platform/upstream/llvm.git
22 months ago[NFC][flang] Lowering options clean-up.
Slava Zakharin [Tue, 1 Nov 2022 22:12:43 +0000 (15:12 -0700)]
[NFC][flang] Lowering options clean-up.

This change-set defines the LoweringOptions the same way
other options are defined in Flang.

Differential Revision: https://reviews.llvm.org/D137207

22 months ago[LoongArch] Fix codegen for [su]itofp instructions
wanglei [Thu, 3 Nov 2022 03:10:41 +0000 (11:10 +0800)]
[LoongArch] Fix codegen for [su]itofp instructions

This patch fixes codegen for `[su]itofp` instructions.

In LoongArch, a legal int-to-float conversion is done in two steps:

1. Move the data from `GPR` to `FPR`. (FRLen >= GRLen)
2. Conversion in `FPR`. (the data in `FPR` is treated as a signed value)

Based on the above features, when the type's BitWidth meets the
requirements, all `SINT_TO_FP` are legal, all `UINT_TO_FP` are expand
and lowered to libcall when appropriate.
The only special case is, LoongArch64 with `+f,-d` features. At this
point, custom processing is required for `[SU]INT_TO_FP`. Of course, we
can also ignore it and use libcall directly.

Differential Revision: https://reviews.llvm.org/D136916

22 months ago[X86] Remove `IntrArgMemOnly` from cmpccxadd intrinsics
Phoebe Wang [Thu, 3 Nov 2022 02:39:31 +0000 (10:39 +0800)]
[X86] Remove `IntrArgMemOnly` from cmpccxadd intrinsics

CMPSXADD will modify memory, so we can't use `IntrArgMemOnly` here.
Found it during review D137250.

22 months ago[sanitizer] Fix vfork interception on loongarch64
Xi Ruoyao [Thu, 3 Nov 2022 03:07:53 +0000 (11:07 +0800)]
[sanitizer] Fix vfork interception on loongarch64

Fix a brown paper bag error made by me in D129418.  I didn't set
ASAN_INTERCEPT_VFORK correctly for loongarch64, but created an all-zero
object for __interception::real_vfork.  This caused anything calling
vfork() to die instantly.

Fix this issue by setting ASAN_INTERCEPT_VFORK and remove the bad
all-zero definition.  Other ports have an all-zero common definition but
we don't need it at least for now.

And, enable ASAN vfork test for loongarch64 to prevent regression in the
future.

Differential Revision: https://reviews.llvm.org/D137160

22 months ago[asan] Use proper shadow offset for loongarch64 in instrumentation passes
Youling Tang [Thu, 3 Nov 2022 03:03:37 +0000 (11:03 +0800)]
[asan] Use proper shadow offset for loongarch64 in instrumentation passes

Instrumentation passes now use the proper shadow offset. There will be many
asan test failures without this patch. For example:

```
$ ./lib/asan/tests/LOONGARCH64LinuxConfig/Asan-loongarch64-calls-Test
AddressSanitizer:DEADLYSIGNAL
=================================================================
==651209==ERROR: AddressSanitizer: SEGV on unknown address 0x1ffffe2dfa9b (pc 0x5555585e151c bp 0x7ffffb9ec070 sp 0x7ffffb9ebfd0 T0)
==651209==The signal is caused by a UNKNOWN memory access.
```

Before the patch:

```
$ make check-asan
Testing Time: 36.13s
  Unsupported      : 205
  Passed           :  83
  Expectedly Failed:   1
  Failed           : 239
```

After the patch:
```
$ make check-asan
Testing Time: 58.98s
  Unsupported      : 205
  Passed           : 421
  Expectedly Failed:   1
  Failed           :  89
```

Differential Revision: https://reviews.llvm.org/D137013

22 months ago[asan] Default to -fsanitize-address-use-odr-indicator for non-Windows
Fangrui Song [Thu, 3 Nov 2022 02:21:33 +0000 (19:21 -0700)]
[asan] Default to -fsanitize-address-use-odr-indicator for non-Windows

This enables odr indicators on all platforms and private aliases on non-Windows.
Note that GCC also uses private aliases: this fixes bogus
`The following global variable is not properly aligned.` errors for interposed global variables

Fix https://github.com/google/sanitizers/issues/398
Fix https://github.com/google/sanitizers/issues/1017
Fix https://github.com/llvm/llvm-project/issues/36893 (we can restore D46665)

Global variables of non-hasExactDefinition() linkages (i.e.
linkonce/linkonce_odr/weak/weak_odr/common/external_weak) are not instrumented.
If an instrumented variable gets interposed to an uninstrumented variable due to
symbol interposition (e.g. in issue 36893, _ZTS1A in foo.so is resolved to _ZTS1A in
the executable), there may be a bogus error.

With private aliases, the register code will not resolve to a definition in
another module, and thus prevent the issue.

Cons: minor size increase. This is mainly due to extra `__odr_asan_gen_*` symbols.
(ELF) In addition, in relocatable files private aliases replace some relocations
referencing global symbols with .L symbols and may introduce some STT_SECTION symbols.

For lld, with -g0, the size increase is 0.07~0.09% for many configurations I
have tested: -O0, -O1, -O2, -O3, -O2 -ffunction-sections -fdata-sections
-Wl,--gc-sections. With -g1 or above, the size increase ratio will be even smaller.

This patch obsoletes D92078.

Don't migrate Windows for now: the static data member of a specialization
`std::num_put<char>::id` is a weak symbol, as well as its ODR indicator.
Unfortunately, link.exe (and lld without -lldmingw) generally doesn't support
duplicate weak definitions (weak symbols in different TUs likely pick different
defined external symbols and conflict).

Differential Revision: https://reviews.llvm.org/D137227

22 months ago[flang][RFC] Add lowering design for procdure pointers
Peixin-Qiao [Thu, 3 Nov 2022 01:00:52 +0000 (09:00 +0800)]
[flang][RFC] Add lowering design for procdure pointers

This document aims to give insights at the representation of procdure
pointers in FIR.

