, 21 Nov 2018 19:37:19 +0000 (19:37 +0000)]
[MergeFuncs] Generate alias instead of thunk if possible

The MergeFunctions pass was originally intended to emit aliases
instead of thunks where possible (unnamed_addr). However, for a
long time this functionality was behind a flag hardcoded to false,
bitrotted and was eventually removed in r309313.

Originally the functionality was first disabled in r108417 due to
lack of support for aliases in Mach-O. I believe that this is no
longer the case nowadays, but not really familiar with this area.

In the interest of being conservative, this patch reintroduces the
aliasing functionality behind a default disabled -mergefunc-use-aliases
flag.

Differential Revision: https://reviews.llvm.org/D53285

llvm-svn: 347407

5 years ago[x86] add tests for select-of-FP-constants; NFC
Sanjay Patel [Wed, 21 Nov 2018 19:14:38 +0000 (19:14 +0000)]
[x86] add tests for select-of-FP-constants; NFC

llvm-svn: 347406

5 years ago[OPENMP] Support relational-op != (not-equal) as one of the canonical
Kelvin Li [Wed, 21 Nov 2018 19:10:48 +0000 (19:10 +0000)]
[OPENMP] Support relational-op != (not-equal) as one of the canonical
forms of random access iterator

In OpenMP 4.5, only 4 relational operators are supported: <, <=, >,
and >=.  This work is to enable support for relational operator
!= (not-equal) as one of the canonical forms.

Patch by Anh Tuyen Tran

Differential Revision: https://reviews.llvm.org/D54441

llvm-svn: 347405

5 years ago[ELF] Write IPLT header in -static -z retpolineplt mode
Fangrui Song [Wed, 21 Nov 2018 18:10:00 +0000 (18:10 +0000)]
[ELF] Write IPLT header in -static -z retpolineplt mode

Summary:
This fixes PR39711: -static -z retpolineplt does not produce retpoline PLT header.
-z now is not relevant.

Statically linked executable does not have PLT, but may have IPLT with no header. When -z retpolineplt is specified, however, the repoline PLT header should still be emitted.

I've checked that this fixes the FreeBSD reproduce in PR39711 and a Linux program statically linked against glibc. The programm print "Hi" rather than SIGILL/SIGSEGV.

getPltEntryOffset may look dirty after this patch, but it can be cleaned up later.

Another possible improvement is that when there are non-preemptible IFUNC symbols (rare case, e.g. -Bsymbolic), both In.Plt and In.Iplt can be non-empty and we'll emit the retpoline PLT header twice.

Reviewers: espindola, emaste, chandlerc, ruiu

Reviewed By: emaste

Subscribers: emaste, arichardson, krytarowski, llvm-commits

Differential Revision: https://reviews.llvm.org/D54782

llvm-svn: 347404

5 years ago[x86] fix predicate for avoiding vblendv
Sanjay Patel [Wed, 21 Nov 2018 18:02:50 +0000 (18:02 +0000)]
[x86] fix predicate for avoiding vblendv

It only makes sense to produce the logic ops when 1 of the
constants is +0.0. Otherwise, go with vblendv to reduce code.

llvm-svn: 347403

5 years agoMark lambda decl as invalid if a captured variable has an invalid type.
Jorge Gorbe Moya [Wed, 21 Nov 2018 17:49:37 +0000 (17:49 +0000)]
Mark lambda decl as invalid if a captured variable has an invalid type.

