platform/upstream/llvm.git
2 years ago[mlir][tensor] Extend reshape utils.
Stephan Herhut [Mon, 14 Feb 2022 16:46:35 +0000 (17:46 +0100)]
[mlir][tensor] Extend reshape utils.

This change changes the handling of trailing dimensions with unknown
extent. Users of the changessociationIndicesForReshape helper should
see benefits when transforming reshape like operations into
expand/collapse pairs if the higher-rank type has trailing unknown
dimensions.

The motivating example is a reshape from tensor<16x1x?xi32> to
tensor<16xi32> that can be modeled as collapsing the three dimensions.

Differential Revision: https://reviews.llvm.org/D119730

2 years ago[NFC][Clang/test] add target triple to CodeGen/analyze_noundef.cpp
hyeongyukim [Fri, 18 Feb 2022 08:52:56 +0000 (17:52 +0900)]
[NFC][Clang/test] add target triple to CodeGen/analyze_noundef.cpp

2 years ago[asan] Add support for disable_sanitizer_instrumentation attribute
Alexander Potapenko [Thu, 17 Feb 2022 14:48:35 +0000 (15:48 +0100)]
[asan] Add support for disable_sanitizer_instrumentation attribute

For ASan this will effectively serve as a synonym for
__attribute__((no_sanitize("address"))).

Adding the disable_sanitizer_instrumentation to functions will drop the
sanitize_XXX attributes on the IR level.

This is the third reland of https://reviews.llvm.org/D114421.
Now that TSan test is fixed (https://reviews.llvm.org/D120050) there
should be no deadlocks.

Differential Revision: https://reviews.llvm.org/D120055

2 years ago[NFC] Fix a buildbot failure after b529744
hyeongyukim [Fri, 18 Feb 2022 08:38:50 +0000 (17:38 +0900)]
[NFC] Fix a buildbot failure after b529744

2 years ago[mlir] Propagate NaNs in PolynomialApproximation
Tres Popp [Thu, 17 Feb 2022 09:20:50 +0000 (10:20 +0100)]
[mlir] Propagate NaNs in PolynomialApproximation

Previously, NaNs would be dropped in favor of bounded values which was
strictly incorrect. Now the min/max operation propagate this
information. Not all uses of min/max need this, but the given change
will help protect future additions, and this prevents the need for an
additional cmpf and select operation to handle NaNs.

Differential Revision: https://reviews.llvm.org/D120020

2 years ago[Clang] Rename `disable-noundef-analysis` flag to `-[no-]enable-noundef-analysis`
hyeongyukim [Fri, 18 Feb 2022 08:00:00 +0000 (17:00 +0900)]
[Clang] Rename `disable-noundef-analysis` flag to `-[no-]enable-noundef-analysis`

This flag was previously renamed `enable_noundef_analysis` to
`disable-noundef-analysis,` which is not a conventional name. (Driver and
CC1's boolean options are using [no-] prefix)
As discussed at https://reviews.llvm.org/D105169, this patch reverts its
name to `[no-]enable_noundef_analysis` and enables noundef-analysis as
default.

Reviewed By: MaskRay

Differential Revision: https://reviews.llvm.org/D119998

2 years agoFix buildbots after https://reviews.llvm.org/D119797
Greg Clayton [Fri, 18 Feb 2022 07:59:15 +0000 (23:59 -0800)]
Fix buildbots after https://reviews.llvm.org/D119797

This value error is no longer needed with the new version of the patch

2 years ago[flang] Lower scalar negation
Valentin Clement [Fri, 18 Feb 2022 07:08:34 +0000 (08:08 +0100)]
[flang] Lower scalar negation

Handle negation on scalar expression.

```
res = -a
```

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D120071

Co-authored-by: Jean Perier <jperier@nvidia.com>
Co-authored-by: Eric Schweitz <eschweitz@nvidia.com>
2 years ago[gn build] Port 0a4184909a8c
LLVM GN Syncbot [Fri, 18 Feb 2022 06:12:10 +0000 (06:12 +0000)]
[gn build] Port 0a4184909a8c

2 years agoReland "[memprof] Extend the index prof format to include memory profiles."
Snehasish Kumar [Fri, 18 Feb 2022 00:01:31 +0000 (16:01 -0800)]
Reland "[memprof] Extend the index prof format to include memory profiles."

This patch adds support for optional memory profile information to be
included with and indexed profile. The indexed profile header adds a new
field which points to the offset of the memory profile section (if
present) in the indexed profile. For users who do not utilize this
feature the only overhead is a 64-bit offset in the header.

The memory profile section contains (1) profile metadata describing the
information recorded for each entry (2) an on-disk hashtable containing
the profile records indexed via llvm::md5(function_name). We chose to
introduce a separate hash table instead of the existing one since the
indexing for the instrumented fdo hash table is based on a CFG hash
which itself is perturbed by memprof instrumentation.

This commit also includes the changes reviewed separately in D120093.

Differential Revision: https://reviews.llvm.org/D120103

2 years ago[JITLink][RISCV] fix the extractBits behavior and add R_RISCV_JAL relocation.
fourdim [Fri, 18 Feb 2022 06:06:48 +0000 (14:06 +0800)]
[JITLink][RISCV] fix the extractBits behavior and add R_RISCV_JAL relocation.

This patch supports the R_RISCV_JAL relocation.
Moreover, it will fix the extractBits function's behavior as it extracts Size + 1 bits.
In the test ELF_jal.s:
Before:
```
Hi: 4294836480
extractBits(Hi, 12, 8): 480
```
After:
```
Hi: 4294836480
extractBits(Hi, 12, 8): 224
```

Reviewed By: StephenFan

Differential Revision: https://reviews.llvm.org/D117975

2 years ago[MemoryDependency] Simplfy re-ordering condition. Cleanup. NFC.
Serguei Katkov [Thu, 17 Feb 2022 05:06:23 +0000 (12:06 +0700)]
[MemoryDependency] Simplfy re-ordering condition. Cleanup. NFC.

Make the reading of condition for restricting re-ordering simpler.

Reviewers: reames
Reviewed By: reames
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D120005

2 years ago[XCOFF][llvm-objdump] change the priority of symbols with
esmeyi [Fri, 18 Feb 2022 05:29:10 +0000 (00:29 -0500)]
[XCOFF][llvm-objdump] change the priority of symbols with
                      the same address by symbol types.

