platform/upstream/llvm.git
23 months ago[libunwind] Fix a few libunwind includes
Ryan Prichard [Wed, 7 Sep 2022 21:27:57 +0000 (17:27 -0400)]
[libunwind] Fix a few libunwind includes

In UnwindCursor.hpp, include config.h before checking _LIBUNWIND_SUPPORT_SEH_UNWIND.

Include libunwind_ext.h for UNW_STEP_SUCCESS.

Differential Revision: https://reviews.llvm.org/D86766

23 months ago[flang] Write semantics test for atomic_fetch_add
Katherine Rasmussen [Wed, 31 Aug 2022 19:12:00 +0000 (12:12 -0700)]
[flang] Write semantics test for atomic_fetch_add

Write a semantics test for the atomic intrinsic subroutine,
atomic_fetch_add.

Reviewed By: rouson

Differential Revision: https://reviews.llvm.org/D133139

23 months agoApply clang-tidy fixes for readability-identifier-naming in CRunnerUtils.cpp (NFC)
Mehdi Amini [Mon, 29 Aug 2022 11:24:57 +0000 (11:24 +0000)]
Apply clang-tidy fixes for readability-identifier-naming in CRunnerUtils.cpp (NFC)

23 months agoApply clang-tidy fixes for performance-unnecessary-value-param in VectorDistribute...
Mehdi Amini [Mon, 29 Aug 2022 11:22:27 +0000 (11:22 +0000)]
Apply clang-tidy fixes for performance-unnecessary-value-param in VectorDistribute.cpp (NFC)

23 months ago[mlir:PassTiming] Always use parentInfo for determining pipeline parent scope
River Riddle [Tue, 30 Aug 2022 22:25:55 +0000 (15:25 -0700)]
[mlir:PassTiming] Always use parentInfo for determining pipeline parent scope

This fixes a bug where, depending on thread usage, a pipeline may be
misattributed to a timer that wasn't it's parent.

Differential Revision: https://reviews.llvm.org/D132979

23 months ago[BOLT] Distinguish sections in heatmap
Fabian Parzefall [Wed, 7 Sep 2022 23:28:01 +0000 (16:28 -0700)]
[BOLT] Distinguish sections in heatmap

Output different letters for different sections in the heatmap to
visually separate sections.

Reviewed By: maksfb

Differential Revision: https://reviews.llvm.org/D133068

23 months ago[AMDGPU] Drop _oneuse checks from med3 patterns
Justin Bogner [Wed, 24 Aug 2022 22:35:02 +0000 (15:35 -0700)]
[AMDGPU] Drop _oneuse checks from med3 patterns

We use _oneuse checks to make sure combines won't accidentally
increase code size, but this prevents the optimization in cases where
we happen to want to clamp multiple values to the same range

It's safe to drop these checks for two reasons:

1. The pattern of max/min operations for med3 is complicated enough
   it's unlikely to come up by accident, so this will still only fire
   when appropriate to do so
2. Even if every intermediate is used and we don't save a single
   operation, we still won't end up with more operations since the
   med3 replaces the final max/min.

In pathological cases we could potentially end up with a larger
encoding size or possibly slightly increased vgpr pressure, but the
risk of that is low, especially considering the upside.

Differential Revision: https://reviews.llvm.org/D132621

23 months ago[AMDGPU] Fix liveness verifier error in hazard recognizer
Stanislav Mekhanoshin [Wed, 7 Sep 2022 22:24:11 +0000 (15:24 -0700)]
[AMDGPU] Fix liveness verifier error in hazard recognizer

After D133067 we are inserting swaps to use a new physical
register. I have noticed verifier errors about undefined
physical register uses if we are tracking liveness post RA.

We have no access to LIS at this point, so mark new register
uses as undef to calm down the verifier. Liveness should not
matter at this point anyway.

Note the description of the RegState::Undef: "Value of the
register doesn't matter." I.e. it does not say it is strictly
undefined. In fact that is what we really need: this value
does not matter.

I also had to modify the test a bit since with tracking enabled
it does not pass verification even before the recognizer.

Differential Revision: https://reviews.llvm.org/D133459

23 months ago[libc][math] Implement asinf function correctly rounded for all rounding modes.
Tue Ly [Wed, 7 Sep 2022 06:20:45 +0000 (02:20 -0400)]
[libc][math] Implement asinf function correctly rounded for all rounding modes.

Implement asinf function correctly rounded for all rounding modes.

For `|x| <= 0.5`, we approximate `asin(x)` by
```
  asin(x) = x * P(x^2)
```
where `P(X^2) = Q(X)` is a degree-20 minimax even polynomial approximating
`asin(x)/x` on `[0, 0.5]` generated by Sollya with:
```
  > Q = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20|],
                 [|1, D...|], [0, 0.5]);
```

When `|x| > 0.5`, we perform range reduction as follow:
Assume further that `0.5 < x <= 1`, and let:
```
  y = asin(x)
```
We will use the double angle formula:
```
  cos(2X) = 1 - 2 sin^2(X)
```
and the complement angle identity:
```
  x = sin(y) = cos(pi/2 - y)
              = 1 - 2 sin^2 (pi/4 - y/2)
```
So:
```
  sin(pi/4 - y/2) = sqrt( (1 - x)/2 )
```
And hence:
```
  pi/4 - y/2 = asin( sqrt( (1 - x)/2 ) )
```
Equivalently:
```
  asin(x) = y = pi/2 - 2 * asin( sqrt( (1 - x)/2 ) )
```
Let `u = (1 - x)/2`, then
```
  asin(x) = pi/2 - 2 * asin(u)
```
Moreover, since `0.5 < x <= 1`,
```
  0 <= u < 1/4, and 0 <= sqrt(u) < 0.5.
```
And hence we can reuse the same polynomial approximation of `asin(x)` when
`|x| <= 0.5`:
```
  asin(x) = pi/2 - 2 * u * P(u^2).
```

Performance benchmark using `perf` tool from the CORE-MATH project on Ryzen 1700:
```
$ CORE_MATH_PERF_MODE="rdtsc" ./perf.sh asinf
CORE-MATH reciprocal throughput   : 23.418
System LIBC reciprocal throughput : 27.310
LIBC reciprocal throughput        : 22.741

$ CORE_MATH_PERF_MODE="rdtsc" ./perf.sh asinf --latency
GNU libc version: 2.35
GNU libc release: stable
CORE-MATH latency   : 58.884
System LIBC latency : 62.055
LIBC latency        : 62.037
```

Reviewed By: orex, zimmermann6

Differential Revision: https://reviews.llvm.org/D133400

23 months ago[mlir][sparse] change variable dimension to fixed attribute pointers/indices
Aart Bik [Wed, 7 Sep 2022 22:08:29 +0000 (15:08 -0700)]
[mlir][sparse] change variable dimension to fixed attribute pointers/indices

The "sparsification" pass does not need the ability to use runtime values for
the dimension, so the only source for variability would have been user code.
Restricting the dimension to constants simplifies code generation.

