platform/upstream/llvm.git
14 months ago[LLVM][Uniformity] Improve detection of uniform registers
Sameer Sahasrabuddhe [Tue, 16 May 2023 04:07:04 +0000 (09:37 +0530)]
[LLVM][Uniformity] Improve detection of uniform registers

The MachineUA now queries the target to determine if a given register holds a
uniform value. This is determined using the corresponding register bank if
available, or by a combination of the register class and value type. This
assumes that the target is optimizing for performance by choosing registers, and
the target is responsible for any mismatch with the inferred uniformity.

For example, on AMDGPU, an SGPR is now treated as uniform, except if the
register bank is VCC (i.e., the register holds a wave-wide vector of 1-bit
values) or equivalently if it has a value type of s1.

 - This does not always work with inline asm, where the register bank or the
   value type might not be present. We assume that the SGPR is uniform, because
   it is not expected to be s1 in the vast majority of cases.
 - The pseudo branch instruction SI_LOOP is now hard-coded to be always
   divergent, although its condition is an SGPR.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D150438

14 months ago[SimpleLoopUnswitch] Skip trivial select conds for selects
Joshua Cao [Tue, 16 May 2023 03:28:40 +0000 (20:28 -0700)]
[SimpleLoopUnswitch] Skip trivial select conds for selects

Fixes https://github.com/llvm/llvm-project/issues/62715

If a select's condition is a trivial select:
```
%s = select %cond, i1 true, i1 false
```
Unswitch on %cond, rather than %s. This fixes crashes where there is a
disparity in finding candidates and and the transformation logic.

14 months ago[Docs][llvm-exegesis] Specify platform support for different modes
Aiden Grossman [Tue, 16 May 2023 03:19:12 +0000 (03:19 +0000)]
[Docs][llvm-exegesis] Specify platform support for different modes

llvm-exegesis has both a capture mode and an analysis mode that can be
used independently of each other. This patch makes it clear that
analysis mode will work on other platforms that LLVM supports in the
documentation which was unclear before.

Reviewed By: courbet

Differential Revision: https://reviews.llvm.org/D150536

14 months ago[NFC] [C++20] [Modules] Rename ASTContext::getNamedModuleForCodeGen to ASTContext...
Chuanqi Xu [Tue, 16 May 2023 03:22:02 +0000 (11:22 +0800)]
[NFC] [C++20] [Modules] Rename ASTContext::getNamedModuleForCodeGen to ASTContext::getCurrentNamedModule

The original name "ASTContext::getNamedModuleForCodeGen" is not properly
reflecting the usage of the interface. This interface can be used to
judge the current module unit in both sema analysis and code generation.
So the original name was not so correct.

14 months ago[clang-format] Stop comment disrupting indentation of Verilog ports
sstwcw [Tue, 16 May 2023 02:50:07 +0000 (02:50 +0000)]
[clang-format] Stop comment disrupting indentation of Verilog ports

Before:

```
module x
    #( //
        parameter x)
    ( //
        input y);
endmodule
```

After:

```
module x
    #(//
      parameter x)
    (//
     input y);
endmodule
```

If the first line in a port or parameter list is not a comment, the
following lines will be aligned to the first line as intended:

```
module x
    #(parameter x1,
      parameter x2)
    (input y,
     input y2);
endmodule
```

Previously, the indentation would be changed to an extra continuation
indentation relative to the start of the parenthesis or the hash if
the first token inside the parentheses was a comment.  It is a feature
introduced in ddaa9be97839.  The feature enabled one to insert a `//`
comment right after an opening parentheses to put the function
arguments on a new line with a small indentation regardless of how
long the function name is, like this:

```
someFunction(anotherFunction( // Force break.
    parameter));
```

People are unlikely to use this feature in a Verilog port list because
the formatter already puts the port list on its own lines.  A comment
at the start of a port list is probably a comment for the port on the
next line.

We also removed the space before the comment so that its indentation
would be same as that for a line comment anywhere else in the port
list.

Reviewed By: HazardyKnusperkeks

Differential Revision: https://reviews.llvm.org/D149562

14 months ago[clangd] Fix test failure when it's built with compiler flags unknown by clang
Xi Ruoyao [Tue, 16 May 2023 02:56:14 +0000 (03:56 +0100)]
[clangd] Fix test failure when it's built with compiler flags unknown by clang

If LLVM is built with a compiler other than clang, the `compile_commands.json`
file may contain compiler flags unknown by clang.  When a clangd test is copied
into the build directory and checked, clangd will pick the unknown flag from
the file and cause a test failure.  Create an empty `compile_commands.json` in
the test directory nested in the build directory to override it.