Summary: In XCOFF, each section comes with a default symbol
         with the same name as the section. It doesn't bind
         to code locations and it may cause incorrect display
         of symbol names under `llvm-objdump -d`.
         This patch changes the priority of symbols with the
         same address by symbol type.

Reviewed By: jhenderson, shchenz

Differential Revision: https://reviews.llvm.org/D117642

2 years ago[gn build] Port 8e979460bb27
LLVM GN Syncbot [Fri, 18 Feb 2022 04:18:16 +0000 (04:18 +0000)]
[gn build] Port 8e979460bb27

2 years ago[libc++][ranges] Implement `std::sortable`.
Konstantin Varlamov [Fri, 18 Feb 2022 04:15:02 +0000 (20:15 -0800)]
[libc++][ranges] Implement `std::sortable`.

Differential Revision: https://reviews.llvm.org/D119619

2 years ago[gn build] Port eea3d90af181
LLVM GN Syncbot [Fri, 18 Feb 2022 04:12:48 +0000 (04:12 +0000)]
[gn build] Port eea3d90af181

2 years ago[libc++][ranges] Implement `std::mergeable`.
Konstantin Varlamov [Fri, 18 Feb 2022 04:11:46 +0000 (20:11 -0800)]
[libc++][ranges] Implement `std::mergeable`.

Differential Revision: https://reviews.llvm.org/D119489

2 years ago[GlobalDCE] [VFE] Avoid dropping vfunc dependencies when an invalid vtable entry...
Kuba Mracek [Fri, 18 Feb 2022 03:41:46 +0000 (19:41 -0800)]
[GlobalDCE] [VFE] Avoid dropping vfunc dependencies when an invalid vtable entry is present

When we scan vtables for a particular vload in ScanVTableLoad and an entry in
one possible vtable is invalid (null or non-fptr), we bail in a wrong way -- we
completely stop the scanning of vtables and this results in dropped dependencies
and incorrectly removed vfuncs from vtables. Let's fix that by correcting the
bailing logic to keep iterating and only skip the invalid entries.

Differential Revision: https://reviews.llvm.org/D120006

2 years ago[NFC] Add myself to CREDITS.TXT
Weining Lu [Fri, 18 Feb 2022 02:18:13 +0000 (10:18 +0800)]
[NFC] Add myself to CREDITS.TXT

2 years ago[MachineOutliner] Add testcase for instruction mapping stats
Jessica Paquette [Fri, 18 Feb 2022 02:26:16 +0000 (18:26 -0800)]
[MachineOutliner] Add testcase for instruction mapping stats

I forgot to attach the testcase for 12389e375811d46ce41d949857f8b469d6563114!

2 years ago[MachineOutliner] Add statistics for unsigned vector size
Jessica Paquette [Fri, 18 Feb 2022 02:12:39 +0000 (18:12 -0800)]
[MachineOutliner] Add statistics for unsigned vector size

Useful for debugging + evaluating improvements to the outliner.

Stats are the number of illegal, legal, and invisible instructions in the
unsigned vector, and it's total length.

2 years ago[msan] strsignal interceptor
Vitaly Buka [Thu, 17 Feb 2022 21:06:11 +0000 (13:06 -0800)]
[msan] strsignal interceptor

Reviewed By: kstoimenov

Differential Revision: https://reviews.llvm.org/D120082

2 years ago[LoongArch] Fix several instruction definition errors in initial patches
Weining Lu [Fri, 18 Feb 2022 02:04:28 +0000 (10:04 +0800)]
[LoongArch] Fix several instruction definition errors in initial patches

This patch corrects some instruction definitions that I incorrectly wrote
in initial patches including bstr{ins/pick}.{w/d}, ll.{w/d} and sc.{w/d}.

Reviewed By: xen0n

Differential Revision: https://reviews.llvm.org/D119817

2 years ago[LoongArch] Add missing dollar prefix to register name in InstPrinter
Weining Lu [Thu, 17 Feb 2022 01:04:47 +0000 (09:04 +0800)]
[LoongArch] Add missing dollar prefix to register name in InstPrinter

This patch adds a '$' prefix to register name in InstPrinter that I missed in initial patches.

Reviewed By: xen0n

Differential Revision: https://reviews.llvm.org/D119813

2 years ago[LICM][PhaseOrder] Don't speculate in LICM until after running loop rotate
William S. Moses [Sat, 12 Feb 2022 21:33:22 +0000 (16:33 -0500)]
[LICM][PhaseOrder] Don't speculate in LICM until after running loop rotate

LICM will speculatively hoist code outside of loops. This requires removing information, like alias analysis (https://github.com/llvm/llvm-project/issues/53794), range information (https://bugs.llvm.org/show_bug.cgi?id=50550), among others. Prior to https://reviews.llvm.org/D99249 , LICM would only be run after LoopRotate. Running Loop Rotate prior to LICM prevents a instruction hoist from being speculative, if it was conditionally executed by the iteration (as is commonly emitted by clang and other frontends). Adding the additional LICM pass first, however, forces all of these instructions to be considered speculative, even if they are not speculative after LoopRotate. This destroys information, resulting in performance losses for discarding this additional information.

This PR modifies LICM to accept a ``speculative'' parameter which allows LICM to be set to perform information-loss speculative hoists or not. Phase ordering is then modified to not perform the information-losing speculative hoists until after loop rotate is performed, preserving this additional information.

Reviewed By: lebedev.ri

Differential Revision: https://reviews.llvm.org/D119965

2 years ago[mlir][Vector] Switch ShuffleOp to the declarative assembly format
Benjamin Kramer [Fri, 18 Feb 2022 00:35:25 +0000 (01:35 +0100)]
[mlir][Vector] Switch ShuffleOp to the declarative assembly format

This also requires implementing return type deduction.

2 years ago[mlir][Vector] Add InferTypeOpInterface to vector dialect and remove the now redundan...
Benjamin Kramer [Fri, 18 Feb 2022 00:15:14 +0000 (01:15 +0100)]
[mlir][Vector] Add InferTypeOpInterface to vector dialect and remove the now redundant builders

2 years ago[RISCV][NFC] Add some tail agnostic tests for nomask operations.
Zakk Chen [Thu, 17 Feb 2022 04:49:25 +0000 (20:49 -0800)]
[RISCV][NFC] Add some tail agnostic tests for nomask operations.