Reviewed By: Peiming, wrengr

Differential Revision: https://reviews.llvm.org/D133458

23 months ago[libc] Return correct values for hypot when overflowed.
Tue Ly [Tue, 6 Sep 2022 18:18:18 +0000 (14:18 -0400)]
[libc] Return correct values for hypot when overflowed.

Hypot incorrectly returns +Inf when overflowed with FE_DOWNWARD and
FE_TOWARDZERO rounding modes.

Reviewed By: sivachandra, zimmermann6

Differential Revision: https://reviews.llvm.org/D133370

23 months ago[InstCombine] Baseline tests for folding x + (x | -x) to x & (x - 1)
Marc Auberer [Wed, 7 Sep 2022 23:12:23 +0000 (19:12 -0400)]
[InstCombine] Baseline tests for folding x + (x | -x) to x & (x - 1)

Baseline tests for D133362

Differential Revision: https://reviews.llvm.org/D133449

23 months agoRevert D133266 "[MinGW] Reject explicit non-default visibility applied to dllexport...
Fangrui Song [Wed, 7 Sep 2022 23:06:19 +0000 (16:06 -0700)]
Revert D133266 "[MinGW] Reject explicit non-default visibility applied to dllexport/dllimport declaration"

This reverts commit 91d8324366f405e871aa8174ab61fc66912964dd.

The combo dllexport protected makes sense and is used by PlayStation.
Will change the patch to allow dllexport protected.

23 months ago[gn build] Port 97c2220565ab
LLVM GN Syncbot [Wed, 7 Sep 2022 23:00:40 +0000 (23:00 +0000)]
[gn build] Port 97c2220565ab

23 months ago[gn build] Port 3e7350f3173b
LLVM GN Syncbot [Wed, 7 Sep 2022 23:00:39 +0000 (23:00 +0000)]
[gn build] Port 3e7350f3173b

23 months ago[InstCombine] Fix a crash in -kcfi debug block
Sami Tolvanen [Wed, 7 Sep 2022 19:22:17 +0000 (19:22 +0000)]
[InstCombine] Fix a crash in -kcfi debug block

Don't attempt to print out DebugLoc as we may not have one.

23 months ago[gn build] port aa484c90cf59
Nico Weber [Wed, 7 Sep 2022 22:55:07 +0000 (18:55 -0400)]
[gn build] port aa484c90cf59

23 months ago[gn build] port e321c8dd2cea
Nico Weber [Wed, 7 Sep 2022 22:52:06 +0000 (18:52 -0400)]
[gn build] port e321c8dd2cea

23 months ago[Libomptarget] Implement OpenMP 5.2 semantics for device pointers
Joseph Huber [Wed, 7 Sep 2022 18:37:26 +0000 (13:37 -0500)]
[Libomptarget] Implement OpenMP 5.2 semantics for device pointers

In OpenMP 5.2, §5.8.6, page 160 line 32-33, when a device pointer
allocated by omp_target_alloc has implicitly been included on a target
construct as a zero-length array, the pointer initialisation should not
find a matching mapped list item, and so should retain its value as a
firstprivate variable. Previously, we would return a null pointer if the
list item was not found. This patch updates the map handling to the
OpenMP 5.2 semantics.

Reviewed By: jdoerfert, ye-luo

Differential Revision: https://reviews.llvm.org/D133447

23 months ago[Hexagon] Move vectorization checks from subtarget to TTI
Krzysztof Parzyszek [Wed, 7 Sep 2022 21:46:22 +0000 (14:46 -0700)]
[Hexagon] Move vectorization checks from subtarget to TTI

23 months ago[ODRHash diagnostics] Move `ODRDiagsEmitter` to libAST in separate files. NFC.
Volodymyr Sapsai [Tue, 28 Jun 2022 03:06:07 +0000 (20:06 -0700)]
[ODRHash diagnostics] Move `ODRDiagsEmitter` to libAST in separate files. NFC.

Intend to use `ODRDiagsEmitter` during parsing to diagnose a parsed
definition differing from a definition with the same name from a hidden
[sub]module.

Differential Revision: https://reviews.llvm.org/D128695

23 months ago[AMDGPU] W/a hazard if 64 bit shift amount is a highest allocated VGPR
Stanislav Mekhanoshin [Wed, 31 Aug 2022 21:14:05 +0000 (14:14 -0700)]
[AMDGPU] W/a hazard if 64 bit shift amount is a highest allocated VGPR

In this case gfx90a uses v0 instead of the correct register. Swap
the value temporarily with a lower register and then swap it back.

Unfortunately hazard recognizer works after wait count insertion,
so we cannot simply reuse an arbitrary register, hence w/a also
includes a full waitcount. This can be avoided if we run it from
expandPostRAPseudo, but that is a complete misplacement.

Differential Revision: https://reviews.llvm.org/D133067

23 months agoC++/ObjC++: switch to gnu++17 as the default standard
Fangrui Song [Wed, 7 Sep 2022 21:17:52 +0000 (14:17 -0700)]
C++/ObjC++: switch to gnu++17 as the default standard

Clang's default C++ standard is now `gnu++17` instead of `gnu++14`:
https://discourse.llvm.org/t/c-objc-switch-to-gnu-17-as-the-default-dialect/64360

* CUDA/HIP are unchanged: C++14 from D103221.
* Sony PS4/PS5 are unchanged: https://discourse.llvm.org/t/c-objc-switch-to-gnu-17-as-the-default-dialect/64360/6
* lit feature `default-std-cxx` is added to keep CLANG_DEFAULT_STD_CXX=xxx tests working.
  Whether the cmake variable should be retained is disccused in D133375.