Reviewed By: thesamesam

Differential Revision: https://reviews.llvm.org/D150582

14 months agoRevert "[NFC] [C++20] [Modules] Refactor Sema::isModuleUnitOfCurrentTU into"
Chuanqi Xu [Tue, 16 May 2023 02:36:08 +0000 (10:36 +0800)]
Revert "[NFC] [C++20] [Modules] Refactor Sema::isModuleUnitOfCurrentTU into"

This reverts commit f109b1016801e2b0dbee278f3c517057c0b1d441 as required
in
https://github.com/llvm/llvm-project/commit/f109b1016801e2b0dbee278f3c517057c0b1d441#commitcomment-113477829.

14 months ago[NFC] [C++20] [Modules] Refactoring b6c7177145bc to make it not
Chuanqi Xu [Tue, 16 May 2023 02:19:27 +0000 (10:19 +0800)]
[NFC] [C++20] [Modules] Refactoring b6c7177145bc to make it not
dependent on f109b10

Given
https://github.com/llvm/llvm-project/commit/f109b1016801e2b0dbee278f3c517057c0b1d441#commitcomment-113477829,
we need to revert f109b10. So it will be better to make this patch not
dependent on f109b10 as much as possible.

14 months ago[flang] Add check for constraints on event-stmts
Katherine Rasmussen [Tue, 16 May 2023 01:08:17 +0000 (18:08 -0700)]
[flang] Add check for constraints on event-stmts

In the CoarrayChecker, add checks for the constraints C1177 and
C1178 for event-wait-stmt. Add event-post-stmt to the check
for the constraints for sync-stat-list. Add a check for the
constraint C1176 on event-variable.

Reviewed By: PeteSteinfeld

Differential Revision: https://reviews.llvm.org/D137204

14 months ago[RISCV][MC] .debug_line/.debug_frame/.eh_frame: emit relocations for assembly input...
Fangrui Song [Tue, 16 May 2023 01:44:55 +0000 (18:44 -0700)]
[RISCV][MC] .debug_line/.debug_frame/.eh_frame: emit relocations for assembly input files with relaxation

When assembling `.debug_line` for both explicitly specified and synthesized
`.loc` directives. the integrated assembler may incorrectly omit relocations for
-mrelax.

For an assembly file, we have a `MCAssembler` object and `evaluateAsAbsolute`
will incorrectly fold `AddrDelta` to a constant (which is not true in the
presence of linker relaxation).
`MCDwarfLineAddr::Emit` will emit a special opcode, which does not take into
account of linker relaxation. This is a sufficiently complex function that
I think should be called in any "fast paths" for linker relaxation aware assembling.

The following script demonstrates the bugs.

```
cat > x.c <<eof
void f();
void _start() {
  f();
  f();
  f();
}
eof
# C to object file: correct DW_LNS_fixed_advance_pc
clang --target=riscv64 -g -c x.c
llvm-dwarfdump --debug-line -v x.o | grep \ DW_LNS_fixed_advance_pc -q

# Assembly to object file with synthesized line number information: incorrect special opcodes
clang --target=riscv64 -S x.c && clang --target=riscv64 -g -c x.s
llvm-dwarfdump --debug-line -v x.o | grep \ DW_LNS_fixed_advance_pc -q; test $? -eq 1

# Assembly with .loc to object file: incorrect special opcodes
clang --target=riscv64 -S -g x.c && clang --target=riscv64 -c x.s
llvm-dwarfdump --debug-line -v x.o | grep \ DW_LNS_fixed_advance_pc -q; test $? -eq 1
```

The `MCDwarfLineAddr::Emit` code path is an old optimization in commit
57ab708bdd3231b23a8ef4978b11ff07616034a2 (2010) that seems no longer relevant.
It don't trigger for direct machine code emission (label differences are not
foldable without a `MCAssembler`). MCDwarfLineAddr::Emit does complex operations
that are repeated in MCAssembler::relaxDwarfLineAddr, which an intricate RISCV
override.

Let's remove the "fast path". Assembling the assembly output of
X86ISelLowering.cpp with `-g` may be 2% slower, but I think the cost is fine.
There are opportunities to make the "slow path" faster, e.g.