Improve test coverage for tail agnostic nomask vslidedown/up, vmv.s.x
vfmv.s.f and vcompress.

Reviewed By: rogfer01

Differential Revision: https://reviews.llvm.org/D120008

2 years ago[RISCV][NFC] Add tail agnostic tests for nomask Vector Reduction IR intrinsics.
Zakk Chen [Wed, 16 Feb 2022 08:38:51 +0000 (00:38 -0800)]
[RISCV][NFC] Add tail agnostic tests for nomask Vector Reduction IR intrinsics.

Improve test coverage for tail agnostic nomask Vector Reduction IR.

Reviewed By: monkchiang

Differential Revision: https://reviews.llvm.org/D119920

2 years ago[WebAssembly] Make EH/SjLj vars unconditionally thread local
Heejin Ahn [Thu, 17 Feb 2022 00:11:29 +0000 (16:11 -0800)]
[WebAssembly] Make EH/SjLj vars unconditionally thread local

This makes three thread local variables (`__THREW__`, `__threwValue`,
and `__wasm_lpad_context`) unconditionally thread local. If the target
doesn't support TLS, they will be downgraded to normal variables in
`stripThreadLocals`. This makes the object not linkable with other
objects using shared memory, which is what we intend here; these
variables should be thread local when used with shared memory. This is
what we initially tried in D88262.

But D88323 changed this: It only created these variables when threads
were supported, because `__THREW__` and `__threwValue` were always
generated even if Emscripten EH/SjLj was not used, making all objects
built without threads not linkable with shared memory, which was too
restrictive. But sometimes this is not safe. If we build an object using
variables such as `__THREW__` without threads, it can be linked to other
objects using shared memory, because the original object's `__THREW__`
was not created thread local to begin with.

So this CL basically reverts D88323 with some additional improvements:
- This checks each of the functions and global variables created within
  `LowerEmscriptenEHSjLj` pass and removes it if it's not used at the
  end of the pass. So only modules using those variables will be
  affected.
- Moves `CoalesceFeaturesAndStripAtomics` and `AtomicExpand` passes
  after all other IR pasess that can create thread local variables. It
  is not sufficient to move them to the end of `addIRPasses`, because
  `__wasm_lpad_context` is created in `WasmEHPrepare`, which runs inside
  `addPassesToHandleExceptions`, which runs before `addISelPrepare`. So
  we override `addISelPrepare` and move atomic/TLS stripping and
  expanding passes there.

This also removes merges `TLS` and `NO-TLS` FileCheck lines into one
`CHECK` line, because in the bitcode level we always create them as
thread local. Also some function declarations are deleted `CHECK` lines
because they are unused.

Reviewed By: tlively, sbc100

Differential Revision: https://reviews.llvm.org/D120013

2 years ago[gn build] Port 19bdf44d8508
LLVM GN Syncbot [Thu, 17 Feb 2022 23:56:17 +0000 (23:56 +0000)]
[gn build] Port 19bdf44d8508

2 years agoRevert "Reland "[memprof] Extend the index prof format to include memory profiles.""
Snehasish Kumar [Thu, 17 Feb 2022 23:45:10 +0000 (15:45 -0800)]
Revert "Reland "[memprof] Extend the index prof format to include memory profiles.""

This reverts commit 807ba7aace188ada83ddb4477265728e97346af1.

2 years agoRevert "[memprof] Fix frame deserialization on big endian systems."
Snehasish Kumar [Thu, 17 Feb 2022 23:42:02 +0000 (15:42 -0800)]
Revert "[memprof] Fix frame deserialization on big endian systems."

This reverts commit c74389b4b58d8db3f8262ce15b9d514d62fe265c.

This broke the ml-opt-x86-64 build.
https://lab.llvm.org/buildbot#builders/9/builds/4127

2 years agoAMDGPU: Try to fix expensive_checks bot tests failures
Matt Arsenault [Thu, 17 Feb 2022 23:42:06 +0000 (18:42 -0500)]
AMDGPU: Try to fix expensive_checks bot tests failures

2 years ago[memprof] Fix frame deserialization on big endian systems.
Snehasish Kumar [Thu, 17 Feb 2022 23:14:47 +0000 (15:14 -0800)]
[memprof] Fix frame deserialization on big endian systems.

We write the memprof internal call frame data in little endian format.
However when reading the frame information we were casting it directly
to a MemProfRecord::Frame pointer. In this change we add a separate
deserialization method which uses an endian reader to read the bytes as
little endian.

This fixes https://lab.llvm.org/buildbot/#/builders/100/builds/12940

Differential Revision: https://reviews.llvm.org/D120093

2 years agoRegAllocGreedy: Fix last chance recolor assert in impossible case
Matt Arsenault [Thu, 17 Feb 2022 21:52:02 +0000 (16:52 -0500)]
RegAllocGreedy: Fix last chance recolor assert in impossible case

This example is not compilable without handling eviction of specific
subregisters. Last chance recoloring was deciding it could try
evicting an overlapping superregister, which doesn't help make any
progress. The LiveIntervalUnion would then assert due to an
overlapping / identical range when trying the new assignment.

Unfortunately this is also producing a verifier error after the
allocation fails. I've seen a number of these, and not sure if we
should just start deleting the function on error rather than trying to
figure out how to put together valid MIR.

I'm not super confident this is the right place to fix this. I also
have a number of failing testcases I need to fix by handling partial
evictions of superregisters.