Reviewed By: PeteSteinfeld, jeanPerier, kiranchandramohan

Differential Revision: https://reviews.llvm.org/D136840

22 months agoLangRef: Fix typo in backtick placement
Matt Arsenault [Wed, 2 Nov 2022 23:13:12 +0000 (16:13 -0700)]
LangRef: Fix typo in backtick placement

22 months ago[Driver] Don't preprocess source files when reproducing linker crashes
Alex Brachet [Thu, 3 Nov 2022 00:12:10 +0000 (00:12 +0000)]
[Driver] Don't preprocess source files when reproducing linker crashes

It's not necessary to redo the source file preprocessing for reproducing linker
crashes because we must have successfully created the object file by this point.
Skip this step, and also don't report the preprocessed source file or create
the clang invocation shell script. The latter is no longer sensible without the
preprocessed source, or helpful given the linker reproducer will have it's own
shell script.

Differential Revision: https://reviews.llvm.org/D137289

22 months ago[ConstraintElimination] Skip compares with scalable vector types.
Florian Hahn [Wed, 2 Nov 2022 23:57:06 +0000 (23:57 +0000)]
[ConstraintElimination] Skip compares with scalable vector types.

Materializing scalable vectors with boolean values is not implemented
yet. Skip those cases for now and leave a TODO.

22 months ago[Clang] follow-up D128745, remove ClangABICompat checks
Yuanfang Chen [Wed, 2 Nov 2022 18:54:19 +0000 (11:54 -0700)]
[Clang] follow-up D128745, remove ClangABICompat checks

Per discussions in D128745, remove ClangABICompat checks for implementations
of DR692/DR1395/DR1432. This is a potentially breaking changes, so the release
note is updated accordingly.

Reviewed By: aaron.ballman

Differential Revision: https://reviews.llvm.org/D136120

22 months agoSPARC: Register null target streamer
Matt Arsenault [Sat, 29 Oct 2022 19:52:24 +0000 (12:52 -0700)]
SPARC: Register null target streamer

Fixes null dereference in emitFunctionBodyStart for 64-bit

22 months agoWebAssembly: Remove unnecessary set check
Matt Arsenault [Wed, 2 Nov 2022 21:29:14 +0000 (14:29 -0700)]
WebAssembly: Remove unnecessary set check

The empty set will be default constructed if this wasn't
in the map already.

22 months agoWebAssembly: Move exception handling code together
Matt Arsenault [Wed, 2 Nov 2022 20:44:57 +0000 (13:44 -0700)]
WebAssembly: Move exception handling code together

22 months agoFunctionLoweringInfo: Use TLI member instead of finding it
Matt Arsenault [Wed, 2 Nov 2022 20:17:30 +0000 (13:17 -0700)]
FunctionLoweringInfo: Use TLI member instead of finding it

22 months ago[mlir] Add call_intrinsic op to LLVMIIR
electriclilies [Wed, 2 Nov 2022 22:56:28 +0000 (15:56 -0700)]
[mlir] Add call_intrinsic op to LLVMIIR

The call_intrinsic op allows us to call LLVM intrinsics from the LLVMDialect without implementing a new op every time.

Reviewed By: lattner, rriddle

Differential Revision: https://reviews.llvm.org/D137187

22 months ago[libc] Add implementation of ungetc.
Siva Chandra Reddy [Wed, 2 Nov 2022 20:19:04 +0000 (20:19 +0000)]
[libc] Add implementation of ungetc.

A bug in the file read logic has also been fixed along the way. Parts
of the ungetc tests will fail without that bug fixed.

Reviewed By: michaelrj

Differential Revision: https://reviews.llvm.org/D137286

22 months ago[ARM] Fix vector ule zero lowering
David Green [Wed, 2 Nov 2022 22:34:05 +0000 (22:34 +0000)]
[ARM] Fix vector ule zero lowering

The instruction icmp ule <4 x i32> %0, zeroinitializer will usually be
simplified to icmp eq <4 x i32> %0, zeroinitializer. It is not
guaranteed though, and the code for lowering vector compares could pick
the wrong form of the instruction if this happened. I've tried to make
the code more explicit about the supported conditions.

This fixes NEON being unable to select VCMPZ with HS conditions, and
fixes some incorrect MVE patterns.

Fixes #58514.

Differential Revision: https://reviews.llvm.org/D136447

22 months ago[libc++][Android] strong_order_long_double.verify.cpp: disable on i686
Ryan Prichard [Wed, 2 Nov 2022 21:55:56 +0000 (14:55 -0700)]
[libc++][Android] strong_order_long_double.verify.cpp: disable on i686

This target (as well as 32-bit ARM Android) have sizeof(long double)
equal to sizeof(double).

Reviewed By: ldionne, #libc

Differential Revision: https://reviews.llvm.org/D137135

22 months ago[libc++][Android] XFAIL aligned_alloc and timespec_get tests
Ryan Prichard [Wed, 2 Nov 2022 21:53:51 +0000 (14:53 -0700)]
[libc++][Android] XFAIL aligned_alloc and timespec_get tests

Mark tests XFAIL that use APIs that are unsupported on old versions of
Android:
 - aligned_alloc isn't available until API 28.
 - timespec_get isn't available until API 29.

Reviewed By: ldionne, #libc

Differential Revision: https://reviews.llvm.org/D137134

22 months ago[OPENMP]Initial support for error directive.
Jennifer Yu [Tue, 1 Nov 2022 21:46:12 +0000 (14:46 -0700)]
[OPENMP]Initial support for error directive.

Differential Revision: https://reviews.llvm.org/D137209

22 months ago[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator.
Kirill Stoimenov [Wed, 2 Nov 2022 20:58:36 +0000 (20:58 +0000)]
[Sanitizers] Modified __aarch64__ to use the 64 bit version of the allocator.

This change will switch SizeClassAllocator32 to SizeClassAllocator64 on ARM. This might potentially affect ARM platforms with 39-bit address space. This addresses [[ https://github.com/google/sanitizers/issues/703  | issues/703  ]], but unlike [[ https://reviews.llvm.org/D60243 | D60243 ]] it defaults to 64 bit allocator.