* Optimizing the current new MC*Fragment pattern that allocates new fragments on
  the heap.
* Reducing the number of relaxation times for .debug_line and .debug_frame, as
  well as possibly other sections using LEB128. For instance, LEB128 can have a
  one-byte estimate to avoid the second relaxation iteration.

For assembly input with -mno-relax, in theory we can prefer special opcodes to
DW_LNS_fixed_advance_pc to decrease the size of .debug_line, but such a change
may be overkill and unnecessarily diverge from -mrelax behaviors and GCC.

---

For .debug_frame/.eh_frame, MCDwarf currently emits DW_CFA_advance_loc without
relocations. Remove the special case to enable relocations. Similar to
.debug_line, even without the bug fix, the MCDwarfFrameEmitter::encodeAdvanceLoc
special case is a sufficiently complex code path that should be avoided.

---

When there are more than one section, we generate .debug_rnglists for
DWARF v5. We currently emit DW_RLE_start_length using ULEB128, which
is incorrect. The new test gen-dwarf.s adds a TODO.

---

About other `evaluateAsAbsolute` uses. `MCObjectStreamer::emit[SU]LEB128Value`
have similar code to MCDwarfLineAddr. They are fine to keep as we don't have
LEB128 relocations to correctly represent link-time non-constants anyway.

---

In the future, we should investigate ending a MCFragment for a relaxable
instruction, to further clean up the assembler support for linker relaxation
and fix `evaluateAsAbsolute`.

See bbea64250f65480d787e1c5ff45c4de3ec2dcda8 for some of the related code.

Reviewed By: enh, barannikov88

Differential Revision: https://reviews.llvm.org/D150004

14 months ago[mlir][sparse] change runners to c_runners
Aart Bik [Tue, 16 May 2023 00:54:47 +0000 (17:54 -0700)]
[mlir][sparse] change runners to c_runners

Reviewed By: Peiming

Differential Revision: https://reviews.llvm.org/D150628

14 months ago[mlir][tosa] Fold consecutive negate as no-op
Kai Sasaki [Tue, 16 May 2023 00:41:28 +0000 (09:41 +0900)]
[mlir][tosa] Fold consecutive negate as no-op

Consecutive element-wise negate should be canonicalized as no-op.

Reviewed By: eric-k256

Differential Revision: https://reviews.llvm.org/D150518

14 months ago[RISCV] Support vreinterpret intrinsics between vector boolean type and m1 vector...
eopXD [Tue, 2 May 2023 10:16:44 +0000 (03:16 -0700)]
[RISCV] Support vreinterpret intrinsics between vector boolean type and m1 vector integer type

Link to specification: [riscv-non-isa/rvv-intrinsic-doc#221](https://github.com/riscv-non-isa/rvv-intrinsic-doc/pull/221)

Reviewed By: craig.topper

Differential Revision: https://reviews.llvm.org/D149642

14 months agoRevert "[MergeICmps] Adapt to non-eq comparisons, bugfix"
Arthur Eubanks [Tue, 16 May 2023 00:44:53 +0000 (17:44 -0700)]
Revert "[MergeICmps] Adapt to non-eq comparisons, bugfix"

This reverts commit ae337ed5951c896164e07618d651d086f978ff2c.

Still causes miscompiles, see D141188.

14 months ago[clang][deps] Do not cache PCM files
Jan Svoboda [Mon, 15 May 2023 21:26:10 +0000 (14:26 -0700)]
[clang][deps] Do not cache PCM files

On incremental scan, caching an out-of-date PCM on the VFS layer causes each TU and each module to recompile the PCM again. This is huge performance problem. Stop caching ".pcm" files.

Reviewed By: Bigcheese

Differential Revision: https://reviews.llvm.org/D150615

14 months agoRevert "[clang][deps] Only cache files with specific extension"
Jan Svoboda [Mon, 15 May 2023 21:24:08 +0000 (14:24 -0700)]
Revert "[clang][deps] Only cache files with specific extension"

This reverts commit d1e00b6f136ec71a4c95a7eb4fd81ec0ab547962.

Internally, there were issues with caching stat failures for .framework directories. We need some time for investigation to pinpoint what exactly was going wrong.

14 months ago[lldb] Fix lua build after 27b6a4e63afe
Alex Langford [Mon, 15 May 2023 23:06:50 +0000 (16:06 -0700)]
[lldb] Fix lua build after 27b6a4e63afe

This applies the same trick for Lua that I did for python in
27b6a4e63afe.