2 years agoFix typo in Swift.def, it is swift5_protocol_conformances not swift5_protocol_confrom...
Shubham Sandeep Rastogi [Thu, 17 Feb 2022 22:59:50 +0000 (14:59 -0800)]
Fix typo in Swift.def, it is swift5_protocol_conformances not swift5_protocol_confromances

2 years ago[clang] Remove Address::deprecated() in emitVoidPtrDirectVAArg()
Arthur Eubanks [Thu, 17 Feb 2022 23:00:43 +0000 (15:00 -0800)]
[clang] Remove Address::deprecated() in emitVoidPtrDirectVAArg()

2 years ago[mlir][sparse] remove unused test matrix file
Aart Bik [Thu, 17 Feb 2022 17:43:10 +0000 (09:43 -0800)]
[mlir][sparse] remove unused test matrix file

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D120069

2 years agoRevert "[CUDA][SPIRV] Assign global address space to CUDA kernel arguments"
Matthew Voss [Thu, 17 Feb 2022 22:18:54 +0000 (14:18 -0800)]
Revert "[CUDA][SPIRV] Assign global address space to CUDA kernel arguments"

This reverts commit 9de4fc0f2d3b60542956f7e5254951d049edeb1f.

Reverting due to test failure: https://lab.llvm.org/buildbot/#/builders/139/builds/17199

2 years ago[LLDB] Adding skipif for arm linux for TestStructuredBinding.py
Shafik Yaghmour [Thu, 17 Feb 2022 22:30:18 +0000 (14:30 -0800)]
[LLDB] Adding skipif for arm linux for TestStructuredBinding.py

2 years ago[clang] Remove Address::deprecated() in CGCXXABI.h
Arthur Eubanks [Thu, 17 Feb 2022 22:22:39 +0000 (14:22 -0800)]
[clang] Remove Address::deprecated() in CGCXXABI.h

2 years ago[M68k] Adopt VarLenCodeEmitter for bits instructions
Min-Yih Hsu [Thu, 17 Feb 2022 22:13:52 +0000 (14:13 -0800)]
[M68k] Adopt VarLenCodeEmitter for bits instructions

And introduce operand encoding fragments (i.e. MxEncMemOp record) for
addressing modes 'o' and 'e'.

2 years ago[libcxx][test] Let the library indicate support for int128
Casey Carter [Thu, 17 Feb 2022 20:23:58 +0000 (12:23 -0800)]
[libcxx][test] Let the library indicate support for int128

Define `TEST_HAS_NO_INT128` accordingly.

Differential Revision: https://reviews.llvm.org/D120010

2 years ago[libcxx] [test] Use proper UTF-8 locales on Windows
Martin Storsjö [Tue, 25 Jan 2022 13:58:43 +0000 (13:58 +0000)]
[libcxx] [test] Use proper UTF-8 locales on Windows

Since Windows 10 version 1803 (10.0.17134.0) (or Windows Server 2019),
the Windows Universal C Runtime (UCRT) actually does support UTF-8
locales - they're available e.g. via the same names as commonly on Unices,
e.g. "en_US.UTF-8".

The UTF-8 locale support unfortunately has a bug which breaks a couple
tests that were passing previously. That bug is fixed in the very
latest version of the UCRT (in UCRT 10.0.20348.0, available in Windows
11 or Windows Server 2022), so it will get resolved at some point
eventually, provided that the CI environment does get upgraded to a
newer version of Windows Server.

While the net number of xfailed/passing tests in this patch is a loss,
this does allow fixing a lot more locale tests properly for Windows
in later patches.

Intentionally not touching the ISO-8859-1/2 locales used for testing;
they're not detected and tested/used right now, and fixing that up
is another project.

Differential Revision: https://reviews.llvm.org/D119930

2 years ago[NFC][Flang] Add colon to CHECK-LABEL to exercise the check
Kiran Chandramohan [Thu, 17 Feb 2022 21:55:07 +0000 (21:55 +0000)]
[NFC][Flang] Add colon to CHECK-LABEL to exercise the check

Reviewed By: clementval

Differential Revision: https://reviews.llvm.org/D119995

2 years ago[clang] Remove Address::deprecated() call in CGExprCXX.cpp
Arthur Eubanks [Thu, 17 Feb 2022 21:57:46 +0000 (13:57 -0800)]
[clang] Remove Address::deprecated() call in CGExprCXX.cpp

2 years agoRemove __uncvref; use __uncvref_t instead
Nikolas Klauser [Thu, 17 Feb 2022 21:53:20 +0000 (22:53 +0100)]
Remove __uncvref; use __uncvref_t instead

Reviewed By: Quuxplusone, #libc

Spies: libcxx-commits

Differential Revision: https://reviews.llvm.org/D119958

2 years ago[MTE] Instrument use-after-scope for optnone functions.
Florian Mayer [Thu, 17 Feb 2022 00:32:09 +0000 (16:32 -0800)]
[MTE] Instrument use-after-scope for optnone functions.

We always need the Dominator and PostDominatorTree for use-after-scope.

Reviewed By: eugenis

Differential Revision: https://reviews.llvm.org/D119993

2 years ago[clang] Pass more flags to ld64.lld
Nico Weber [Sat, 12 Feb 2022 03:13:28 +0000 (22:13 -0500)]
[clang] Pass more flags to ld64.lld

* ld64.lld now completely supports -export_dynamic (D119372), so map -rdynamic
  to -export_dynamic like already done for ld64

* ld64.lld has been supporting -object_path_lto for well over a year (D92537),
  so pass it like already done for ld64

Differential Revision: https://reviews.llvm.org/D119612

2 years ago[gn build] Port 807ba7aace18
LLVM GN Syncbot [Thu, 17 Feb 2022 21:41:09 +0000 (21:41 +0000)]
[gn build] Port 807ba7aace18

2 years ago[MLIR][GPU] Remove call to cudaSetDevice(), which no longer exists
Krzysztof Drewniak [Thu, 17 Feb 2022 21:37:15 +0000 (21:37 +0000)]
[MLIR][GPU] Remove call to cudaSetDevice(), which no longer exists

Differential Revision: https://reviews.llvm.org/D120085

2 years ago[ASan] Fix TLS teardown.
Mitch Phillips [Thu, 17 Feb 2022 21:27:13 +0000 (13:27 -0800)]
[ASan] Fix TLS teardown.

TLS teardown is currently broken, as we unpoison the shadow a little bit
and to the right of the TLS section, rather than the full TLS section
itself. This currently breaks at -O0, and breaks with some upcoming
globals code that I have.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D120080

2 years agounbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 32b73bc6ab82
Nico Weber [Thu, 17 Feb 2022 21:33:24 +0000 (16:33 -0500)]
unbreak Modules/cxx20-export-import.cpp with LLVM_APPEND_VC_REV after 32b73bc6ab82

See revision b8b7a9dcdcbc for prior art.