Reviewed By: vitalybuka, MaskRay

Differential Revision: https://reviews.llvm.org/D137136

22 months ago[libc++][test] Fixes transitive includes.
Mark de Wever [Wed, 2 Nov 2022 18:43:53 +0000 (19:43 +0100)]
[libc++][test] Fixes transitive includes.

These were accidentally set to generating in
243da90ea5357c1ca324f714ea4813dc9029af27

Reviewed By: #libc, philnik

Differential Revision: https://reviews.llvm.org/D137278

22 months ago[gn build] Port cf239c2f1777
LLVM GN Syncbot [Wed, 2 Nov 2022 21:01:56 +0000 (21:01 +0000)]
[gn build] Port cf239c2f1777

22 months ago[mlir] Remove eliminateBlockingAwaitOps option in AsyncToAsyncRuntime pass
yijiagu [Wed, 2 Nov 2022 20:51:48 +0000 (13:51 -0700)]
[mlir] Remove eliminateBlockingAwaitOps option in AsyncToAsyncRuntime pass

Remove the eliminateBlockingAwaitOps option in AsyncToAsyncRuntime pass

Today the AsyncToAsyncRuntime pass does two things: one is converting normal funcs with async ops to coroutine cfg; the other is lowering high level async operations to async.coro and async.runtime operations. This patch removes the converting step from AsyncToAsyncRuntime pass.

In the next step we will create a new asyncfication pass for converting normal funcs to the newly added async.func operation.

Reviewed By: ezhulenev

Differential Revision: https://reviews.llvm.org/D137282

22 months ago[NFC] Make format() more amenable to format attributes
Félix Cloutier [Fri, 9 Sep 2022 21:08:19 +0000 (14:08 -0700)]
[NFC] Make format() more amenable to format attributes

This change modifies the implementation of the format() function
so that vendor forks committed to building with compilers that
support __attribute__((format)) on non-variadic functions can
check the format() function with it.

rdar://84571523

22 months ago[RISCV] Prevent autovectorization using vscale with Zvl32b.
Craig Topper [Wed, 2 Nov 2022 20:09:52 +0000 (13:09 -0700)]
[RISCV] Prevent autovectorization using vscale with Zvl32b.

RVVBitsPerBlock is 64. If VLen==32, VLen/RVVBitsPerBlock is 0.

Reviewed By: reames

Differential Revision: https://reviews.llvm.org/D137280

22 months ago[clang][pp] only __is_target_environment(unknown) should match unknown target triple...
Alex Lorenz [Wed, 2 Nov 2022 20:40:52 +0000 (13:40 -0700)]
[clang][pp] only __is_target_environment(unknown) should match unknown target triple environment

22 months ago[VectorCombine] add tests for load+shuffle and update to typeless ptr; NFC
Sanjay Patel [Wed, 2 Nov 2022 19:55:51 +0000 (15:55 -0400)]
[VectorCombine] add tests for load+shuffle and update to typeless ptr; NFC

22 months ago[OpenMP][libomp] Fix disabled affinity
Jonathan Peyton [Tue, 1 Nov 2022 17:29:17 +0000 (12:29 -0500)]
[OpenMP][libomp] Fix disabled affinity

Fix setting affinity type and topology method when affinity is disabled
and fix places that were not taking into account that affinity can be
explicitly disabled by putting proper KMP_AFFINITY_CAPABLE() check.

Differential Revision: https://reviews.llvm.org/D137176

22 months ago[clang-format] Don't skip #else/#elif of #if 0
Owen Pan [Sun, 30 Oct 2022 23:45:00 +0000 (16:45 -0700)]
[clang-format] Don't skip #else/#elif of #if 0

Fixes #58188.

Differential Revision: https://reviews.llvm.org/D137052

22 months agoRevert "[Attributor][NFCI] Move MemIntrinsic handling into the initializer"
Joseph Huber [Wed, 2 Nov 2022 20:28:50 +0000 (15:28 -0500)]
Revert "[Attributor][NFCI] Move MemIntrinsic handling into the initializer"

This was causing failures when optimizing codes with complex numbers.
Revert until a fix can be implemented.

This reverts commit 7fdf3564c04075d3e6be2d9540e5a6f1e084be9f.

22 months ago[LinkerWrapper] report on missing libraries
Joseph Huber [Tue, 1 Nov 2022 18:16:47 +0000 (18:16 +0000)]
[LinkerWrapper] report on missing libraries

The linker wrapper does its own library searching for static archives
that can contain device code. The device linking phases happen before
the host linking phases so that we can generate the necessary
registration code and link it in with the rest of the code. Previously,
If a library containing needed device code was not found the execution
would continue silently until it failed with undefined symbols. This
patch allows the linker wrapper to perform its own check beforehand to
catch these errors.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D137180

22 months ago[clang] Fix inline builtin functions of an __asm__ renamed function with symbol prefixes
Martin Storsjö [Mon, 31 Oct 2022 10:18:21 +0000 (12:18 +0200)]
[clang] Fix inline builtin functions of an __asm__ renamed function with symbol prefixes

If a function is renamed with `__asm__`, the name provided is the
exact symbol name, without any extra implicit symbol prefixes.
If the target does use symbol prefixes, the IR level symbol gets
an `\01` prefix to indicate that it's a literal symbol name to be
taken as is.

When a builtin function is specialized by providing an inline
version of it, that inline function is named `<funcname>.inline`.

When the base function has been renamed due to `__asm__`, the inline
function ends up named `<asmname>.inline`. Up to this point,
things did work as expected before.

However, for targets with symbol prefixes, one codepath that produced
the combined name `<asmname>.inline` used the mangled `asmname` with
`\01` prefix, while others didn't. This patch fixes this.

This fixes the combination of asm renamed builtin function, with
inline override of the function, on any target with symbol
prefixes (such as i386 windows and any Darwin target).