This causes the compiler to crash when trying to compute a layout for
the lambda closure type (see included test).

llvm-svn: 347402

5 years ago[x86] add test for FP select with constant; NFC
Sanjay Patel [Wed, 21 Nov 2018 17:47:18 +0000 (17:47 +0000)]
[x86] add test for FP select with constant; NFC

llvm-svn: 347401

5 years ago[libcxx] Make sure operator+ is declared with the right visibility attribute
Louis Dionne [Wed, 21 Nov 2018 17:31:55 +0000 (17:31 +0000)]
[libcxx] Make sure operator+ is declared with the right visibility attribute

Otherwise, Clang complains about internal_linkage not being applied to the
first declaration of the operator (and rightfully so).

llvm-svn: 347400

5 years ago[libcxx] Mark stray symbols as hidden to try and fix the build
Louis Dionne [Wed, 21 Nov 2018 17:00:52 +0000 (17:00 +0000)]
[libcxx] Mark stray symbols as hidden to try and fix the build

r347395 changed the ABI list on Linux, but two of those symbols are still
being exported from the shared object:

    _ZSt18make_exception_ptrINSt3__112future_errorEESt13exception_ptrT_
    _ZNSt3__1plIcNS_11char_traitsIcEENS_9allocatorIcEEEENS_12basic_stringIT_T0_T1_EERKS9_PKS6_

This commit makes sure those symbols are not exported, as they should be.

llvm-svn: 347399

5 years ago[mips][mc] Add basic support for R_MIPS_JALR/R_MICROMIPS_JALR
Vladimir Stefanovic [Wed, 21 Nov 2018 16:38:34 +0000 (16:38 +0000)]
[mips][mc] Add basic support for R_MIPS_JALR/R_MICROMIPS_JALR

R_MIPS_JALR/R_MICROMIPS_JALR can now be parsed in .s files and emitted to .o.
They are still not generated with JALR.

Differential revision: https://reviews.llvm.org/D54721

llvm-svn: 347398

5 years ago[MC] Support labels as offsets in .reloc directive
Vladimir Stefanovic [Wed, 21 Nov 2018 16:28:39 +0000 (16:28 +0000)]
[MC] Support labels as offsets in .reloc directive

Currently, expressions like

  .reloc 1f, R_MIPS_JALR, foo
  1: nop

are not allowed, ie. an offset in .reloc can only be absolute value.
This patch adds support for labels as offsets.
If offset is a forward declared label, MCObjectStreamer keeps the fixup locally
and adds it to the fixups vector after the label (and its offset) is defined.
label+number is not supported yet.

Differential revision: https://reviews.llvm.org/D53990

llvm-svn: 347397

5 years ago[NFC][libcxx] Add revision number to ABI changelog
Louis Dionne [Wed, 21 Nov 2018 16:27:14 +0000 (16:27 +0000)]
[NFC][libcxx] Add revision number to ABI changelog

llvm-svn: 347396

5 years ago[libcxx] Make sure we can build with -fvisibility=hidden on Linux
Louis Dionne [Wed, 21 Nov 2018 16:24:46 +0000 (16:24 +0000)]
[libcxx] Make sure we can build with -fvisibility=hidden on Linux

Summary:
This commit marks a few functions as hidden and removes them from the ABI list
on Linux such that libc++ can be built with -fvisibility=hidden. The functions
marked as hidden by this patch were exported from the shared object only
because they were implicitly instantiated function templates. It is safe
to stop exporting those symbols from the shared object because nobody could
actually depend on them: implicit instantiations are not taken from shared
objects.

The symbols removed in this commit are basically the same that had been
removed in https://reviews.llvm.org/D53868, but that patch had to be reverted
because it broke the build (because the functions were not marked as hidden
like this patch does).

Reviewers: EricWF, mclow.lists

Subscribers: christof, jkorous, dexonsmith, libcxx-commits

Differential Revision: https://reviews.llvm.org/D54639

llvm-svn: 347395

5 years ago[x86] add checks for asm to test; NFC
Sanjay Patel [Wed, 21 Nov 2018 15:26:35 +0000 (15:26 +0000)]
[x86] add checks for asm to test; NFC

llvm-svn: 347394

5 years ago[TargetLowering] SimplifyDemandedBits - only reduce known bits for integer constants
Simon Pilgrim [Wed, 21 Nov 2018 14:26:19 +0000 (14:26 +0000)]
[TargetLowering] SimplifyDemandedBits - only reduce known bits for integer constants

Avoids fuzzing crash found by Mikael Holmén.

llvm-svn: 347393

5 years ago[PM] Port Scalarizer to the new pass manager.
Mikael Holmen [Wed, 21 Nov 2018 14:00:17 +0000 (14:00 +0000)]
[PM] Port Scalarizer to the new pass manager.