Depends on D131464

Close https://github.com/llvm/llvm-project/issues/56946

Reviewed By: #clang-language-wg, aaron.ballman

Differential Revision: https://reviews.llvm.org/D131465

23 months ago[ASan] Show memory rather than tag addresses in tag dump
Florian Mayer [Wed, 7 Sep 2022 00:44:20 +0000 (17:44 -0700)]
[ASan] Show memory rather than tag addresses in tag dump

Reviewed By: eugenis

Differential Revision: https://reviews.llvm.org/D133391

23 months ago[MLIR] NFC: add back `exports_files(["LICENSE.txt"])`.
Christian Sigg [Wed, 7 Sep 2022 19:57:02 +0000 (21:57 +0200)]
[MLIR] NFC: add back `exports_files(["LICENSE.txt"])`.

23 months ago[MLIR] NFC: expand comment in bazel BUILD file
Christian Sigg [Wed, 7 Sep 2022 19:57:02 +0000 (21:57 +0200)]
[MLIR] NFC: expand comment in bazel BUILD file

... and remove unnecessry `exports_files()`.

23 months ago[Bazel] Make clang/unittests:lex_tests depend on //llvm:TestingSupport
Fangrui Song [Wed, 7 Sep 2022 19:51:31 +0000 (12:51 -0700)]
[Bazel] Make clang/unittests:lex_tests depend on //llvm:TestingSupport

23 months agoApply clang-tidy fixes for llvm-else-after-return in SparseTensorRewriting.cpp (NFC)
Mehdi Amini [Mon, 29 Aug 2022 11:12:46 +0000 (11:12 +0000)]
Apply clang-tidy fixes for llvm-else-after-return in SparseTensorRewriting.cpp (NFC)

23 months agoApply clang-tidy fixes for performance-unnecessary-value-param in Utils.cpp (NFC)
Mehdi Amini [Mon, 29 Aug 2022 11:09:54 +0000 (11:09 +0000)]
Apply clang-tidy fixes for performance-unnecessary-value-param in Utils.cpp (NFC)

23 months ago[Clang] Introduce -fexperimental-sanitize-metadata=
Marco Elver [Tue, 6 Sep 2022 13:49:39 +0000 (15:49 +0200)]
[Clang] Introduce -fexperimental-sanitize-metadata=

Introduces the frontend flag -fexperimental-sanitize-metadata=, which
enables SanitizerBinaryMetadata instrumentation.

The first intended user of the binary metadata emitted will be a variant
of GWP-TSan [1]. The plan is to open source a stable and production
quality version of GWP-TSan. The development of which, however, requires
upstream compiler support.

[1] https://llvm.org/devmtg/2020-09/slides/Morehouse-GWP-Tsan.pdf

Until the tool has been open sourced, we mark this kind of
instrumentation as "experimental", and reserve the option to change
binary format, remove features, and similar.

Reviewed By: vitalybuka, MaskRay

Differential Revision: https://reviews.llvm.org/D130888

23 months ago[SanitizerBinaryMetadata] Introduce SanitizerBinaryMetadata instrumentation pass
Marco Elver [Tue, 6 Sep 2022 13:49:33 +0000 (15:49 +0200)]
[SanitizerBinaryMetadata] Introduce SanitizerBinaryMetadata instrumentation pass

Introduces the SanitizerBinaryMetadata instrumentation pass which uses
the new MD_pcsections metadata kinds to instrument certain types of
instructions and functions required for breakpoint-based sanitizers.

The first intended user of the binary metadata emitted will be a variant
of GWP-TSan [1]. GWP-TSan will require information about atomic
accesses; to unambiguously determine if an access is atomic or not, we
also require "covered" information which code has been compiled with
SanitizerBinaryMetadata instrumentation enabled.

[1] https://llvm.org/devmtg/2020-09/slides/Morehouse-GWP-Tsan.pdf

Reviewed By: dvyukov

Differential Revision: https://reviews.llvm.org/D130887

23 months ago[Libomptarget] Remove leftover ELF header from x86 plugin
Joseph Huber [Wed, 7 Sep 2022 18:40:54 +0000 (13:40 -0500)]
[Libomptarget] Remove leftover ELF header from x86 plugin

Summary:
We removed the linking support for `gelf.h` in a previous patch. This
header was incorrectly leftover causing build problems on some systems.

23 months ago[RISCV] Remove space before colon in error message.
Craig Topper [Wed, 7 Sep 2022 18:30:03 +0000 (11:30 -0700)]
[RISCV] Remove space before colon in error message.

23 months ago[lld-macho] Hardlink -object_path_lto files to cache when possible
Leonard Grey [Fri, 2 Sep 2022 17:50:01 +0000 (13:50 -0400)]
[lld-macho] Hardlink -object_path_lto files to cache when possible

This is a follow-up to https://reviews.llvm.org/D131624 (specifically to https://reviews.llvm.org/D131624#3716584)

Differential revision: https://reviews.llvm.org/D133218

23 months ago[libc++] Fixes CI.
Mark de Wever [Wed, 7 Sep 2022 17:56:52 +0000 (19:56 +0200)]
[libc++] Fixes CI.

It seems merging the changes in transitive macros and recent commits
conflicted.

23 months ago[mlir][sparse] Refactoring: remove dependence on tuple type when lowering sparse...
Peiming Liu [Wed, 7 Sep 2022 00:49:44 +0000 (00:49 +0000)]
[mlir][sparse] Refactoring: remove dependence on tuple type when lowering sparse tensors.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D133390

23 months ago[Libomptarget] Replace libelf with LLVM's Elf libraries
Joseph Huber [Fri, 5 Aug 2022 18:32:42 +0000 (14:32 -0400)]
[Libomptarget] Replace libelf with LLVM's Elf libraries

This patch replaces the dependency on `libelf` with LLVM's ELF support.
With this patch the user no-longer needs to have `libelf` on their
system to build and configure OpenMP offloading. The replacement is
mostly mechanical, with the exception of the hash table support which
was added in D131309.