Differential Revision: https://reviews.llvm.org/D137073

22 months ago[RISCV] Move RVVBitsPerBlock to TargetParser.h so we can use it in clang. NFC
Craig Topper [Wed, 2 Nov 2022 16:18:41 +0000 (09:18 -0700)]
[RISCV] Move RVVBitsPerBlock to TargetParser.h so we can use it in clang. NFC

Differential Revision: https://reviews.llvm.org/D137266

22 months ago[LV][RISCV] Add test showing poor choice of VF for short loop
Philip Reames [Wed, 2 Nov 2022 20:00:35 +0000 (13:00 -0700)]
[LV][RISCV] Add test showing poor choice of VF for short loop

22 months ago[wasm-ld] Update supported features in the generic CPU configuration
Dan Gohman [Tue, 18 Oct 2022 18:10:19 +0000 (11:10 -0700)]
[wasm-ld] Update supported features in the generic CPU configuration

Accompanying https://reviews.llvm.org/D125728, this updates LLVM
Codegen's "generic" CPU to enable the same new features.

Differential Revision: https://reviews.llvm.org/D125729

22 months ago[flang][NFC] Fix typo in filename
Valentin Clement [Wed, 2 Nov 2022 19:47:19 +0000 (20:47 +0100)]
[flang][NFC] Fix typo in filename

22 months ago[flang] Fix for polymoprhic pointer component
Valentin Clement [Wed, 2 Nov 2022 19:46:09 +0000 (20:46 +0100)]
[flang] Fix for polymoprhic pointer component

Fix path that generates MutableBox for
pointer component.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D137270

22 months ago[mlir][llvmir] Support FastmathFlags for LLVM intrinsic operations.
Slava Zakharin [Wed, 2 Nov 2022 18:42:46 +0000 (11:42 -0700)]
[mlir][llvmir] Support FastmathFlags for LLVM intrinsic operations.

This is required for D126305 code to propagate fastmath attributes
for Arith operations that are converted to LLVM IR intrinsics
operations.

LLVM IR intrinsic operations are using custom assembly format now
to avoid printing {fastmathFlags = #llvm.fastmath<none>}, which
is too verbose.

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D136225

22 months ago[libc++] Implement P2438R2 (std::string::substr() &&)
Nikolas Klauser [Tue, 9 Aug 2022 11:17:30 +0000 (13:17 +0200)]
[libc++] Implement P2438R2 (std::string::substr() &&)

This doesn't affect our ABI because `std::string::substr()` isn't in the dylib and the mangling of `substr() const` and `substr() const&` are different.

Reviewed By: ldionne, Mordante, var-const, avogelsgesang, #libc

Spies: arphaman, huixie90, libcxx-commits

Differential Revision: https://reviews.llvm.org/D131668

22 months ago[bolt] update bazel rules for f71d32a0eea4
Augie Fackler [Wed, 2 Nov 2022 16:29:24 +0000 (12:29 -0400)]
[bolt] update bazel rules for f71d32a0eea4

Differential Revision: https://reviews.llvm.org/D137281

22 months ago[OpenMP][OpenMPIRBuilder] Migrate loadOffloadInfoMetadata from clang to OMPIRbuilder
Akash Banerjee [Thu, 27 Oct 2022 18:45:26 +0000 (19:45 +0100)]
[OpenMP][OpenMPIRBuilder] Migrate loadOffloadInfoMetadata from clang to OMPIRbuilder

This patch moves the implementation of the loadOffloadInfoMetadata to the OMPIRbuilder.

Differential Revision: https://reviews.llvm.org/D136872

22 months ago[clangd] Add scoped enum constants to all-scopes-completion
Tom Praschan [Mon, 31 Oct 2022 20:36:18 +0000 (21:36 +0100)]
[clangd] Add scoped enum constants to all-scopes-completion

This was originally part of https://reviews.llvm.org/D136925, but we decided to move it to a separate patch.
In case it turns out to be controversial, it can be reverted more easily.

Differential Revision: https://reviews.llvm.org/D137104

22 months agoAdd Async Function to the Async Dialect
yijiagu [Wed, 2 Nov 2022 18:27:26 +0000 (11:27 -0700)]
Add Async Function to the Async Dialect

Add Async Function to the Async Dialect

Today `async.execute` operation semantics requires attached region to be executed in a thread managed by the runtime, and always returns an `!async.token` result. We need to model async functions that are not necessarily executed in a runtime-managed threads, but eventually lowered to llvm coroutines.

Example:

```
async.func @foo(%arg0: !async.value<f32>) -> !async.token {
  %0 = async.await %arg0: !async.value<f32>
  "do_something_with_f32"(%0)
  return
}
```

If `arg0` is available this function will be executed in the caller thread. If it's not available it will be suspended and resumed later later on a thread managed by the async runtime. Currently this is not representable with `async.execute` operations.

The longer term goal is to make async dialect more like https://github.com/lewissbaker/cppcoro to be able to represent structured host concurrency in MLIR.

(1) Add async.func, async.call, and async.return operations in Async Dialect

Reviewed By: ezhulenev, rriddle

Differential Revision: https://reviews.llvm.org/D137189

22 months agoFix LLVM sphinx build bot
Aaron Ballman [Wed, 2 Nov 2022 18:32:47 +0000 (14:32 -0400)]
Fix LLVM sphinx build bot

This should address the issue found in:
https://lab.llvm.org/buildbot/#/builders/30/builds/27824

22 months ago[TSan] Adjust `TSAN_RTL_CFLAGS` before it gets copied
Julian Lettner [Wed, 2 Nov 2022 18:21:34 +0000 (11:21 -0700)]
[TSan] Adjust `TSAN_RTL_CFLAGS` before it gets copied

Add `COMPILER_RT_LIBDISPATCH_CFLAGS` to `TSAN_RTL_CFLAGS` before it gets
duplicated to `TSAN_RTL_DYNAMIC_CFLAGS` so both versions have the
necessary flags.

Reviewed By: wrotki, rsundahl

Differential Revision: https://reviews.llvm.org/D137183

22 months ago[ConstraintElimination] Generate true/false vectors for vector cmps.
Florian Hahn [Wed, 2 Nov 2022 18:13:34 +0000 (18:13 +0000)]
[ConstraintElimination] Generate true/false vectors for vector cmps.

This fixes crashes when vector compares can be simplified to true/false.

22 months ago[mlir][linalg] Add support for vectorizing convs that have different types.
Hanhan Wang [Wed, 2 Nov 2022 18:02:48 +0000 (11:02 -0700)]
[mlir][linalg] Add support for vectorizing convs that have different types.