Differential Revision: https://reviews.llvm.org/D150624

14 months ago[StructuralHash] Track global variables
Arthur Eubanks [Mon, 15 May 2023 23:57:10 +0000 (16:57 -0700)]
[StructuralHash] Track global variables

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D149209

14 months ago[llvm-exegesis] Create a proper LLVM IR Function for MachineFunctions
Arthur Eubanks [Mon, 15 May 2023 23:50:56 +0000 (16:50 -0700)]
[llvm-exegesis] Create a proper LLVM IR Function for MachineFunctions

I have upcoming changes break with invalid Function definitions.

Reviewed By: aidengrossman, courbet

Differential Revision: https://reviews.llvm.org/D149207

14 months agoFix MLIR build (typo in VectorOps.cpp)
Mehdi Amini [Fri, 12 May 2023 21:58:25 +0000 (22:58 +0100)]
Fix MLIR build (typo in VectorOps.cpp)

14 months ago[mlir][spirv] Check type legality using converter for vectors
Lei Zhang [Mon, 15 May 2023 22:36:08 +0000 (22:36 +0000)]
[mlir][spirv] Check type legality using converter for vectors

This allows `index` vectors to be converted to SPIR-V.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D150616

14 months ago[CMake] Use LLVM own tools in extract_symbols.py
Igor Kudrin [Sat, 22 Apr 2023 03:01:10 +0000 (20:01 -0700)]
[CMake] Use LLVM own tools in extract_symbols.py

As for now, 'extract_symbols.py' can use several tools to extract
symbols from object files and libraries and to guess if the target is
32-bit Windows. The tools are being found via PATH, so in most cases,
they are just system tools. This approach has a number of limitations,
in particular:

* System tools may not be able to handle the target format in case of
  cross-platform builds,
* They cannot read symbols from LLVM bitcode files, so the staged LTO
  build with plugins is not supported,
* The auto-selected tools may be suboptimal (see D113557),
* Support for multiple tools for a single task increases the complexity
  of the script code.

The patch proposes using LLVM's own tools to solve these issues.
Specifically, 'llvm-readobj' detects the target platform, and 'llvm-nm'
reads symbols from all supported formats, including bitcode files. The
tools can be built in Release mode for the host platform or overridden
using CMake settings 'LLVM_READOBJ' and 'LLVM_NM' respectively. The
implementation also supports using precompiled tools via
'LLVM_NATIVE_TOOL_DIR'.

Differential Revision: https://reviews.llvm.org/D149119

14 months ago[llvm-reduce] Only write reduced output after we've saved it
Arthur Eubanks [Mon, 15 May 2023 22:45:54 +0000 (15:45 -0700)]
[llvm-reduce] Only write reduced output after we've saved it

Otherwise we're just rewriting the currently saved module instead of the newly reduced one.

Doesn't affect the final output since we separately write that at the end.

14 months agoAdopt Properties to store operations inherent Attributes in the X86Vector dialect
Mehdi Amini [Sat, 22 Apr 2023 01:08:16 +0000 (19:08 -0600)]
Adopt Properties to store operations inherent Attributes in the X86Vector dialect

This is part of an on-going migration to adopt Properties inside MLIR.

Differential Revision: https://reviews.llvm.org/D148976

14 months agoAdopt Properties to store operations inherent Attributes in the Vector dialect
Mehdi Amini [Fri, 21 Apr 2023 07:17:20 +0000 (01:17 -0600)]
Adopt Properties to store operations inherent Attributes in the Vector dialect

This is part of an on-going migration to adopt Properties inside MLIR.

Differential Revision: https://reviews.llvm.org/D148939

14 months agoAdopt Properties to store operations inherent Attributes in the Transform dialect
Mehdi Amini [Fri, 21 Apr 2023 07:13:20 +0000 (01:13 -0600)]
Adopt Properties to store operations inherent Attributes in the Transform dialect

This is part of an on-going migration to adopt Properties inside MLIR.

Differential Revision: https://reviews.llvm.org/D148938

14 months ago[MachineOutliner] NFC: Add debug output to MachineOutliner::outline
Jessica Paquette [Mon, 15 May 2023 22:25:12 +0000 (15:25 -0700)]
[MachineOutliner] NFC: Add debug output to MachineOutliner::outline

Add some debug output to `outline` to assist in debugging + understanding the
code.