Depends on D131309

Reviewed By: JonChesterfield, saiislam

Differential Revision: https://reviews.llvm.org/D131401

23 months ago[Libomptarget] Add utility functions for loading an ELF symbol by name
Joseph Huber [Sat, 3 Sep 2022 16:38:26 +0000 (11:38 -0500)]
[Libomptarget] Add utility functions for loading an ELF symbol by name

The `SHT_HASH` sections in an ELF are used to look up a symbol in the
symbol table using a symbol's name. This is done by obtaining the
`SHT_HASH` section and using its `sh_link` attribute to access the
associated symbol table, from which we can access the string table
containing the associated name. We can then search for the symbol using
the hash of the name and the buckets and chains in the hash table
itself

This patch adds utility functions that allow us to look up a symbol in
an ELF file by name. It will first attempt to look through the hash
tables, and then search the section tables manually if failed. This
allows us to pull out constants necessary for setting up offloading
without first loading the object.

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D131309

23 months ago[AArch64] Add tests for using tbl for fp conversions.
Florian Hahn [Wed, 7 Sep 2022 17:35:29 +0000 (18:35 +0100)]
[AArch64] Add tests for using tbl for fp conversions.

23 months ago[RISCV] Remove unnecessary word from error message.
Craig Topper [Wed, 7 Sep 2022 17:29:58 +0000 (10:29 -0700)]
[RISCV] Remove unnecessary word from error message.

23 months ago[Lex/DependencyDirectivesScanner] Keep track of the presence of tokens between the...
Argyrios Kyrtzidis [Mon, 5 Sep 2022 21:28:32 +0000 (14:28 -0700)]
[Lex/DependencyDirectivesScanner] Keep track of the presence of tokens between the last scanned directive and EOF

Directive `dependency_directives_scan::tokens_present_before_eof` is introduced to indicate there were tokens present before
the last scanned dependency directive and EOF.
This is useful to ensure we correctly identify the macro guards when lexing using the dependency directives.

Differential Revision: https://reviews.llvm.org/D133357

23 months ago[amdgpu] Always, instead of mostly, remove unused LDS symbols
Jon Chesterfield [Wed, 7 Sep 2022 17:28:14 +0000 (18:28 +0100)]
[amdgpu] Always, instead of mostly, remove unused LDS symbols

Currently LDS variables are removed by the lower module pass
if they have a use which is caught by the replace with struct control flow.
This makes tests brittle to changes to that control flow which induces
noise when trying to improve lowering. Some tests already check that
variables are removed, while others checked that they are not removed.

LDS variables are not (currently) externally accessible, and if that
changes the machinery which makes them externally accessible will look
like a use. This change therefore breaks no applications.

Reviewed By: rampitec

Differential Revision: https://reviews.llvm.org/D133028

23 months ago[mlir][sparse] minor zero test refactoring in rewriting
Aart Bik [Tue, 6 Sep 2022 23:09:17 +0000 (16:09 -0700)]
[mlir][sparse] minor zero test refactoring in rewriting

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D133382

23 months ago[mlir][sparse] fix python indentation in test
Aart Bik [Tue, 6 Sep 2022 23:57:19 +0000 (16:57 -0700)]
[mlir][sparse] fix python indentation in test

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D133385

23 months ago[gn build] Port e5d2d3eafbb3
LLVM GN Syncbot [Wed, 7 Sep 2022 16:59:10 +0000 (16:59 +0000)]
[gn build] Port e5d2d3eafbb3

23 months ago[libc++][locale] Removes an transitive include.
Mark de Wever [Wed, 7 Sep 2022 16:52:41 +0000 (18:52 +0200)]
[libc++][locale] Removes an transitive include.

Removes <cstdarg> transitive include from <locale> in C++23.

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D133254

23 months ago[AArch64] Add additional tests for tbl expansion.
Florian Hahn [Wed, 7 Sep 2022 16:47:59 +0000 (17:47 +0100)]
[AArch64] Add additional tests for tbl expansion.

Add test coverage to make sure tbl expansion isn't used when optimizing
for size or when the zext is only executed conditionally in a loop.

23 months agoAvoid __builtin_assume_aligned crash when the 1st arg is array type
yronglin [Wed, 7 Sep 2022 16:46:20 +0000 (12:46 -0400)]
Avoid __builtin_assume_aligned crash when the 1st arg is array type

Avoid __builtin_assume_aligned crash when the 1st arg is array type (or
string literal).

Fixes Issue #57169

Differential Revision: https://reviews.llvm.org/D133202

23 months ago[libc++][chrono] Implements formatter day.
Mark de Wever [Sun, 20 Mar 2022 12:40:02 +0000 (13:40 +0100)]
[libc++][chrono] Implements formatter day.

This implements the enabled specializaton
template<class charT> struct formatter<chrono::day, charT>;

and
template<class charT, class traits>
    basic_ostream<charT, traits>&
      operator<<(basic_ostream<charT, traits>& os, const day& d);

Implements:
- LWG 3241 chrono-spec grammar ambiguity in §[time.format]

Partially implements:
- P1361 Integration of chrono with text formatting

Reviewed By: #libc, ldionne

Differential Revision: https://reviews.llvm.org/D128577

23 months ago[libc++][format] Updates feature-test macros.
Mark de Wever [Sun, 4 Sep 2022 11:56:36 +0000 (13:56 +0200)]
[libc++][format] Updates feature-test macros.

During the discussion on the SG-10 mailinglist regarding the format
feature-test macros voted in during the last plenary it turns out libc++
can't mark the format feature-test macro as implemented.

According to
  https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations#__cpp_lib_format
the not yet implemented paper
  P1361R2 Integration of chrono with text formatting
affects the feature test macro.

Note that P1361R2 doesn't mention the feature-test macro nor is there an
LWG-issue to address the issue. The reporter of the issue didn't recall
where this requirement exactly has been decided.