Reviewed By: dcaballe

Differential Revision: https://reviews.llvm.org/D137208

22 months ago[llvm-tblgen] NFC: Small code refactor in DecoderEmitter.
James Y Knight [Wed, 2 Nov 2022 16:29:42 +0000 (12:29 -0400)]
[llvm-tblgen] NFC: Small code refactor in DecoderEmitter.

Extracts part of populateInstruction into a separate
addOneOperandFields function.

22 months agoRevert "[cmake][msvc] Enable standards-conforming preprocessor"
Jan Svoboda [Wed, 2 Nov 2022 17:57:24 +0000 (10:57 -0700)]
Revert "[cmake][msvc] Enable standards-conforming preprocessor"

This reverts commit 12d8e7c6ade55bba241259312e3e4bdcf6aeab81.

The Windows MLIR buildbot started failing with:

C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winbase.h(9531): error C2220: the following warning is treated as an error
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winbase.h(9531): warning C5105: macro expansion producing 'defined' has undefined behavior
C:\Program Files (x86)\Windows Kits\10\include\10.0.19041.0\um\winbase.h(9531): note: to simplify migration, consider the temporary use of /Wv:18 flag with the version of the compiler with which you used to build without warnings

22 months agollvm-reduce: Require x86 to run file ouput test
Matt Arsenault [Wed, 2 Nov 2022 17:39:50 +0000 (10:39 -0700)]
llvm-reduce: Require x86 to run file ouput test

The MIR test somewhat depends on target support.

22 months agoAMDGPU: Directly pass Function to mayUseAGPRs
Matt Arsenault [Wed, 2 Nov 2022 17:38:29 +0000 (10:38 -0700)]
AMDGPU: Directly pass Function to mayUseAGPRs

This was taking the MachineFunction, but only inspecting the
underlying IR.

22 months ago[NewPM][Pipeline] Add PipelineTuningOption to set inliner threshold
Arthur Eubanks [Sun, 30 Oct 2022 16:52:20 +0000 (09:52 -0700)]
[NewPM][Pipeline] Add PipelineTuningOption to set inliner threshold

The legacy PM allowed you to set a custom inliner threshold via
  builder.Inliner = llvm::createFunctionInliningPass(inline_threshold);

This allows the same thing to be done with the new PM optimization pipelines.

Reviewed By: asbirlea

Differential Revision: https://reviews.llvm.org/D137038

22 months ago[docs][NewPM] Move pass plugin documentation into existing new PM docs
Arthur Eubanks [Mon, 24 Oct 2022 17:45:51 +0000 (10:45 -0700)]
[docs][NewPM] Move pass plugin documentation into existing new PM docs

Reviewed By: awarzynski, asbirlea

Differential Revision: https://reviews.llvm.org/D136626

22 months ago[opt] Add -p alias for -passes
Arthur Eubanks [Sun, 23 Oct 2022 19:02:59 +0000 (12:02 -0700)]
[opt] Add -p alias for -passes

See [1] for background.

Some people have complained that `opt -passes=instcombine` is a lot more typing than `opt -instcombine`.
As a compromise that nobody has objected to in [1], allow `opt -p instcombine`.

[1] https://discourse.llvm.org/t/rfc-legacy-opt-pass-syntax-with-new-pass-manager/65863

Reviewed By: bjope, asbirlea

Differential Revision: https://reviews.llvm.org/D136616

22 months ago[lldb] Add information on type systems to statistics dump command
Alex Langford [Mon, 31 Oct 2022 21:50:04 +0000 (14:50 -0700)]
[lldb] Add information on type systems to statistics dump command

Context: I plan on using this change primarily downstream in the apple
fork of llvm to track swift module loading time.

Reviewed By: clayborg, tschuett

Differential Revision: https://reviews.llvm.org/D137191

22 months ago[llvm-config] Remove --src-root option
Michał Górny [Tue, 1 Nov 2022 03:23:36 +0000 (04:23 +0100)]
[llvm-config] Remove --src-root option

Remove the `--src-root` option from the deprecated llvm-config tool.
None of the llvm-project projects use this option anymore. The value
was only meaningful for in-tree use and usually became no longer correct
once LLVM was installed -- either because it was built in a temporary
directory, or installed from a binary package and built on a different
system entirely.  Therefore, third-party tools could not have been
relying on it anyway.

The LLVM_SRC_ROOT #define is left intact, as it is used to compute
includedir when llvm-config is used in-source.

Differential Revision: https://reviews.llvm.org/D137144

22 months ago[ASAN] Removed special case controlling allocator constants for __aarch64__.
Kirill Stoimenov [Wed, 2 Nov 2022 16:13:02 +0000 (16:13 +0000)]
[ASAN] Removed special case controlling allocator constants for __aarch64__.

This patch should land before D137136 to make sure that the leak sanitizer allocator works correctly. This patch is NFC without D137136.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D137265

22 months ago[PGO] Add a threshold for number of critical edges in PGO
Rong Xu [Wed, 2 Nov 2022 17:11:34 +0000 (10:11 -0700)]
[PGO] Add a threshold for number of critical edges in PGO

For some auto-generated sources, we have a huge number of critical
edges (like from switch statements). We have seen instance of 183777
critical edges in one function.

After we split the critical edges in PGO instrumentation/profile-use
pass, the CFG is so large that we have compiler time issues in
downstream passes (like in machine CSE and block placement). Here I
add a threshold to skip PGO if the number of critical edges are too
large.

The threshold is large enough so that it will not affect the majority
of PGO compilation.

Also sync the logic for skipping instrumentation and profile-use. I
think this is the correct thing to do.

Differential Revision: https://reviews.llvm.org/D137184

22 months ago[Assignment Tracking][2/*] Add flags to enable Assignment Tracking
OCHyams [Wed, 2 Nov 2022 17:00:58 +0000 (17:00 +0000)]
[Assignment Tracking][2/*] Add flags to enable Assignment Tracking

The Assignment Tracking debug-info feature is outlined in this RFC:

https://discourse.llvm.org/t/
rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir

Enable in clang: -Xclang -fexperimental-assignment-tracking
Enable in llvm tools: -experimental-assignment-tracking

When assignment tracking is enabled in clang it will pass on the flag to enable
the feature in lllvm. It's undefined behaviour to read IR that contains
assignment tracking metadata without specifying the feature flags.