This will say

- How many things we found worth turning into outlined functions
- Whether or not candidates were pruned via the outlining algorithm
- The function created (if it was created)
- Where the calls were inserted
- What instruction was used to create the call

Sample output below:

```
NUMBER OF POTENTIAL FUNCTIONS: 5
WALKING FUNCTION LIST
PRUNED: 0/2 candidates
OUTLINE: Expected benefit (12 B) > threshold (1 B)
NEW FUNCTION: OUTLINED_FUNCTION_0
CREATE OUTLINED CALLS
  CALL: OUTLINED_FUNCTION_0 in bar:<unknown>
   .. BL @OUTLINED_FUNCTION_0, implicit-def $lr, implicit $sp
  CALL: OUTLINED_FUNCTION_0 in bar:<unknown>
   .. BL @OUTLINED_FUNCTION_0, implicit-def $lr, implicit $sp
PRUNED: 2/2 candidates
SKIP: Expected benefit (0 B) < threshold (1 B)
PRUNED: 0/2 candidates
OUTLINE: Expected benefit (8 B) > threshold (1 B)
NEW FUNCTION: OUTLINED_FUNCTION_1
CREATE OUTLINED CALLS
  CALL: OUTLINED_FUNCTION_1 in bar:<unknown>
   .. BL @OUTLINED_FUNCTION_1, implicit-def $lr, implicit $sp
  CALL: OUTLINED_FUNCTION_1 in bar:<unknown>
   .. BL @OUTLINED_FUNCTION_1, implicit-def $lr, implicit $sp
PRUNED: 2/2 candidates
SKIP: Expected benefit (0 B) < threshold (1 B)
PRUNED: 2/2 candidates
SKIP: Expected benefit (0 B) < threshold (1 B)
```

14 months ago[lldb] Refine call to decl printing helper (NFC)
Dave Lee [Mon, 8 May 2023 17:13:33 +0000 (10:13 -0700)]
[lldb] Refine call to decl printing helper (NFC)

When `ValueObjectPrinter` calls its `m_decl_printing_helper`, not all state is passed to
the helper. In particular, the helper doesn't have access to `m_curr_depth`, and thus
can't act on the logic within `ShouldShowName`.

To address this, this change passes in a modified copy of `m_options`. The modified copy
has has `m_hide_name` set according to the results of `ShouldShowName`. This allows
helper functions to know whether the name should be shown or hidden, without having
access to `ValueObjectPrinter`'s full state.

This is NFC in mainline lldb, as the only decl printing helper doesn't make use of this.
However in swift-lldb at least, there are decl printing helpers that do need this
information passed to them. See https://github.com/apple/llvm-project/pull/6795 where a
test is also included.

Differential Revision: https://reviews.llvm.org/D150129

14 months ago[TOSA] Add QuantizationDialect to TOSA's dependentDialects
Tai Ly [Mon, 15 May 2023 22:08:29 +0000 (15:08 -0700)]
[TOSA] Add QuantizationDialect to TOSA's dependentDialects

This adds QuantizationDialect to the dependent dialects of TOSA
This fixes the intermittent bug when creating uniform quantized type when none was parsed in.

LLVM ERROR: can't create type 'mlir::quant::UniformQuantizedType' because storage uniquer isn't initialized: the dialect was likely not loaded, or the type wasn't added with addTypes<...>() in the Dialect::initialize() method.

This happens, for example, in convert-tfl-uint8 pass when trying to create uniform quantized type i8 with zero-point=-128 to convert from ui8 type.

Signed-off-by: Tai Ly <tai.ly@arm.com>
Change-Id: I204248a45fd728d0cec8dc20214cb0b74de81e7b

Reviewed By: eric-k256

Differential Revision: https://reviews.llvm.org/D149354

14 months ago[gn build] Port 7ace54e64bb6
LLVM GN Syncbot [Mon, 15 May 2023 21:49:25 +0000 (21:49 +0000)]
[gn build] Port 7ace54e64bb6

14 months ago[libc++][PSTL] Implement std::copy{,_n}
Nikolas Klauser [Mon, 15 May 2023 17:38:01 +0000 (10:38 -0700)]
[libc++][PSTL] Implement std::copy{,_n}

Reviewed By: ldionne, #libc

Spies: jloser, libcxx-commits

Differential Revision: https://reviews.llvm.org/D149706

14 months ago[Demangle] fix comment NFC
Nick Desaulniers [Mon, 15 May 2023 21:32:38 +0000 (14:32 -0700)]
[Demangle] fix comment NFC

The second and third parameter of itaniumDemangle were removed in
commit 7277a72b908d ("[Demangle] remove unused params of itaniumDemangle")
Update a comment to reflect this.