2 years ago[MLIR][GPU] Add gpu.set_default_device op
Krzysztof Drewniak [Tue, 15 Feb 2022 20:23:44 +0000 (20:23 +0000)]
[MLIR][GPU] Add gpu.set_default_device op

This op is added to allow MLIR code running on multi-GPU systems to
select the GPU they want to execute operations on when no GPU is
otherwise specified.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D119883

2 years agoFix race condition when launching and attaching.
Greg Clayton [Tue, 15 Feb 2022 01:05:03 +0000 (17:05 -0800)]
Fix race condition when launching and attaching.

We discovered that when using "launchCommands" or "attachCommands" that there was an issue where these commands were not being run synchronously. There were further problems in this case where we would get thread events for the process that was just launched or attached before the IDE was ready, which is after "configurationDone" was sent to lldb-vscode.

This fix introduces the ability to wait for the process to stop after the run or attach to ensure that we have a stopped process at the entry point that is ready for the debug session to proceed. This also allows us to run the normal launch or attach without needing to play with the async flag the debugger. We spin up the thread that listens for process events before we start the launch or attach, but we stop the first eStateStopped (with stop ID of zero) event from being delivered through the DAP protocol because the "configurationDone" request handler will deliver it manually as the IDE expects a stop after configuration done. The request_configurationDone will also only deliver the stop packet if the "stopOnEntry" is False in the launch configuration.

Also added a new "timeout" to the launch and attach launch configuration arguments that can be set and defaults to 30 seconds. Since we now poll to detect when the process is stopped, we need a timeout that can be changed in case certain workflows take longer that 30 seconds to attach. If the process is not stopped by the timeout, an error will be retured for the launch or attach.

Added a flag to the vscode.py protocol classes that detects and ensures that no "stopped" events are sent prior to "configurationDone" has been sent and will raise an error if it does happen.

This should make our launching and attaching more reliable and avoid some deadlocks that were being seen (https://reviews.llvm.org/D119548).

Differential Revision: https://reviews.llvm.org/D119797

2 years agoReland "[memprof] Extend the index prof format to include memory profiles."
Snehasish Kumar [Mon, 14 Feb 2022 19:52:42 +0000 (11:52 -0800)]
Reland "[memprof] Extend the index prof format to include memory profiles."

This reverts commit 85355a560a33897453df2ef959e255ee725eebce.

This patch adds support for optional memory profile information to be
included with and indexed profile. The indexed profile header adds a new
field which points to the offset of the memory profile section (if
present) in the indexed profile. For users who do not utilize this
feature the only overhead is a 64-bit offset in the header.

The memory profile section contains (1) profile metadata describing the
information recorded for each entry (2) an on-disk hashtable containing
the profile records indexed via llvm::md5(function_name). We chose to
introduce a separate hash table instead of the existing one since the
indexing for the instrumented fdo hash table is based on a CFG hash
which itself is perturbed by memprof instrumentation.

Differential Revision: https://reviews.llvm.org/D118653

2 years ago[clang-format] Do not add space after return-like keywords in macros.
Marek Kurdej [Thu, 17 Feb 2022 10:10:41 +0000 (11:10 +0100)]
[clang-format] Do not add space after return-like keywords in macros.

Fixes https://github.com/llvm/llvm-project/issues/33336.

Reviewed By: HazardyKnusperkeks, owenpan

Differential Revision: https://reviews.llvm.org/D120028

2 years ago[libc++][CI] Upload ABI lists for all jobs
Louis Dionne [Thu, 17 Feb 2022 15:00:29 +0000 (10:00 -0500)]
[libc++][CI] Upload ABI lists for all jobs

Some jobs might not produce those, but it makes the blocks easier to
copy-paste and makes sure that if a job does produce an ABI list, it
will be updloaded in the artifacts.

Differential Revision: https://reviews.llvm.org/D120056

2 years ago[libc++] Implement P1165R1 (Make stateful allocator propagation more consistent)
Nikolas Klauser [Fri, 11 Feb 2022 18:24:31 +0000 (19:24 +0100)]
[libc++] Implement P1165R1 (Make stateful allocator propagation more consistent)

Reviewed By: Quuxplusone, ldionne, #libc

Spies: libcxx-commits

Differential Revision: https://reviews.llvm.org/D119112

2 years ago[libcxx][test] Silence signed/unsigned comparison warnings
Casey Carter [Thu, 17 Feb 2022 20:55:51 +0000 (12:55 -0800)]
[libcxx][test] Silence signed/unsigned comparison warnings

2 years ago[clang][SemaTemplate] Fix a stack use after scope
Kadir Cetinkaya [Thu, 17 Feb 2022 16:18:43 +0000 (17:18 +0100)]
[clang][SemaTemplate] Fix a stack use after scope

Differential Revision: https://reviews.llvm.org/D120065

2 years ago[mlir][memref] Switch ViewOp to the declarative assembly format
Benjamin Kramer [Thu, 17 Feb 2022 20:16:51 +0000 (21:16 +0100)]
[mlir][memref] Switch ViewOp to the declarative assembly format

2 years agoUse functions with prototypes when appropriate; NFC
Aaron Ballman [Thu, 17 Feb 2022 20:32:15 +0000 (15:32 -0500)]
Use functions with prototypes when appropriate; NFC

Now that the AST printer properly handles functions with no parameters
in C code, all of the tests relying on AST printing can be updated to
use prototypes where appropriate.

2 years ago[mlir] NFC NamedAttrList append with StringAttr
Mogball [Thu, 17 Feb 2022 20:29:25 +0000 (20:29 +0000)]
[mlir] NFC NamedAttrList append with StringAttr

NamedAttrList.append(StringAttr, StringAttr) fails to compile because it is
matched to the IteratorT append. Fixes the method to only match if the type is
an iterator.