Patch by: markus (Markus Lavin)

Reviewers: chandlerc, fedor.sergeev

Reviewed By: fedor.sergeev

Subscribers: llvm-commits, Ka-Ka, bjope

Differential Revision: https://reviews.llvm.org/D54695

llvm-svn: 347392

5 years agoRevert 347365, its prerequisite 347364 got reverted.
Nico Weber [Wed, 21 Nov 2018 12:50:13 +0000 (12:50 +0000)]
Revert 347365, its prerequisite 347364 got reverted.

llvm-svn: 347391

5 years agoRevert 347366, its prerequisite 347364 got reverted.
Nico Weber [Wed, 21 Nov 2018 12:49:22 +0000 (12:49 +0000)]
Revert 347366, its prerequisite 347364 got reverted.

llvm-svn: 347390

5 years agoRevert r347364 again, the fix was incomplete.
Nico Weber [Wed, 21 Nov 2018 12:47:43 +0000 (12:47 +0000)]
Revert r347364 again, the fix was incomplete.

llvm-svn: 347389

5 years ago[ARM] Change REQUIRES: ARM to Requires: arm
Peter Smith [Wed, 21 Nov 2018 11:30:03 +0000 (11:30 +0000)]
[ARM] Change REQUIRES: ARM to Requires: arm

When REQUIRES: ARM is used the test is skipped as ARM is not recognized.
Change to REQUIRES: arm so that it is run. This required updating one of the
tests due to changes in expected output.

Differential Revision: https://reviews.llvm.org/D54786

llvm-svn: 347388

5 years ago[nios2] Add missing Nios2CodeGen -> Nios2AsmPrinter linkage
Michal Gorny [Wed, 21 Nov 2018 11:25:01 +0000 (11:25 +0000)]
[nios2] Add missing Nios2CodeGen -> Nios2AsmPrinter linkage

Add missing linkage from Nios2CodeGen library to Nios2AsmPrinter
library.  The missing dependency causes shared-lib build to fail with
the following reason:

  lib/Target/Nios2/CMakeFiles/LLVMNios2CodeGen.dir/Nios2AsmPrinter.cpp.o: In function `(anonymous namespace)::Nios2AsmPrinter::PrintAsmMemoryOperand(llvm::MachineInstr const*, unsigned int, unsigned int, char const*, llvm::raw_ostream&)':
  Nios2AsmPrinter.cpp:(.text._ZN12_GLOBAL__N_115Nios2AsmPrinter21PrintAsmMemoryOperandEPKN4llvm12MachineInstrEjjPKcRNS1_11raw_ostreamE+0x2b): undefined reference to `llvm::Nios2InstPrinter::getRegisterName(unsigned int)'
  lib/Target/Nios2/CMakeFiles/LLVMNios2CodeGen.dir/Nios2AsmPrinter.cpp.o: In function `(anonymous namespace)::Nios2AsmPrinter::PrintAsmOperand(llvm::MachineInstr const*, unsigned int, unsigned int, char const*, llvm::raw_ostream&)':
  Nios2AsmPrinter.cpp:(.text._ZN12_GLOBAL__N_115Nios2AsmPrinter15PrintAsmOperandEPKN4llvm12MachineInstrEjjPKcRNS1_11raw_ostreamE+0x97): undefined reference to `llvm::Nios2InstPrinter::getRegisterName(unsigned int)'
  collect2: error: ld returned 1 exit status