Reviewed By: ldionne, #libc

Differential Revision: https://reviews.llvm.org/D133271

23 months ago[RISCV][MC] Add minimal support for Ztso extension
Philip Reames [Wed, 7 Sep 2022 15:45:50 +0000 (08:45 -0700)]
[RISCV][MC] Add minimal support for Ztso extension

This is a minimalist implementation which simply adds the extension (in the experimental namespace since its not ratified), and wires up the setting of the required ELF header flag. Future changes will include codegen changes to exploit the stronger memory model.

This is intended to implement v0.1 of the proposed specification which can be found in Chapter 25 of https://github.com/riscv/riscv-isa-manual/releases/download/draft-20220723-10eea63/riscv-spec.pdf.

Differential Revision: https://reviews.llvm.org/D133239

23 months ago[SCCP] convert signed div/rem to unsigned for non-negative operands, 2nd try
Sanjay Patel [Wed, 7 Sep 2022 15:44:27 +0000 (11:44 -0400)]
[SCCP] convert signed div/rem to unsigned for non-negative operands, 2nd try

The original commit ( fe1f3cfc2669 ) was reverted because it could
crash / assert when trying to fold a value that was replaced
by a constant. In that case, there might not be an entry for the
constant in the solver yet.

This version adds a check for that possibility along with tests to
exercise that pattern (they used to crash).

Original commit message:
This extends the transform added with D81756 to handle div/rem opcodes.
For example:
https://alive2.llvm.org/ce/z/cX6za6

This replicates part of what CVP already does, but the motivating example
from issue #57472 demonstrates a phase ordering problem - we convert
branches to select before CVP runs and miss the transform.

Differential Revision: https://reviews.llvm.org/D133198

23 months ago[mlir] Use std::size instead of llvm::array_lengthof
Joe Loser [Wed, 7 Sep 2022 00:01:39 +0000 (18:01 -0600)]
[mlir] Use std::size instead of llvm::array_lengthof

LLVM contains a helpful function for getting the size of a C-style
array: `llvm::array_lengthof`. This is useful prior to C++17, but not as
helpful for C++17 or later: `std::size` already has support for C-style
arrays.

Change call sites to use `std::size` instead.

Differential Revision: https://reviews.llvm.org/D133387

23 months ago[clangd] Add Macro Expansion to Hover
Qingyuan Zheng [Wed, 7 Sep 2022 15:44:32 +0000 (17:44 +0200)]
[clangd] Add Macro Expansion to Hover

This patch adds macro expansion preview to hover info. Basically, the refactor infrastructure for expanding macro is used for this purpose. The following steps are added to getHoverContents for macros:
1. calling AST.getTokens().expansionStartingAt(...) to get expanded tokens
2. calling reformat(...) to format expanded tokens

Some opinions are wanted:
1. Should we present macro expansion before definition in the hover card?
2. Should we truncate/ignore macro expansion if it's too long? For performance and presentation reason, it might not be a good idea to expand pages worth of tokens in hover card. If so, what's the preferred threshold?

Also, some limitation applies:
1. Expansion isn't available in macro definition/arguments as the refactor code action isn't either.

Differential Revision: https://reviews.llvm.org/D127082

23 months agoRemoving myself from the CODE_OWNERS.TXT file as I am no longer working
Eric Schweitz [Wed, 7 Sep 2022 15:40:55 +0000 (08:40 -0700)]
Removing myself from the CODE_OWNERS.TXT file as I am no longer working
on the flang project. Ownership reverts to Steve Scalpone for now.

23 months ago[clangd] Fix hover crashing on integral or enumeral casts
Georg Kotheimer [Mon, 5 Sep 2022 09:14:27 +0000 (11:14 +0200)]
[clangd] Fix hover crashing on integral or enumeral casts

When pretty printing the value of an expression, we cannot infer from
the type of the expression the type of the constant that the expression
evaluates to, as the expression might contain a type cast.

23 months ago[clangd] Improve Selection testcase, pin to C++17
Sam McCall [Wed, 7 Sep 2022 14:00:21 +0000 (16:00 +0200)]
[clangd] Improve Selection testcase, pin to C++17

17 vs 14 have different ASTs, this causes D131465 to have to touch this test.
While here, make sure we're being clear about *which* nodes we're matching.

Differential Revision: https://reviews.llvm.org/D133423

23 months ago[Sema] Move Diags.isIgnored() checks off hot paths, it's not free. NFC
Sam McCall [Wed, 13 Jul 2022 19:11:44 +0000 (21:11 +0200)]
[Sema] Move Diags.isIgnored() checks off hot paths, it's not free. NFC

This speeds up clangd's buildAST() (i.e. parsing with a preamble) by 5% on
clangd/AST.cpp, by avoiding filling up the diagnostic state map with entries for
all the files where templates are being instantiated from.

(I would assume it has a similar effect on PCH and modules compiles).

This approach is obviously pretty fragile, and we should find ways to make
isIgnored() cheaper instead. But these changes in particular don't seem to make
the code worse in any case.

Differential Revision: https://reviews.llvm.org/D129683

23 months agoOne-shot-bufferize: allow non-tensor arguments in scg.while/for.
Johannes Reifferscheid [Wed, 7 Sep 2022 12:35:50 +0000 (14:35 +0200)]
One-shot-bufferize: allow non-tensor arguments in scg.while/for.

Currently, one-shot-bufferize crashes as soon as there's
a mixture of tensor and non-tensor arguments. This seems
to happen for no good reason.

Reviewed By: springerm

Differential Revision: https://reviews.llvm.org/D133419

23 months ago[mlir][complex] Canonicalization for complex.sub adding same numbers
Kai Sasaki [Wed, 7 Sep 2022 12:55:28 +0000 (14:55 +0200)]
[mlir][complex] Canonicalization for complex.sub adding same numbers

Canonicalization for complex.sub adding same numbers. This canonicalization supports the case like complex.sub(complex.add(a, b), b) -> a.

Reviewed By: pifon2a

Differential Revision: https://reviews.llvm.org/D133397

23 months ago[mlir][Math] Fix RoundEven constant folder.
jacquesguan [Wed, 7 Sep 2022 12:45:57 +0000 (12:45 +0000)]
[mlir][Math] Fix RoundEven constant folder.