Reviewed By: nathanchance

Differential Revision: https://reviews.llvm.org/D149975

14 months ago[lldb] Set CMAKE_CXX_STANDARD before including LLDBStandalone
Jonas Devlieghere [Mon, 15 May 2023 21:29:20 +0000 (14:29 -0700)]
[lldb] Set CMAKE_CXX_STANDARD before including LLDBStandalone

Set the C++ language standard before including LLDBStandalone.cmake.
Otherwise we risk building some of our dependencies (such as llvm_gtest)
without C++ 17 support.

This should fix the standalone bot [1] which is currently failing with the
following error:

  test-port.h:841:12: error: no member named 'tuple' in namespace 'std'
  using std::tuple;

[1] https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake-standalone

14 months ago[lldb] Change definition of DisassemblerCreateInstance
Alex Langford [Wed, 10 May 2023 00:04:37 +0000 (17:04 -0700)]
[lldb] Change definition of DisassemblerCreateInstance

DissassemblerCreateInstance is a function pointer whos return type is
`Disassembler *`. But Disassembler::FindPlugin always returns a
DisassemblerSP, so there's no reason why we can't just create a
DisassemblerSP in the first place.

Differential Revision: https://reviews.llvm.org/D150235

14 months ago[clang][modules] NFC: Only sort interesting identifiers
Jan Svoboda [Mon, 15 May 2023 20:28:07 +0000 (13:28 -0700)]
[clang][modules] NFC: Only sort interesting identifiers

In 9c254184 `ASTWriter` stopped writing identifiers that are not interesting. Taking it a bit further, we don't need to sort the whole identifier table, just the interesting identifiers. This reduces the size of sorted vector from ~10k (including lots of builtins) to 2 (`__VA_ARGS__` and `__VA_OPT__`) in a typical Xcode project, improving `clang-scan-deps` performance.

Reviewed By: benlangmuir

Differential Revision: https://reviews.llvm.org/D150494

14 months ago[LLDB] Fix TestDataFormatterSynthVal.py for AArch64/Windows
Muhammad Omair Javaid [Mon, 15 May 2023 20:09:13 +0000 (00:09 +0400)]
[LLDB] Fix TestDataFormatterSynthVal.py for AArch64/Windows

Since 44363f2 various tests have started passing but introduced a
expression evaluation failure in TestDataFormatterSynthVal.py.
This patch marks the expression evaluation part as skipped while rest
of the test passes.
This patch aslo introduces a new helper isAArch64Windows in lldbtest.py.

14 months agoasan-rt: Silence a few -Wformat=pedantic's in asan_mac.cpp
Jon Roelofs [Mon, 15 May 2023 19:57:39 +0000 (12:57 -0700)]
asan-rt: Silence a few -Wformat=pedantic's in asan_mac.cpp

Differential revision: https://reviews.llvm.org/D150604

14 months ago[clang-tidy] Extract areStatementsIdentical
Piotr Zegar [Mon, 15 May 2023 19:23:56 +0000 (19:23 +0000)]
[clang-tidy] Extract areStatementsIdentical

Move areStatementsIdentical from BranchCloneCheck into ASTUtils.
Add small improvments. Use it in LoopConvertUtils.

Reviewed By: carlosgalvezp

Differential Revision: https://reviews.llvm.org/D148995

14 months agoRevert "Emit the correct flags for the PROC CodeView Debug Symbol"
Muhammad Omair Javaid [Mon, 15 May 2023 19:13:19 +0000 (23:13 +0400)]
Revert "Emit the correct flags for the PROC CodeView Debug Symbol"

This reverts commit e48826e016e2f427f3b7b1274166aa9aa0ea7f4f.

https://lab.llvm.org/buildbot/#/builders/219/builds/2520

ldb-shell :: SymbolFile/PDB/function-nested-block.test

Differential Revision: https://reviews.llvm.org/D148761

14 months ago[libc++] Revert moving the pre-release checklist
Louis Dionne [Mon, 15 May 2023 19:35:15 +0000 (12:35 -0700)]
[libc++] Revert moving the pre-release checklist

I had not seen https://reviews.llvm.org/D150585 which supersedes
this, and I want to avoid merge conflicts for D150585.