Differential Revision: https://reviews.llvm.org/D47810

llvm-svn: 347387

5 years ago[X86][AVX] Remove BROADCAST if we only need the 0'th element
Simon Pilgrim [Wed, 21 Nov 2018 11:00:09 +0000 (11:00 +0000)]
[X86][AVX] Remove BROADCAST if we only need the 0'th element

We don't catch this with target shuffle simplification if the src/dst types are different.

llvm-svn: 347386

5 years agoTest commit: Delete trailing space in comment
Nikita Popov [Wed, 21 Nov 2018 10:57:22 +0000 (10:57 +0000)]
Test commit: Delete trailing space in comment

llvm-svn: 347385

5 years ago[NFC] More complex tests for LoopSimplifyCFG
Max Kazantsev [Wed, 21 Nov 2018 09:55:09 +0000 (09:55 +0000)]
[NFC] More complex tests for LoopSimplifyCFG

llvm-svn: 347384

5 years agotsan: add pthread_tryjoin_np and pthread_timedjoin_np interceptors
Dmitry Vyukov [Wed, 21 Nov 2018 09:31:21 +0000 (09:31 +0000)]
tsan: add pthread_tryjoin_np and pthread_timedjoin_np interceptors

Add pthread_tryjoin_np() and pthread_timedjoin_np() interceptors on Linux,
so that ThreadSanitizer can handle programs using these functions.

Author: Yuri Per (yuri)
Reviewed in: https://reviews.llvm.org/D54521

llvm-svn: 347383

5 years agoAdd header <atomic> which is needed to compile with some older library versions.
Douglas Yung [Wed, 21 Nov 2018 08:54:40 +0000 (08:54 +0000)]
Add header <atomic> which is needed to compile with some older library versions.

llvm-svn: 347382

5 years ago[NFC] Add some sophisticated tests on LoopSimplifyCFG
Max Kazantsev [Wed, 21 Nov 2018 07:22:06 +0000 (07:22 +0000)]
[NFC] Add some sophisticated tests on LoopSimplifyCFG

llvm-svn: 347381

5 years ago[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and...
Craig Topper [Wed, 21 Nov 2018 07:01:22 +0000 (07:01 +0000)]
[X86] In getScalarMaskingNode, replace scalar_to_vector with a bitcast to v8i1 and an extract_subvector to convert i8 to v1i1.

The bitcast can be nicely merged with any i8 loads that exist for argument passing in 32 mode for example.

llvm-svn: 347380

5 years ago[LVI] run transfer function for binary operator even when the RHS isn't a constant
John Regehr [Wed, 21 Nov 2018 05:24:12 +0000 (05:24 +0000)]
[LVI] run transfer function for binary operator even when the RHS isn't a constant

LVI was symbolically executing binary operators only when the RHS was
constant, missing the case where we have a ConstantRange for the RHS,
but not an actual constant. Tested using check-all and by
bootstrapping. Compile time is not impacted measurably.

Differential Revision: https://reviews.llvm.org/D19859

llvm-svn: 347379

5 years ago[Driver] Link sanitizer runtime deps on Fuchsia when needed
Petr Hosek [Wed, 21 Nov 2018 04:33:44 +0000 (04:33 +0000)]
[Driver] Link sanitizer runtime deps on Fuchsia when needed

Even though these deps weren't needed, this makes Fuchsia driver
better match other drivers, and it may be necessary when trying to
use different C libraries on Fuchsia.

Differential Revision: https://reviews.llvm.org/D54741

llvm-svn: 347378

5 years ago[libc++] Implement P0487R1 - Fixing operator>>(basic_istream&, CharT*)
Zhihao Yuan [Wed, 21 Nov 2018 03:30:10 +0000 (03:30 +0000)]
[libc++] Implement P0487R1 - Fixing operator>>(basic_istream&, CharT*)

Summary:
Avoid buffer overflow by replacing the pointer interface with an array reference interface in C++2a.
Tentatively ready on Batavia2018.