Tests will come with later patches that add assignment tracking features.

Reviewed By: jmorse

Differential Revision: https://reviews.llvm.org/D132221

22 months ago[mlir][scf] support 1:N type conversion for scf.if/while/condition
Peiming Liu [Mon, 31 Oct 2022 16:26:32 +0000 (16:26 +0000)]
[mlir][scf] support 1:N type conversion for scf.if/while/condition

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D137100

22 months ago[ValueLattice] Fix typo in condition (NFC)
Nikita Popov [Wed, 2 Nov 2022 16:50:02 +0000 (17:50 +0100)]
[ValueLattice] Fix typo in condition (NFC)

Fix typo pointed out by Roman Divacky.

There should be no functional change, as the rest of the code will
return nullptr for undef anyway. The condition is just there for
clarity.

22 months ago[mlir][arith] Use declarative asm format for fastmath flags
rkayaith [Mon, 31 Oct 2022 16:50:17 +0000 (12:50 -0400)]
[mlir][arith] Use declarative asm format for fastmath flags

Switch to using an optional group in the assembly format instead of a
custom directive. There's no change to the actual printed format.

Reviewed By: vzakhari

Differential Revision: https://reviews.llvm.org/D137087

22 months ago[mlir][Transform] Fix ASAN heap-use-after-free
Nicolas Vasilache [Wed, 2 Nov 2022 16:39:16 +0000 (09:39 -0700)]
[mlir][Transform] Fix ASAN heap-use-after-free

22 months ago[mlir][scf] refactor scf structuralOpConversion to better support 1:N type conversion
Peiming Liu [Fri, 28 Oct 2022 23:39:47 +0000 (23:39 +0000)]
[mlir][scf] refactor scf structuralOpConversion to better support 1:N type conversion

This patch moves the 1:N type mapping into its own classes to allow better code reuse in D137100.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D137099

22 months ago[openmp][AMDGPU] - Correct getNumberOfBlocks calculation.
Ethan Stewart [Wed, 2 Nov 2022 16:37:42 +0000 (11:37 -0500)]
[openmp][AMDGPU] - Correct getNumberOfBlocks calculation.

This patch fixes the 6 amdgpu buildbot lit test failures
introduced by https://reviews.llvm.org/D135444.
      libomptarget :: amdgcn-amd-amdhsa :: mapping/reduction_implicit_map.cpp
      libomptarget :: amdgcn-amd-amdhsa :: offloading/cuda_no_devices.c
      libomptarget :: amdgcn-amd-amdhsa :: offloading/target-teams-atomic.c
      libomptarget :: amdgcn-amd-amdhsa-LTO :: mapping/reduction_implicit_map.cpp
      libomptarget :: amdgcn-amd-amdhsa-LTO :: offloading/cuda_no_devices.c
      libomptarget :: amdgcn-amd-amdhsa-LTO :: offloading/target-teams-atomic.c

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D137261

22 months ago[cmake][msvc] Enable standards-conforming preprocessor
Jan Svoboda [Wed, 2 Nov 2022 16:35:00 +0000 (09:35 -0700)]
[cmake][msvc] Enable standards-conforming preprocessor

Since we now only support Visual Studio 2019 16.7 and newer, we're able to use the /Zc:preprocessor flag that turns on the standards-conforming preprocessor. Among other things, it correctly expands __VA_ARGS__. This enables removal of some boilerplate in D135128.

Reviewed By: Bigcheese, thieta

Differential Revision: https://reviews.llvm.org/D135128

22 months ago[lldb] Update TestDump.test for gnuwin32's 'file' command output
Jonas Devlieghere [Wed, 2 Nov 2022 16:30:55 +0000 (09:30 -0700)]
[lldb] Update TestDump.test for gnuwin32's 'file' command output

Remove the colon from the CHECK line for the output of the file command.
On macOS and Linux, the file command uses a colon as the separator
between the path and the file type, but gnuwin32's file command uses a
semicolon.

22 months ago[mlir][Pass] Disallow mixing -pass-pipeline with other pass options
rkayaith [Tue, 11 Oct 2022 19:23:48 +0000 (15:23 -0400)]
[mlir][Pass] Disallow mixing -pass-pipeline with other pass options

Currently `-pass-pipeline` can be specified multiple times and mixed
with the individual `-pass-name` options. Removing this feature will
allow for including the pipeline anchor as part of the option
argument (see D134900).

Reviewed By: rriddle

Differential Revision: https://reviews.llvm.org/D135745

22 months ago[mlir] Don't mix -pass-pipeline with other pass options
rkayaith [Tue, 11 Oct 2022 21:29:39 +0000 (17:29 -0400)]
[mlir] Don't mix -pass-pipeline with other pass options

These are test updates required for D135745, which disallows mixing
`-pass-pipeline` and the individual `-pass-name` options.

Reviewed By: rriddle, mehdi_amini

Differential Revision: https://reviews.llvm.org/D135746

22 months ago[OpenMP][libomptarget] Fix AsyncInfoTy object in omp_target_memcpy
Kevin Sala Penadés [Wed, 2 Nov 2022 16:03:34 +0000 (12:03 -0400)]
[OpenMP][libomptarget] Fix AsyncInfoTy object in omp_target_memcpy

The AsyncInfoTy should be created in the same device as the async operation will be issued. In omp_target_memcpy, the AsyncInfoTy for the host to destination device transfer was created referring to the source device.

Reviewed By: tianshilei1992

Differential Revision: https://reviews.llvm.org/D137225

22 months ago[flang] Handle dispatch on nopass procedure from array element
Valentin Clement [Wed, 2 Nov 2022 15:52:05 +0000 (16:52 +0100)]
[flang] Handle dispatch on nopass procedure from array element

When a fir.dispatch is done with a NOPASS type-bound
procedure, the object in fir.dispatch is just used to gather the vtable.
Therefore, no transformation is done on it and the original entity
is used. The current code generation didn't expect the entity to be an array.
This patch update the code generation to be able to retrieve the vtable
accordingly.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D137255

22 months agollvm-reduce: Add explicit triple to test
Matt Arsenault [Wed, 2 Nov 2022 15:30:21 +0000 (08:30 -0700)]
llvm-reduce: Add explicit triple to test

Some host targets are managing to get through MIR parsing
without constructing their MachineFunctionInfo.