14 months ago[clang] Fix emitVoidPtrVAArg for non-zero default alloca address space
Jessica Clarke [Mon, 15 May 2023 19:26:49 +0000 (20:26 +0100)]
[clang] Fix emitVoidPtrVAArg for non-zero default alloca address space

Indirect arguments are passed on the stack and so va_arg should use the
default alloca address space, not hard-code 0, for pointers to those.
The only in-tree target with a non-zero default alloca address space is
AMDGPU, but that does not support variadic arguments, so we cannot test
this upstream. However, downstream in CHERI LLVM (and Morello LLVM, a
further fork of that) we have targets that do both and so require this
change.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D132247

14 months agoFix ConstShapeOp::inferReturnTypes to be resilient to lack of properties
Mehdi Amini [Mon, 15 May 2023 19:12:28 +0000 (12:12 -0700)]
Fix ConstShapeOp::inferReturnTypes to be resilient to lack of properties

The Python bindings test aren't using properties yet, this is a bit
of a hack to support this here, but hopefully it'll be temporary.

14 months agoAdd an operator == and != to properties, use it in DuplicateFunctionElimination
Mehdi Amini [Mon, 15 May 2023 18:03:24 +0000 (11:03 -0700)]
Add an operator == and != to properties, use it in DuplicateFunctionElimination

Differential Revision: https://reviews.llvm.org/D150596

14 months agoRe-land "[-Wunsafe-buffer-usage] Remove an unnecessary const-qualifier"
ziqingluo-90 [Mon, 15 May 2023 18:55:35 +0000 (11:55 -0700)]
Re-land "[-Wunsafe-buffer-usage] Remove an unnecessary const-qualifier"

Re-land 7a0900fd3e2d34bc1d513a97cf8fbdc1754252d7, which includes too
much clang-format changes.  This re-land gets rid of the format changes.

14 months ago[docs] Use doxygen to describe the field `StartAtCycle`. [NFCI]
Francesco Petrogalli [Mon, 15 May 2023 13:49:04 +0000 (15:49 +0200)]
[docs] Use doxygen to describe the field `StartAtCycle`. [NFCI]

Differential Revision: https://reviews.llvm.org/D150572

14 months agoRevert "[lldb] Refactor SBFileSpec::GetDirectory"
Muhammad Omair Javaid [Mon, 15 May 2023 10:25:52 +0000 (14:25 +0400)]
Revert "[lldb] Refactor SBFileSpec::GetDirectory"

This reverts commit 2bea2d7b070dc5df723ce2b92dbc654b8bb1847e.

It introduced following failures on buildbot lldb-aarch64-windows:

lldb-api :: functionalities/process_save_core/TestProcessSaveCore.py
lldb-api :: python_api/symbol-context/TestSymbolContext.py

Differential Revision: https://reviews.llvm.org/D149625

14 months agoRevert "[-Wunsafe-buffer-usage] Remove an unnecessary const-qualifier"
ziqingluo-90 [Mon, 15 May 2023 18:25:52 +0000 (11:25 -0700)]
Revert "[-Wunsafe-buffer-usage] Remove an unnecessary const-qualifier"

This reverts commit 7a0900fd3e2d34bc1d513a97cf8fbdc1754252d7.

The commit includes too much clang-format changes.

14 months agoCleanup uses of getAttrDictionary() in MLIR to use getDiscardableAttrDictionary(...
Mehdi Amini [Mon, 15 May 2023 05:39:50 +0000 (22:39 -0700)]
Cleanup uses of getAttrDictionary() in MLIR to use getDiscardableAttrDictionary() when possible

This also speeds up some benchmarks in compiling simple fortan file by 2x!
Fixes #62687

Differential Revision: https://reviews.llvm.org/D150540

14 months ago[libc++][NFC] Reformat test
Louis Dionne [Mon, 15 May 2023 18:34:54 +0000 (11:34 -0700)]
[libc++][NFC] Reformat test

I didn't notice in the review that clang-format made a poor job at
formatting the test so I went back and did it manually.

14 months ago[libc++][NFC] Use angle brackets to include ranges_mismatch.h
Louis Dionne [Mon, 15 May 2023 18:19:23 +0000 (11:19 -0700)]
[libc++][NFC] Use angle brackets to include ranges_mismatch.h

14 months ago[gn build] Port 61d5671c1697
LLVM GN Syncbot [Mon, 15 May 2023 18:29:44 +0000 (18:29 +0000)]
[gn build] Port 61d5671c1697

14 months ago[gn build] Port 205175578e0d
LLVM GN Syncbot [Mon, 15 May 2023 18:29:43 +0000 (18:29 +0000)]
[gn build] Port 205175578e0d

14 months ago[libc++] Removes _LIBCPP_ABI_OLD_LOGNORMAL_DISTRIBUTION
Mark de Wever [Sun, 7 May 2023 17:50:41 +0000 (19:50 +0200)]
[libc++] Removes _LIBCPP_ABI_OLD_LOGNORMAL_DISTRIBUTION

This was planned for LLVM 15 but was never done.