Use roundToIntegral instead roundeven of libm to avoid window build failed.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D133402

23 months ago[InstCombine] fold add+negate through select into sub
Sanjay Patel [Wed, 7 Sep 2022 11:53:38 +0000 (07:53 -0400)]
[InstCombine] fold add+negate through select into sub

This transform came up as a potential DAGCombine in D133282,
so I wanted to see how it escaped in IR too.

We do general folds in InstCombiner::SimplifySelectsFeedingBinaryOp()
by checking if either arm of a select simplifies when the trailing
binop is threaded into the select.

So as long as one side simplifies, it's a good fold to combine a
negate and add into 1 subtract.

This is an example with a zero arm in the select:
https://alive2.llvm.org/ce/z/Hgu_Tj

And this models the tests with a cancelling 'not' op:
https://alive2.llvm.org/ce/z/BuzVV_

Differential Revision: https://reviews.llvm.org/D133369

23 months ago[InstCombine] add tests for SimplifySelectsFeedingBinaryOp(); NFC
Sanjay Patel [Tue, 6 Sep 2022 17:24:27 +0000 (13:24 -0400)]
[InstCombine] add tests for SimplifySelectsFeedingBinaryOp(); NFC

23 months agoSink/hoist memory instructions between loop fusion candidates
Aaron Kogon [Wed, 7 Sep 2022 11:42:00 +0000 (07:42 -0400)]
Sink/hoist memory instructions between loop fusion candidates

Currently, instructions in the preheader of the second of two fusion
candidates are sunk and hoisted whenever possible, to try to allow the
loops to fuse. Memory instructions are skipped, and are never sunk or
hoisted. This change adds memory instructions for sinking/hoisting
consideration.

This change uses DependenceAnalysis to check if a mem inst in the
preheader of FC1 depends on an instruction in FC0's header, across
which it will be hoisted, or FC1's header, across which it will be
sunk. We reject cases where the dependency is a data hazard.

Differential Revision: https://reviews.llvm.org/D131606

23 months ago[mlir] Change CombiningKind in Vector dialect to EnumAttr.
Oleg Shyshkov [Wed, 7 Sep 2022 11:33:02 +0000 (13:33 +0200)]
[mlir] Change CombiningKind in Vector dialect to EnumAttr.

CombiningKind was implemented before EnumAttr, so it reimplements the same behaviour with the custom code. Except for a few places, EnumAttr is a drop-in replacement.

Reviewed By: nicolasvasilache, pifon2a

Differential Revision: https://reviews.llvm.org/D133343

23 months ago[MCA] Correctly check pipeline availability for partially overlapping resource groups.
Andrea Di Biagio [Wed, 7 Sep 2022 10:27:22 +0000 (11:27 +0100)]
[MCA] Correctly check pipeline availability for partially overlapping resource groups.

This patch mostly reverts commit 70b37f4c03c which fixed PR50725.

In case of explicit consumption of multiple partially overlapping group
resources, the ResourceManager was not correctly checking pipeline
esources availability.

The fix for PR50725 only partially addressed a few instances of that issue.
This is a more general (although, technically slower) fix for that same issue.

It also fixes Issue #57548

Thanks to Haohai Wen for the small reproducible.

23 months ago[clangd] Support renaming virtual methods
Tom Praschan [Wed, 7 Sep 2022 10:03:55 +0000 (12:03 +0200)]
[clangd] Support renaming virtual methods

Fixes https://github.com/clangd/clangd/issues/706

Differential Revision: https://reviews.llvm.org/D132797

23 months ago[CostModel][X86] Merge getTypeBasedIntrinsicInstrCost into getIntrinsicInstrCost
Simon Pilgrim [Wed, 7 Sep 2022 10:27:40 +0000 (11:27 +0100)]
[CostModel][X86] Merge getTypeBasedIntrinsicInstrCost into getIntrinsicInstrCost

For the few non type based intrinsic cases we can just check for !isTypeBasedOnly() to access the args directly.

I don't think we have a need to keep getTypeBasedIntrinsicInstrCost in BasicTTIImpl.h any more and can do a similar merge there as well - but it's a messier refactor and will take a while.

23 months ago[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3 instructions
Dmitry Preobrazhensky [Wed, 7 Sep 2022 10:55:27 +0000 (13:55 +0300)]
[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3 instructions

Differential Revision: https://reviews.llvm.org/D133350

23 months ago[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3.DPP8 instructions
Dmitry Preobrazhensky [Wed, 7 Sep 2022 10:51:31 +0000 (13:51 +0300)]
[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3.DPP8 instructions

Differential Revision: https://reviews.llvm.org/D133353

23 months ago[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3.DPP16 instructions
Dmitry Preobrazhensky [Wed, 7 Sep 2022 10:46:38 +0000 (13:46 +0300)]
[AMDGPU][MC][GFX11][NFC] Update disassembler tests for VOP3.DPP16 instructions

Differential Revision: https://reviews.llvm.org/D133356

23 months ago[Bitcode] Fix constexpr autoupgrade for arrays and structs
Nikita Popov [Wed, 7 Sep 2022 10:46:32 +0000 (12:46 +0200)]
[Bitcode] Fix constexpr autoupgrade for arrays and structs

While vectors use insertelement, structs and arrays should use
insertvalue.

23 months ago[AMDGPU][MC][GFX11][NFC] Update assembler tests for VOPD instructions
Dmitry Preobrazhensky [Wed, 7 Sep 2022 10:41:44 +0000 (13:41 +0300)]
[AMDGPU][MC][GFX11][NFC] Update assembler tests for VOPD instructions

Differential Revision: https://reviews.llvm.org/D133205

23 months ago[Bitcode] Convert constexpr-to-instr.ll to use bitcode input (NFC)
Nikita Popov [Wed, 7 Sep 2022 10:28:26 +0000 (12:28 +0200)]
[Bitcode] Convert constexpr-to-instr.ll to use bitcode input (NFC)

We can't use an IR input once the relevant constant expressions
are no longer supported. Use a bitcode file instead, which will
be auto-upgraded (the whole point of this code...)