 https://wg21.link/lwg2499
 https://wg21.link/p0487

Reviewers: mclow.lists, ldionne, EricWF

Reviewed By: ldionne

Subscribers: libcxx-commits, cfe-commits, christof

Differential Revision: https://reviews.llvm.org/D51268

llvm-svn: 347377

5 years ago[PowerPC] Do not use vectors to codegen bswap with Altivec turned off
Nemanja Ivanovic [Wed, 21 Nov 2018 02:53:50 +0000 (02:53 +0000)]
[PowerPC] Do not use vectors to codegen bswap with Altivec turned off

We have efficient codegen on P9 for lowering bswap that involves moving
the value into a vector reg and moving it back. However, the check under
which we custom lowered it did not adequately reflect the actual requirements.
It required only that the subtarget be an implementation of ISA 3.0 since all
compliant implementations have to provide the vector instructions.
However, the kernel builds have a valid use case for -mno-altivec -mcpu=pwr9
(i.e. don't emit vector code, don't have to save vector regs for context
switch). So we should require the correct features for this lowering.
Fixes https://bugs.llvm.org/show_bug.cgi?id=39334

llvm-svn: 347376

5 years ago[X86] Correct 256 vpmovzx/vpmovsx isel patterns to check HasAVX2 instead of HasAVX...
Craig Topper [Wed, 21 Nov 2018 01:39:38 +0000 (01:39 +0000)]
[X86] Correct 256 vpmovzx/vpmovsx isel patterns to check HasAVX2 instead of HasAVX to prevent fast-isel from using them incorrectly.

These are AVX2 instructions, but have been incorrectly marked in tablegen for a while. This wasn't a problem until r346784 switched the patterns to use target independent ISD opcodes. This made the patterns visible to fast isel.

Fixes PR39733

llvm-svn: 347375

5 years ago[X86] Add a copy of avx512-trunc.ll with -x86-experimental-vector-widening-legalizati...
Craig Topper [Wed, 21 Nov 2018 01:39:35 +0000 (01:39 +0000)]
[X86] Add a copy of avx512-trunc.ll with -x86-experimental-vector-widening-legalization enabled.

llvm-svn: 347374

5 years ago[clang-tidy] Add a test for proper handling of locations in scratch space.
Alexander Kornienko [Wed, 21 Nov 2018 01:11:18 +0000 (01:11 +0000)]
[clang-tidy] Add a test for proper handling of locations in scratch space.

This test examines the behavior change of clang::tooling::Diagnostic in r347372.

llvm-svn: 347373

5 years agoclang::tooling::Diagnostic: Don't store offset in the scratch space.
Alexander Kornienko [Wed, 21 Nov 2018 01:08:46 +0000 (01:08 +0000)]
clang::tooling::Diagnostic: Don't store offset in the scratch space.

These offsets are useless (and even harmful in certain cases) in exported
diagnostics. The test will be added to clang-tidy, since it's the main user of
the clang::tooling::Diagnostic class.

llvm-svn: 347372

5 years agoImplement YAML serialization of notes in clang::tooling::Diagnostic.
Alexander Kornienko [Wed, 21 Nov 2018 01:06:32 +0000 (01:06 +0000)]
Implement YAML serialization of notes in clang::tooling::Diagnostic.

llvm-svn: 347371

5 years ago[WebAssembly] Delete unused using statements (NFC)
Heejin Ahn [Wed, 21 Nov 2018 00:51:37 +0000 (00:51 +0000)]
[WebAssembly] Delete unused using statements (NFC)

Reviewers: sbc100, dschuff

Subscribers: mehdi_amini, jgravelle-google, sunfish, steven_wu, dexonsmith, llvm-commits

Differential Revision: https://reviews.llvm.org/D54683

llvm-svn: 347370

5 years ago[docs] Add C++ Performance Benchmark to test-suite proposals.
Michael Kruse [Wed, 21 Nov 2018 00:34:02 +0000 (00:34 +0000)]
[docs] Add C++ Performance Benchmark to test-suite proposals.