2 years ago[ELF] Support (TYPE=<value>) to customize the output section type
Fangrui Song [Thu, 17 Feb 2022 20:10:58 +0000 (12:10 -0800)]
[ELF] Support (TYPE=<value>) to customize the output section type

The current output section type allows to set the ELF section type to
SHT_PROGBITS or SHT_NOLOAD. This patch allows an arbitrary section value
to be specified. Some common SHT_* literal names are supported as well.

```
SECTIONS {
  note (TYPE=SHT_NOTE) : { BYTE(8) *(note) }
  init_array ( TYPE=14 ) : { QUAD(14) }
  fini_array (TYPE = SHT_FINI_ARRAY) : { QUAD(15) }
}
```

When `sh_type` is specified, it is an error if an input section has a different type.

Our syntax is compatible with GNU ld 2.39 (https://sourceware.org/bugzilla/show_bug.cgi?id=28841).

Reviewed By: peter.smith

Differential Revision: https://reviews.llvm.org/D118840

2 years ago{instsimplify] Precommit some tests for provable inequal pointers derived from allocas
Philip Reames [Thu, 17 Feb 2022 20:00:44 +0000 (12:00 -0800)]
{instsimplify] Precommit some tests for provable inequal pointers derived from allocas

2 years ago[EarlyCSE][OpaquePtr] Check access type when performing DSE
Arthur Eubanks [Thu, 17 Feb 2022 19:18:26 +0000 (11:18 -0800)]
[EarlyCSE][OpaquePtr] Check access type when performing DSE

This will bail out on target specific intrinsics. If those are deemed
important enough for EarlyCSE to handle, we can augment MemIntrinsicInfo
with an access type for TargetTransformInfo::getTgtMemIntrinsic() to
handle.

Reviewed By: #opaque-pointers, nikic

Differential Revision: https://reviews.llvm.org/D120077

2 years ago[lld] Make error handling functions opaque
Fangrui Song [Thu, 17 Feb 2022 19:54:57 +0000 (11:54 -0800)]
[lld] Make error handling functions opaque

The inline `lld::error` expands to two function calls `errorHandler` and `error`
where the latter is opaque. Move the functions to .cpp files to decrease code
size.

My x86-64 lld executable is 9KiB smaller.

Reviewed By: #lld-macho, thakis

Differential Revision: https://reviews.llvm.org/D120002

2 years agoReland "[InstrProf] Make the IndexedInstrProf header backwards compatible."
Snehasish Kumar [Mon, 14 Feb 2022 19:52:40 +0000 (11:52 -0800)]
Reland "[InstrProf] Make the IndexedInstrProf header backwards compatible."

This reverts commit 9fd2cb21fb3f763fc784eab198bf1297a24596fa.

Fixes an issue on big endian systems where the format version
was not converted to little endian prior to passing to GET_VERSION.

Differential Revision: https://reviews.llvm.org/D118390

2 years agoAST: Move __va_list tag back to std conditionally on AArch64.
Peter Collingbourne [Thu, 6 Jan 2022 21:37:21 +0000 (13:37 -0800)]
AST: Move __va_list tag back to std conditionally on AArch64.

In post-commit feedback on D104830 Jessica Clarke pointed out that
unconditionally adding __va_list to the std namespace caused namespace
debug info to be emitted in C, which is not only inappropriate but
turned out to confuse the dtrace tool. Therefore, move __va_list back
to std only in C++ so that the correct debug info is generated. We
also considered moving __va_list to the top level unconditionally
but this would contradict the specification and be visible to AST
matchers and such, so make it conditional on the language mode.

To avoid breaking name mangling for __va_list, teach the Itanium
name mangler to always mangle it as if it were in the std namespace
when targeting ARM architectures. This logic is not needed for the
Microsoft name mangler because Microsoft platforms define va_list as
a typedef of char *.

Depends on D116773

Differential Revision: https://reviews.llvm.org/D116774

2 years agoAST: Make getEffectiveDeclContext() a member function of ItaniumMangleContextImpl...
Peter Collingbourne [Thu, 17 Feb 2022 19:23:33 +0000 (11:23 -0800)]
AST: Make getEffectiveDeclContext() a member function of ItaniumMangleContextImpl. NFCI.

In an upcoming change we are going to need to access mangler state
from the getEffectiveDeclContext() function. Therefore, make it a
member function of ItaniumMangleContextImpl. Any callers that are
not currently members of ItaniumMangleContextImpl or CXXNameMangler
are made members of one or the other depending on where they are
called from.

Differential Revision: https://reviews.llvm.org/D116773

2 years ago[OpenMP] Add RTL function to externalization RAII
Joseph Huber [Thu, 17 Feb 2022 18:20:51 +0000 (13:20 -0500)]
[OpenMP] Add RTL function to externalization RAII

This patch adds the '_kmpc_get_hardware_num_threads_in_block'
OpenMP RTL function to the externalization RAII struct. This was getting
optimized out and then being replaced with an undefined value once added
back in, causing bugs for complex reductions.

Fixes #53909.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D120076

2 years ago[LLDB] Fix TestStructuredBinding.py for libstdc++
Shafik Yaghmour [Thu, 17 Feb 2022 19:29:05 +0000 (11:29 -0800)]
[LLDB] Fix TestStructuredBinding.py for libstdc++

For the tuple case for the TestStructuredBinding.py the result type is different
between libc++ and libstdc++.

2 years ago[ifs] Add --strip-needed flag
Alex Brachet [Thu, 17 Feb 2022 19:24:53 +0000 (19:24 +0000)]
[ifs] Add --strip-needed flag

Reviewed By: haowei, mcgrathr

Differential Revision: https://reviews.llvm.org/D119907

2 years ago[lld-macho] Allow order files and call graph sorting to be used together
Leonard Grey [Fri, 14 Jan 2022 19:37:00 +0000 (14:37 -0500)]
[lld-macho] Allow order files and call graph sorting to be used together

If both an order file and a call graph profile are present, the edges of the
call graph which use symbols present in the order file are not used. All of
the symbols in the order file will appear at the beginning of the section just
as they do currently. In other words, the highest priority derived from the
call graph will be below the lowest priority derived from the order file.

Practically, this change renames CallGraphSort.{h,cpp} to SectionPriorities.{h,cpp},
and most order file and call graph profile related code is moved into the new
file to reduce duplication.

Differential Revision: https://reviews.llvm.org/D117354

2 years ago[DEBUGINFO] [LLDB] Add support for generating debug-info for structured bindings...
Shafik Yaghmour [Thu, 17 Feb 2022 19:13:46 +0000 (11:13 -0800)]
[DEBUGINFO] [LLDB] Add support for generating debug-info for structured bindings of structs and arrays

Currently we are not emitting debug-info for all cases of structured bindings a
C++17 feature which allows us to bind names to subobjects in an initializer.