23 months ago[AMDGPU] Add an operand folding test case from D114232
Jay Foad [Wed, 7 Sep 2022 10:16:40 +0000 (11:16 +0100)]
[AMDGPU] Add an operand folding test case from D114232

23 months ago[AMDGPU] Refactor SIFoldOperands. NFC.
Jay Foad [Wed, 7 Sep 2022 09:49:50 +0000 (10:49 +0100)]
[AMDGPU] Refactor SIFoldOperands. NFC.

Refactor static functions into class methods so they have access to TII, MRI
etc.

23 months ago[ConstantFold] Avoid unary ConstantExpr::get()
Nikita Popov [Wed, 7 Sep 2022 09:40:30 +0000 (11:40 +0200)]
[ConstantFold] Avoid unary ConstantExpr::get()

Call ConstantFoldUnaryInstruction() instead, to only produce a
result if it folds.

23 months agoApply clang-tidy fixes for performance-for-range-copy in TileUsingInterface.cpp ...
Mehdi Amini [Mon, 29 Aug 2022 11:08:54 +0000 (11:08 +0000)]
Apply clang-tidy fixes for performance-for-range-copy in TileUsingInterface.cpp (NFC)

23 months agoApply clang-tidy fixes for llvm-qualified-auto in Bufferize.cpp (NFC)
Mehdi Amini [Mon, 29 Aug 2022 11:06:51 +0000 (11:06 +0000)]
Apply clang-tidy fixes for llvm-qualified-auto in Bufferize.cpp (NFC)

23 months ago[AsmPrinter] Emit PCs into requested PCSections
Marco Elver [Tue, 6 Sep 2022 13:49:28 +0000 (15:49 +0200)]
[AsmPrinter] Emit PCs into requested PCSections

Interpret MD_pcsections in AsmPrinter emitting the requested metadata to
the associated sections. Functions and normal instructions are handled.

Differential Revision: https://reviews.llvm.org/D130879

23 months ago[GlobalISel] Propagate PCSections metadata to MachineInstr
Marco Elver [Tue, 6 Sep 2022 13:49:23 +0000 (15:49 +0200)]
[GlobalISel] Propagate PCSections metadata to MachineInstr

Propagate (most) PC sections metadata to MachineInstr when GlobalISel is
doing instruction selection.

This change results in support for architectures using GlobalISel (such
as -O0 with AArch64). Not all instructions may be supported yet, and
requires further target-specific handling (such as done for AArch64
pseudo-atomics). Expanding supported instructions is planned on a
case-by-case basis and new use cases for PC sections metadata.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D130886

23 months ago[AtomicExpandPass] Always copy pcsections Metadata to expanded atomics
Marco Elver [Tue, 6 Sep 2022 13:49:18 +0000 (15:49 +0200)]
[AtomicExpandPass] Always copy pcsections Metadata to expanded atomics

When expanding IR atomics to target-specific atomics, copy all
!pcsections Metadata to expanded atomics automatically.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D130885

23 months ago[FastISel] Propagate PCSections metadata to MachineInstr
Marco Elver [Tue, 6 Sep 2022 13:49:13 +0000 (15:49 +0200)]
[FastISel] Propagate PCSections metadata to MachineInstr

Propagate PC sections metadata to MachineInstr when FastISel is doing
instruction selection.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D130884

23 months ago[ConstantExpr] Don't create fneg expressions
Nikita Popov [Wed, 7 Sep 2022 08:56:36 +0000 (10:56 +0200)]
[ConstantExpr] Don't create fneg expressions

Don't create fneg expressions unless explicitly requested by IR or
bitcode.

23 months ago[MachineInstrBuilder] Introduce MIMetadata to simplify metadata propagation
Marco Elver [Tue, 6 Sep 2022 13:49:08 +0000 (15:49 +0200)]
[MachineInstrBuilder] Introduce MIMetadata to simplify metadata propagation

In many places DebugLoc and PCSections metadata are just copied along to
propagate them through MachineInstrs. Simplify doing so by bundling them
up in a MIMetadata class that replaces the DebugLoc argument to most
BuildMI() variants.

The DebugLoc-only constructors allow implicit construction, so that
existing usage of `BuildMI(.., DL, ..)` works as before, and the rest of
the codebase using BuildMI() does not require changes.

NFC.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D130883

23 months ago[SelectionDAG] Propagate PCSections through SDNodes
Marco Elver [Tue, 6 Sep 2022 13:49:03 +0000 (15:49 +0200)]
[SelectionDAG] Propagate PCSections through SDNodes

Add a new entry to SDNodeExtraInfo to propagate PCSections through
SelectionDAG.

Reviewed By: vitalybuka

Differential Revision: https://reviews.llvm.org/D130882

23 months ago[AMDGPU] Simplify mad/mac patterns. NFC.
Jay Foad [Mon, 5 Sep 2022 14:11:15 +0000 (15:11 +0100)]
[AMDGPU] Simplify mad/mac patterns. NFC.

Simplify instruction selection patterns for mad/mac:
- Use any_fmad consistently to make it clear that all patterns treat
  fmad and AMDGPUfmad_ftz identically.
- For mad, put the patterns on the instruction definitions. For mac the
  patterns are still out-of-line because we want to set AddedComplexity
  and to have special handling of the source modifiers.

Differential Revision: https://reviews.llvm.org/D133305

23 months ago[TableGen] Document sequence with stride
Jay Foad [Mon, 5 Sep 2022 12:20:11 +0000 (13:20 +0100)]
[TableGen] Document sequence with stride

Document (in comments) the optional fourth "stride" argument to the
sequence operator, which was added in svn r157416.

Differential Revision: https://reviews.llvm.org/D133297

23 months ago[AMDGPU] Add a comment for a missing fold
Jay Foad [Wed, 7 Sep 2022 08:56:55 +0000 (09:56 +0100)]
[AMDGPU] Add a comment for a missing fold

23 months ago[Reassociate] Avoid ConstantExpr::getFNeg() (NFCI)
Nikita Popov [Wed, 7 Sep 2022 08:46:58 +0000 (10:46 +0200)]
[Reassociate] Avoid ConstantExpr::getFNeg() (NFCI)

Use ConstantFoldUnaryOpOperand() instead. Also make the code below
robust against non-instruction users, just in case it doesn't fold.