Fixes at least SystemZ and SPARC (issue 58768)

22 months ago[libc] Add a doc about the libc fullbuild mode.
Siva Chandra Reddy [Mon, 31 Oct 2022 08:25:25 +0000 (01:25 -0700)]
[libc] Add a doc about the libc fullbuild mode.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D137061

22 months ago[libc] Add links to libc Discourse and to the libc Discord channel.
Siva Chandra Reddy [Wed, 2 Nov 2022 07:34:11 +0000 (00:34 -0700)]
[libc] Add links to libc Discourse and to the libc Discord channel.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D137229

22 months ago[libc] Add a contributing guide to the docs.
Siva Chandra Reddy [Fri, 28 Oct 2022 08:09:38 +0000 (01:09 -0700)]
[libc] Add a contributing guide to the docs.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D136961

22 months ago[libc] Add a porting guide to the docs.
Siva Chandra Reddy [Fri, 28 Oct 2022 07:12:08 +0000 (00:12 -0700)]
[libc] Add a porting guide to the docs.

Reviewed By: jeffbailey

Differential Revision: https://reviews.llvm.org/D136960

22 months ago[RISCV] Fix cost of legal fixed length masked load and stores
Philip Reames [Wed, 2 Nov 2022 00:19:29 +0000 (17:19 -0700)]
[RISCV] Fix cost of legal fixed length masked load and stores

We can cost them the same way as a scalable masked load/store. By hitting the default path, we were costing them as if they were being scalarized. This is a significant over estimate.

Differential Revision: https://reviews.llvm.org/D137218

22 months agoFix Assignment Tracking docs error from 33c7ae55e729069be754f56c4d4606cdeddd377b
OCHyams [Wed, 2 Nov 2022 14:08:23 +0000 (14:08 +0000)]
Fix Assignment Tracking docs error from 33c7ae55e729069be754f56c4d4606cdeddd377b

link: https://lab.llvm.org/buildbot/#/builders/30/builds/27812
22 months ago[mlir] Use the same pipeline tuning options as clang for execution engine
Denys Shabalin [Wed, 2 Nov 2022 13:04:37 +0000 (14:04 +0100)]
[mlir] Use the same pipeline tuning options as clang for execution engine

This change make sure that ExecutionEngine's pass pipeline is identical to one
used by clang. Previously, SLPVectorization was not enabled which caused
differences in code...

...generation.

Reviewed By: ftynse

Differential Revision: https://reviews.llvm.org/D137248

22 months ago[MIRVRegNamer] Avoid opcode hash collision
John Brawn [Fri, 28 Oct 2022 13:17:01 +0000 (14:17 +0100)]
[MIRVRegNamer] Avoid opcode hash collision

D121929 happens to cause CodeGen/MIR/AArch64/mirnamer.mir to fail due
to a hash collision caused by adding two extra opcodes. The collision
is only in the top 19 bits of the hashed opcode so fix this by just
using the whole hash (in fixed width hex for consistency) instead of
the top 5 decimal digits.

Differential Revision: https://reviews.llvm.org/D137155

22 months ago[MachineCSE] Allow PRE of instructions that read physical registers
John Brawn [Thu, 27 Oct 2022 13:14:57 +0000 (14:14 +0100)]
[MachineCSE] Allow PRE of instructions that read physical registers

Currently MachineCSE forbids PRE when the instruction reads a physical
register. Relax this so that it's allowed when the value being read is
the same as what would be read in the place the instruction would be
hoisted to.

This is being done in preparation for adding FPCR handling to the
AArch64 backend, in order to prevent it to from worsening the
generated code, but for targets that already have a similar register
it should improve things.

This patch affects code generation in several tests. The new code
looks better except for in Thumb2/LowOverheadLoops/memcall.ll where
we perform PRE but the LowOverheadLoops transformation then undoes
it. Also in AMDGPU/selectcc-opt.ll the CHECK makes things look worse,
but actually the function as a whole is better (as a MOV is PRE'd).

Differential Revision: https://reviews.llvm.org/D136675

22 months ago[CVP] Add test for icmp without constant operand (NFC)
Nikita Popov [Wed, 2 Nov 2022 13:47:57 +0000 (14:47 +0100)]
[CVP] Add test for icmp without constant operand (NFC)

We should still be able to fold comparisons based on ranged in
this case.

22 months ago[Assignment Tracking][1/*] Add initial docs for Assignment Tracking
OCHyams [Wed, 2 Nov 2022 13:42:52 +0000 (13:42 +0000)]
[Assignment Tracking][1/*] Add initial docs for Assignment Tracking

The Assignment Tracking debug-info feature is outlined in this RFC:

https://discourse.llvm.org/t/
  rfc-assignment-tracking-a-better-way-of-specifying-variable-locations-in-ir

Add documentation outlining the intent and design.

22 months ago[InstCombine] fold mul with decremented "shl -1" factor (2nd try)
Sanjay Patel [Wed, 2 Nov 2022 13:15:11 +0000 (09:15 -0400)]
[InstCombine] fold mul with decremented "shl -1" factor (2nd try)

This is a corrected version of:
bc886e9b587b

I made a copy-paste error that created an "add" instead of the
intended "sub" on that attempt. The regression tests showed the
bug, but I overlooked that.

As I said in a comment on issue #58717, the bug reports resulting
from the botched patch confirm that the pattern does occur in
many real-world applications, so hopefully eliminating the multiply
results in better code.

I added one more regression test in this version of the patch,
and here's an Alive2 proof to show that exact example:
https://alive2.llvm.org/ce/z/dge7VC

Original commit message:

This is a sibling to:
6064e92b0a84
...but we canonicalize the shl+add to shl+xor,
so the pattern is different than I expected:
https://alive2.llvm.org/ce/z/8CX16e

I have not found any patterns that are safe
to propagate no-wrap, so that is not included
here.