A structured binding is represented by a DecompositionDecl AST node and the
binding are represented by a BindingDecl. It looks the original implementation
only covered the tuple like case which be represented by a DeclRefExpr which
contains a VarDecl.

If the binding is to a subobject of the struct the binding will contain a
MemberExpr and in the case of arrays it will contain an ArraySubscriptExpr.
This PR adds support emitting debug-info for the MemberExpr and ArraySubscriptExpr
cases as well as llvm and lldb tests for these cases as well as the tuple case.

Differential Revision: https://reviews.llvm.org/D119178

2 years ago[AArch64] Add extra widening mul tests. NFC
David Green [Thu, 17 Feb 2022 19:11:45 +0000 (19:11 +0000)]
[AArch64] Add extra widening mul tests. NFC

Also regenerate arm64-neon-2velem-high.ll.

2 years ago[AMDGPU] Promote recursive loads from kernel argument to constant
Stanislav Mekhanoshin [Fri, 11 Feb 2022 20:00:05 +0000 (12:00 -0800)]
[AMDGPU] Promote recursive loads from kernel argument to constant

Not clobbered pointer load chains are promoted to global now. That
is possible to promote these loads itself into constant address
space. Loaded pointers still need to point to global because we
need to be able to store into that pointer and because an actual
load from it may occur after a clobber.

Differential Revision: https://reviews.llvm.org/D119886

2 years ago[mlir] Switch {collapse,expand}_shape ops to the declarative assembly format
Benjamin Kramer [Thu, 17 Feb 2022 18:56:38 +0000 (19:56 +0100)]
[mlir] Switch {collapse,expand}_shape ops to the declarative assembly format

Same functionality, a lot less code.

2 years ago[NFC] Fix debug-info-hotpatch.cpp failure due to downstream regex issue.
Zahira Ammarguellat [Thu, 17 Feb 2022 16:19:40 +0000 (08:19 -0800)]
[NFC] Fix debug-info-hotpatch.cpp failure due to downstream regex issue.

In our downstream, we discovered that the that the .* wildcard
in debug-info-hotpatch.cpp (added https://reviews.llvm.org/D116511)
ended up matching the entire line on our Windows configurations, causing
the -function-padmin check to already be consumed. After digging into it
we weren't able to find any sort of reason why the platform would matter
here, however we suspect there must be some difference in the regex
matcher between systems.
This NFC patch replaces the regex with a more conservative regex that
prevents this from happening by replacing the . match with an 'everything
but double-quote match, [^"].

https://reviews.llvm.org/D120066

2 years ago[Clang] Add attributes alloc_size and alloc_align to mm_malloc
Dávid Bolvanský [Thu, 17 Feb 2022 18:58:12 +0000 (19:58 +0100)]
[Clang] Add attributes alloc_size and alloc_align to mm_malloc

LLVM optimizes source codes with mm_malloc better, especially due to alignment info.

alloc align https://clang.llvm.org/docs/AttributeReference.html#alloc-align
alloc size https://clang.llvm.org/docs/AttributeReference.html#alloc-size

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D117091

2 years ago[X86ISelLowering] permit BlockAddressSDNode "i" constraints for PIC
Nick Desaulniers [Thu, 17 Feb 2022 18:43:12 +0000 (10:43 -0800)]
[X86ISelLowering] permit BlockAddressSDNode "i" constraints for PIC

When building 32b x86 code as PIC, the existing handling of "i"
constraints is conservative since generally we have to go through the
GOT to find references to functions.

But generally, BlockAddresses from C code refer to the Function in the
current TU.  Permit BlockAddresses to be used with the "i" constraint
for those cases.

I regressed this in
commit 4edb9983cb8c ("[SelectionDAG] treat X constrained labels as i for asm")

Fixes: https://github.com/llvm/llvm-project/issues/53868

Reviewed By: efriedma, MaskRay

Differential Revision: https://reviews.llvm.org/D119905

2 years agoFix the declaration printer to properly handle prototypes in C
Aaron Ballman [Thu, 17 Feb 2022 18:52:07 +0000 (13:52 -0500)]
Fix the declaration printer to properly handle prototypes in C

Previously, we would take a declaration like void f(void) and print it
as void f(). That's correct in C++ as far as it goes, but is incorrect
in C because that converts the function from having a prototype to one
which does not.

This turns out to matter for some of our tests that use the pretty
printer where we'd like to get rid of the K&R prototypes from the test
but can't because the test is checking the pretty printed function
signature, as done with the ARCMT tests.

2 years ago[Attributor][FIX] Ensure stable iteration order
Johannes Doerfert [Thu, 17 Feb 2022 18:48:32 +0000 (12:48 -0600)]
[Attributor][FIX] Ensure stable iteration order

With
https://github.com/llvm/llvm-project/commit/668c5c688be7ab0af37739bbbe2d653be82d5c6f
we introduced an ordering issue revealed by the reverse iteration
buildbot. Depending on the order of the map that tracks the AAIsDead AAs
we ended up with slightly different attributes. This is not totally
unexpected and can happen. We should however be deterministic in our
orderings to avoid such issues.

2 years ago[GlobalDCE] [VFE] Add a test for incorrect VFE behavior in presence of null/invalid...
Kuba Mracek [Thu, 17 Feb 2022 05:20:04 +0000 (21:20 -0800)]
[GlobalDCE] [VFE] Add a test for incorrect VFE behavior in presence of null/invalid vtable entries

Add a test for VFE where there's several vtables, and one of them contains an
invalid entry (from VFE's perspective), and which causes VFE to incorrectly skip
scanning subsequent vtables and drop their dependencies.

2 years ago[RewriteStatepointsForGC] Fix an incorrect assertion
Daniil Suchkov [Wed, 16 Feb 2022 23:21:15 +0000 (23:21 +0000)]
[RewriteStatepointsForGC] Fix an incorrect assertion

The assertion verifying that a newly computed value matches what is
already cached used stripPointerCasts() to strip bitcasts, however the
values can be not only pointers, but also vectors of pointers. That is
problematic because stripPointerCasts() doesn't handle vectors of
pointers. This patch introduces an ad-hoc utility function to strip all
bitcasts regardless of the value type.