23 months ago[mlir] Remove `materializeOpFoldResult` functions.
Alexander Belyaev [Wed, 7 Sep 2022 08:10:48 +0000 (10:10 +0200)]
[mlir] Remove `materializeOpFoldResult` functions.

We can use `getValueOrCreateConstantIndexOp` instead.

Differential Revision: https://reviews.llvm.org/D133403

23 months ago[AArch64] add tests for non-power2 int types; NFC
chenglin.bi [Wed, 7 Sep 2022 08:07:26 +0000 (16:07 +0800)]
[AArch64] add tests for non-power2 int types; NFC

23 months ago[clang][doc] Do not keep a copy of ClangCommandLineReference.rst in tree
serge-sans-paille [Tue, 6 Sep 2022 12:22:40 +0000 (14:22 +0200)]
[clang][doc] Do not keep a copy of ClangCommandLineReference.rst in tree

This file is auto-generated, it's a bit confusing for the maintainers to have it
in tree while it shouldn't be modified *and* already have a rule to be
generated.

Differential Revision: https://reviews.llvm.org/D133349

23 months ago[MLIR] NFC. Introduce mlir::hasEffect and refactor usages dialect util
Uday Bondhugula [Wed, 7 Sep 2022 04:39:47 +0000 (10:09 +0530)]
[MLIR] NFC. Introduce mlir::hasEffect and refactor usages dialect util

Introduce mlir::hasEffect and refactor existing usage to use utility.
NFC.

Reviewed By: rriddle, mehdi_amini

Differential Revision: https://reviews.llvm.org/D132117

23 months ago[CSKY] Fix the compiling error about missing Log2 function with Log2_64
Zi Xuan Wu (Zeson) [Wed, 7 Sep 2022 06:40:01 +0000 (14:40 +0800)]
[CSKY] Fix the compiling error about missing Log2 function with Log2_64

23 months ago[X86][NFC] Refine load/store reg to StackSlot for extensibility
Xiang1 Zhang [Wed, 31 Aug 2022 03:16:09 +0000 (11:16 +0800)]
[X86][NFC] Refine load/store reg to StackSlot for extensibility

Reviewed By: LuoYuanke

Differential Revision: https://reviews.llvm.org/D133078

23 months ago[LLD][COFF] Fix writing a map file when range extension thunks are inserted
Jan Ole Hüser [Wed, 7 Sep 2022 06:17:10 +0000 (09:17 +0300)]
[LLD][COFF] Fix writing a map file when range extension thunks are inserted

Bug: An assertion fails:

    Assertion failed: isa<To>(Val) && "cast<Ty>() argument of incompatible type!",
    file C:\Users\<user>\prog\llvm\llvm-git-lld-bug\llvm\include\llvm/Support/Casting.h, line 578

Bug is triggered, if

    - a map file is requested with /MAP, and
    - Architecture is ARMv7, Thumb, and
    - a relative jump (branch instruction) is greater than 16 MiB (2^24)

The reason for the Bug is:

    - a Thunk is created for the jump
    - a Symbol for the Thunk is created
        - of type `DefinedSynthetic`
        - in file `Writer.cpp`
        - in function `getThunk`
    - the Symbol has no name
    - when creating the map file, the name of the Symbol is queried
    - the function `Symbol::computeName` of the base class `Symbol`
      casts the `this` pointer to type `DefinedCOFF` (a derived type),
      but the acutal type is `DefinedSynthetic`
    - The in the llvm::cast an assertion fails

Changes:

- Modify regression test to trigger this bug
- Give the symbol pointing to the thunk a name, to fix the bug
- Add assertion, that only DefinedCOFF symbols are allowed to have an
  empty name, when the constructor of the base class Symbol is executed

Reviewed By: rnk

Differential Revision: https://reviews.llvm.org/D133201

23 months ago[compiler-rt] [test] Handle missing ld.gold gracefully
Michał Górny [Tue, 6 Sep 2022 13:40:10 +0000 (15:40 +0200)]
[compiler-rt] [test] Handle missing ld.gold gracefully

Fix the is_binutils_lto_supported() function to handle missing
executables gracefully.  Currently, the function does not catch
exceptions from subprocess.Popen() and therefore causes lit to crash
if config.gold_executable does not specify a valid executable:

```
lit: /usr/lib/python3.11/site-packages/lit/TestingConfig.py:136: fatal: unable to parse config file '/tmp/portage/sys-libs/compiler-rt-
15.0.0/work/compiler-rt/test/lit.common.cfg.py', traceback: Traceback (most recent call last):
  File "/usr/lib/python3.11/site-packages/lit/TestingConfig.py", line 125, in load_from_path
    exec(compile(data, path, 'exec'), cfg_globals, None)
  File "/tmp/portage/sys-libs/compiler-rt-15.0.0/work/compiler-rt/test/lit.common.cfg.py", line 561, in <module>
    if is_binutils_lto_supported():
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/tmp/portage/sys-libs/compiler-rt-15.0.0/work/compiler-rt/test/lit.common.cfg.py", line 543, in is_binutils_lto_supported
    ld_cmd = subprocess.Popen([exe, '--help'], stdout=subprocess.PIPE, env={'LANG': 'C'})
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.11/subprocess.py", line 1022, in __init__
    self._execute_child(args, executable, preexec_fn, close_fds,
  File "/usr/lib/python3.11/subprocess.py", line 1899, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: 'GOLD_EXECUTABLE-NOTFOUND'
```

Differential Revision: https://reviews.llvm.org/D133358

23 months ago[NFC][sancov] Rename ModuleSanitizerCoveragePass
Vitaly Buka [Wed, 7 Sep 2022 03:55:39 +0000 (20:55 -0700)]
[NFC][sancov] Rename ModuleSanitizerCoveragePass

23 months ago[NFC][msan] Rename ModuleMemorySanitizerPass
Vitaly Buka [Tue, 6 Sep 2022 05:30:23 +0000 (22:30 -0700)]
[NFC][msan] Rename ModuleMemorySanitizerPass