llvm-svn: 347369

5 years ago[XRay] Add a test for re-initialising FDR mode (NFC)
Dean Michael Berris [Wed, 21 Nov 2018 00:30:26 +0000 (00:30 +0000)]
[XRay] Add a test for re-initialising FDR mode (NFC)

This change adds an end-to-end test that ensures FDR mode can be
re-initialised safely in the face of multiple threads being traced.

llvm-svn: 347368

5 years ago[NFC] Rename lit feature to '-fsized-deallocation' for consistency
Louis Dionne [Wed, 21 Nov 2018 00:03:17 +0000 (00:03 +0000)]
[NFC] Rename lit feature to '-fsized-deallocation' for consistency

The '-faligned-allocation' flag uses a feature with the same name (with a
leading dash).

llvm-svn: 347367

5 years agoUpdate EvaluateAsInt to the new syntax.
Bill Wendling [Tue, 20 Nov 2018 23:24:39 +0000 (23:24 +0000)]
Update EvaluateAsInt to the new syntax.

llvm-svn: 347366

5 years agoUpdate call to EvaluateAsInt() to the new syntax.
Bill Wendling [Tue, 20 Nov 2018 23:24:18 +0000 (23:24 +0000)]
Update call to EvaluateAsInt() to the new syntax.

llvm-svn: 347365

5 years agoReinstate 347294 with a fix for the failures.
Bill Wendling [Tue, 20 Nov 2018 23:24:16 +0000 (23:24 +0000)]
Reinstate 347294 with a fix for the failures.

EvaluateAsInt() is sometimes called in a constant context. When that's the
case, we need to specify it as so.

llvm-svn: 347364

5 years ago[NFC] Reformat availability #defines in __config
Louis Dionne [Tue, 20 Nov 2018 23:18:25 +0000 (23:18 +0000)]
[NFC] Reformat availability #defines in __config

Aligning everything makes what we're doing more obvious.

llvm-svn: 347363

5 years ago[NFC] Fix formatting in availability documentation
Louis Dionne [Tue, 20 Nov 2018 23:07:01 +0000 (23:07 +0000)]
[NFC] Fix formatting in availability documentation

llvm-svn: 347362

5 years ago[X86] Emit a PACKUS instead of a VECTOR_SHUFFLE from LowerTRUNCATE for v16i16->v16i8.
Craig Topper [Tue, 20 Nov 2018 22:57:48 +0000 (22:57 +0000)]
[X86] Emit a PACKUS instead of a VECTOR_SHUFFLE from LowerTRUNCATE for v16i16->v16i8.

We can't guarantee that demanded bits passing through the vector shuffle won't cause the AND in front of this to be removed. This would prevent the PACKUS from being matched during shuffle lowering.

Unfortunately, this adds a packuswb to one of the vector-reduce-mul.ll tests since we were removing the shuffle via SimplifyDemandedVectorElts. We appear to have similar issues with vpmovwb on the same test case on other targets.

llvm-svn: 347361

5 years agoA couple of tests were broken when clang implemented the compiler parts of P0482...
Marshall Clow [Tue, 20 Nov 2018 22:55:40 +0000 (22:55 +0000)]
A couple of tests were broken when clang implemented the compiler parts of P0482 (support for char8_t). Comment out those bits until we implement the corresponding bits in libc++

llvm-svn: 347360

5 years agoFix pointer options mask. It was off by 1 bit.
Zachary Turner [Tue, 20 Nov 2018 22:53:40 +0000 (22:53 +0000)]
Fix pointer options mask.  It was off by 1 bit.

llvm-svn: 347359

5 years agoRevert "[Sanitizer] intercept setvbuf on other platforms where it is supported"
David Carlier [Tue, 20 Nov 2018 22:50:31 +0000 (22:50 +0000)]
Revert "[Sanitizer] intercept setvbuf on other platforms where it is supported"

llvm-svn: 347358