Reviewed By: skatkov, reames

Differential Revision: https://reviews.llvm.org/D119994

2 years agoRevert "[Driver][Fuchsia][NFC] Use GetLinkerPath to see if linker is lld"
Alex Brachet [Thu, 17 Feb 2022 18:41:49 +0000 (18:41 +0000)]
Revert "[Driver][Fuchsia][NFC] Use GetLinkerPath to see if linker is lld"

This reverts commit b9f4dff8ab40250aac2343e86c1289de46af5585.

2 years ago[OpenMP] Diagnose bad 'omp declare variant' that references itself.
Mike Rice [Wed, 16 Feb 2022 21:58:45 +0000 (13:58 -0800)]
[OpenMP] Diagnose bad 'omp declare variant' that references itself.

When an a variant is specified that is the same as the base function
the compiler will end up crashing in CodeGen. Give an error instead.

Differential Revision: https://reviews.llvm.org/D119979

2 years ago[SystemZ] Improve emission of alignment hints.
Jonas Paulsson [Thu, 17 Feb 2022 01:12:06 +0000 (02:12 +0100)]
[SystemZ] Improve emission of alignment hints.

Handle multiple memoperands in lowerAlignmentHint().

Review: Ulrich Weigand

2 years ago[Driver][Fuchsia][NFC] Use GetLinkerPath to see if linker is lld
Alex Brachet [Thu, 17 Feb 2022 18:20:23 +0000 (18:20 +0000)]
[Driver][Fuchsia][NFC] Use GetLinkerPath to see if linker is lld

Reviewed By: phosek

Differential Revision: https://reviews.llvm.org/D120074

2 years ago[analyzer] Fix a crash in NoStateChangeVisitor with body-farmed stack frames.
Artem Dergachev [Thu, 17 Feb 2022 05:09:09 +0000 (21:09 -0800)]
[analyzer] Fix a crash in NoStateChangeVisitor with body-farmed stack frames.

LocationContext::getDecl() isn't useful for obtaining the "farmed" body because
the (synthetic) body statement isn't actually attached to the (natural-grown)
declaration in the AST.

Differential Revision: https://reviews.llvm.org/D119509

2 years ago[InstCombine][OpaquePtr] Check store type in DSE implementation
Arthur Eubanks [Thu, 17 Feb 2022 18:00:26 +0000 (10:00 -0800)]
[InstCombine][OpaquePtr] Check store type in DSE implementation

2 years ago[SLP][NFC]Add another test for swapped main/alternate cmp, NFC.
Alexey Bataev [Thu, 17 Feb 2022 17:34:58 +0000 (09:34 -0800)]
[SLP][NFC]Add another test for swapped main/alternate cmp, NFC.

2 years ago[instsimplify] When compare allocas, consider their minimal size
Philip Reames [Thu, 17 Feb 2022 17:50:32 +0000 (09:50 -0800)]
[instsimplify] When compare allocas, consider their minimal size

The code was using exact sizing only, but since what we really need is just to make sure the offsets are in bounds, a minimum bound on the object size is sufficient.

To demonstrate the difference, support computing minimum sizes from obects of scalable vector type.

2 years ago[CUDA][SPIRV] Assign global address space to CUDA kernel arguments
Shangwu Yao [Thu, 17 Feb 2022 17:38:06 +0000 (09:38 -0800)]
[CUDA][SPIRV] Assign global address space to CUDA kernel arguments

This patch converts CUDA pointer kernel arguments with default address space to
CrossWorkGroup address space (__global in OpenCL). This is because Generic or
Function (OpenCL's private) is not supported as storage class for kernel pointer types.

Differential Revision: https://reviews.llvm.org/D119207

2 years ago[instsimplify] Fix a miscompile with zero sized allocas
Philip Reames [Thu, 17 Feb 2022 17:21:42 +0000 (09:21 -0800)]
[instsimplify] Fix a miscompile with zero sized allocas

Remove some code which tried to handle the case of comparing two allocas where an object size could not be precisely computed.  This code had zero coverage in tree, and at least one nasty bug.

The bug comes from the fact that the code uses the size of the result pointer as a proxy for whether the alloca can be of size zero.  Since the result of an alloca is *always* a pointer type, and a pointer type can *never* be empty, this check was a nop.  As a result, we blindly consider a zero offset from two allocas to never be equal.  They can in fact be equal when one or more of the allocas is zero sized.

This is particularly ugly because instcombine contains the exact opposite rule.  If instcombine reaches the allocas first, it combines them into one (making them equal).  If instsimplify reaches the compare first, it would consider them not equal.  This creates all kinds of fun scenarios for order of optimization reaching different and contradictory conclusions.

2 years ago[flang] Lower simple scalar assignment
Valentin Clement [Thu, 17 Feb 2022 17:23:22 +0000 (18:23 +0100)]
[flang] Lower simple scalar assignment

This patch hanlde lowering of simple scalar assignment.

This patch is part of the upstreaming effort from fir-dev branch.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D120058

Co-authored-by: Jean Perier <jperier@nvidia.com>
2 years ago[libc] Add exit and atexit
Alex Brachet [Thu, 17 Feb 2022 17:21:55 +0000 (17:21 +0000)]
[libc] Add exit and atexit

Often atexit is implemented using __cxa_atexit. I have not implemented __cxa_atexit here because it potentially requires more discussion. It is unique for llvm-libc (I think) that it is an exported symbol that wouldn’t be defined in any spec file because it doesn’t have a header. Implementing it will be trivial given what is here already, but I figured it would be more contentious so it can be implemented later.

Reviewed By: lntue

Differential Revision: https://reviews.llvm.org/D119512

2 years ago[clangd] Fix building SerializationTests unit test on OpenBSD
Brad Smith [Thu, 17 Feb 2022 17:15:14 +0000 (12:15 -0500)]
[clangd] Fix building SerializationTests unit test on OpenBSD

This fixes building the unit tests on OpenBSD. OpenBSD does not support RLIMIT_AS.

Reviewed By: kadircet

Differential Revision: https://reviews.llvm.org/D119989