Differential Revision: https://reviews.llvm.org/D137157

22 months ago[DebugInfo] Fix minor debug info bug in deleteDeadLoop
OCHyams [Thu, 20 Oct 2022 14:40:14 +0000 (15:40 +0100)]
[DebugInfo] Fix minor debug info bug in deleteDeadLoop

Using a DebugVariable as the set key rather than std::pair<DIVariable *,
DIExpression *> ensures we don't accidently confuse multiple instances of
inlined variables.

Reviewed By: jryans

Differential Revision: https://reviews.llvm.org/D133303

22 months ago[BasicAA] Re-enable cs-cs-arm.ll test (PR58738)
Nikita Popov [Wed, 2 Nov 2022 13:21:38 +0000 (14:21 +0100)]
[BasicAA] Re-enable cs-cs-arm.ll test (PR58738)

Fixes https://github.com/llvm/llvm-project/issues/58738.

22 months ago[mlir][linalg] Fix crash in canonicalization pattern
Matthias Springer [Wed, 2 Nov 2022 11:36:54 +0000 (12:36 +0100)]
[mlir][linalg] Fix crash in canonicalization pattern

This crash was due to incorrect usage of `hasTensorSemantics`, which has changed recently with DestinationStyleOpInterface. An op has tensor semantics if all of its inits and inputs are tensors. Previously, only inits needed to be tensors.

Differential Revision: https://reviews.llvm.org/D137243

22 months ago[Test] Add test exposing crash in SimpleLoopUnswitch
Dmitry Makogon [Wed, 2 Nov 2022 12:55:56 +0000 (19:55 +0700)]
[Test] Add test exposing crash in SimpleLoopUnswitch

The test crashes with an assert in unswitchNontrivialInvariants.

22 months ago[InstCombine] use logical-and matcher to avoid crash
Sanjay Patel [Wed, 2 Nov 2022 12:21:42 +0000 (08:21 -0400)]
[InstCombine] use logical-and matcher to avoid crash

Follow-on to:
ec0b406e16c44f1554

This should prevent crashing for example like issue #58552
by not matching a select-of-vectors-with-scalar-condition.

The test that shows a regression seems unlikely to occur
in real code.

This also picks up an optimization in the case where a real
(bitwise) logic op is used. We could already convert some
similar select ops to real logic via impliesPoison(), so
we don't see more diffs on commuted tests. Using commutative
matchers (when safe) might also handle one of the TODO tests.

22 months ago[InstCombine] add tests for logical-and / logical-or folds; NFC
Sanjay Patel [Wed, 2 Nov 2022 12:14:44 +0000 (08:14 -0400)]
[InstCombine] add tests for logical-and / logical-or folds; NFC

Similar to 29661fe94bf12ced1 - there are matching
deficiencies and a potential crash lurking in
these patterns.

22 months ago[wasm] Always treat DWARF expression addresses as load addresses
Philip Pfaffe [Wed, 2 Nov 2022 10:51:18 +0000 (10:51 +0000)]
[wasm] Always treat DWARF expression addresses as load addresses

When resolving absolute addresses for DW_OP_addr or DW_OP_addrx, these are always load addresses rather than file addresses in wasm.

Reviewed By: DavidSpickett

Differential Revision: https://reviews.llvm.org/D135664

22 months agoReenable POSIX builtin library functions in gnu2x mode
Aaron Ballman [Wed, 2 Nov 2022 11:56:43 +0000 (07:56 -0400)]
Reenable POSIX builtin library functions in gnu2x mode

gnu17 and earlier modes automatically expose several POSIX C APIs, and
this was accidentally disabled for gnu2x in
7d644e1215b376ec5e915df9ea2eeb56e2d94626.

This restores the behavior for gnu2x mode (without changing the
behavior in C standards modes instead of GNU modes).

Fixes #56607

22 months ago[InstCombine] Remove redundant splats in InstCombineVectorOps
Matt Devereau [Wed, 12 Oct 2022 09:20:05 +0000 (09:20 +0000)]
[InstCombine] Remove redundant splats in InstCombineVectorOps

Splatting the first vector element of the result of a BinOp, where any of the
BinOp's operands are the result of a first vector element splat can be simplified to
splatting the first vector element of the result of the BinOp

Differential Revision: https://reviews.llvm.org/D135876

22 months ago[AMDGPU][NFC] Split MC tests into promoted from VOP1 to VOP3 and only VOP3
Mirko Brkusanin [Tue, 1 Nov 2022 15:32:55 +0000 (16:32 +0100)]
[AMDGPU][NFC] Split MC tests into promoted from VOP1 to VOP3 and only VOP3

Differential Revision: https://reviews.llvm.org/D136149

22 months ago[AMDGPU][NFC] Split MC tests into promoted from VOP2 to VOP3 and only VOP3
Mirko Brkusanin [Tue, 1 Nov 2022 15:22:45 +0000 (16:22 +0100)]
[AMDGPU][NFC] Split MC tests into promoted from VOP2 to VOP3 and only VOP3

Differential Revision: https://reviews.llvm.org/D136148

22 months ago[RuntimeDyld] Require x86_64-linux for IFUNC test
Moritz Sichert [Wed, 2 Nov 2022 11:08:52 +0000 (12:08 +0100)]
[RuntimeDyld] Require x86_64-linux for IFUNC test

This test was introduced by a667aa4de041816cb4865bce8f523228f2332ffa. It
assumes that it runs on x86_64 on linux, so require that for the test.

22 months ago[AMDGPU] Constify a couple of methods. NFC.
Jay Foad [Wed, 2 Nov 2022 11:03:59 +0000 (11:03 +0000)]
[AMDGPU] Constify a couple of methods. NFC.

22 months ago[LoopPredication][NFCI] Perform 'visited' check before pushing to worklist
Max Kazantsev [Wed, 2 Nov 2022 11:02:59 +0000 (18:02 +0700)]
[LoopPredication][NFCI] Perform 'visited' check before pushing to worklist

This prevents duplicates to be pushed into the stack and hypothetically
should reduce memory footprint on ugly cornercases with multiple repeating
duplicates in 'and' tree.