Reviewed By: #libc, philnik

Differential Revision: https://reviews.llvm.org/D150580

14 months ago[flang][openacc] Lower host_data construct
Valentin Clement [Mon, 15 May 2023 18:22:12 +0000 (11:22 -0700)]
[flang][openacc] Lower host_data construct

Lower host_data construct to the acc.host_data operation.

Depends on D150289

Reviewed By: razvanlupusoru, jeanPerier

Differential Revision: https://reviews.llvm.org/D150290

14 months ago[SLP][NFC]Add missing finalize params in the CostEstimator, NFC.
Alexey Bataev [Fri, 5 May 2023 21:14:39 +0000 (14:14 -0700)]
[SLP][NFC]Add missing finalize params in the CostEstimator, NFC.

Prepare functions for generalization of codegen/cost estimation.

Differential Revision: https://reviews.llvm.org/D150121

14 months ago[libc++] Implement ranges::starts_with
zijunzhao [Mon, 8 May 2023 22:04:00 +0000 (22:04 +0000)]
[libc++] Implement ranges::starts_with

14 months ago[LLD][ELF] change CHECK to CHECK-NEXT in overlay-phdr.test NFCI
Peter Smith [Mon, 15 May 2023 18:01:05 +0000 (19:01 +0100)]
[LLD][ELF] change CHECK to CHECK-NEXT in overlay-phdr.test NFCI

A code-review comment to change a couple of CHECK to CHECK-NEXT that I
forgot to apply prior to committing.

Differential Revision: https://reviews.llvm.org/D150445

14 months agoRevert "[libc++][PSTL] Implement std::copy{,_n}"
Nikolas Klauser [Mon, 15 May 2023 17:33:40 +0000 (10:33 -0700)]
Revert "[libc++][PSTL] Implement std::copy{,_n}"

This reverts commit b049fc0481bc387f57fd61da7239f85ef91096c1.

The wrong patch was landed.

14 months ago[mlir][sparse][gpu] end-to-end integration test of GPU libgen approach
Aart Bik [Mon, 15 May 2023 17:27:39 +0000 (10:27 -0700)]
[mlir][sparse][gpu] end-to-end integration test of GPU libgen approach

Reviewed By: Peiming

Differential Revision: https://reviews.llvm.org/D150172

14 months agoRevert "[AIX][tests] XFAIL -ftime-trace test for now"
Jake Egan [Mon, 15 May 2023 17:48:05 +0000 (13:48 -0400)]
Revert "[AIX][tests] XFAIL -ftime-trace test for now"

The test was fixed by 2f999327534f7cc660d2747ce294f50184dc1f97.

This reverts commit 25dc215ddaa6cb3e206858008fe4bc6844ea0d9c.

14 months ago[flang][runtime] Fixed memory leak in Assign().
Slava Zakharin [Mon, 15 May 2023 16:52:14 +0000 (09:52 -0700)]
[flang][runtime] Fixed memory leak in Assign().

The temporary descriptor must be either Pointer or Allocatable,
otherwise its memory will not be freed.

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D150534

14 months ago[flang][runtime] Fixed dimension offset computation for MayAlias.
Slava Zakharin [Mon, 15 May 2023 16:52:07 +0000 (09:52 -0700)]
[flang][runtime] Fixed dimension offset computation for MayAlias.

Reviewed By: klausler

Differential Revision: https://reviews.llvm.org/D150533

14 months ago[gn build] Port b049fc0481bc
LLVM GN Syncbot [Mon, 15 May 2023 17:38:45 +0000 (17:38 +0000)]
[gn build] Port b049fc0481bc

14 months ago[libc++][docs] Move the pre-release check-list
Louis Dionne [Mon, 15 May 2023 17:34:17 +0000 (10:34 -0700)]
[libc++][docs] Move the pre-release check-list

It was confusing to some contributors because it appeared in a
prominent place on the Contibuting page.

14 months ago[libc++][PSTL] Implement std::copy{,_n}
Nikolas Klauser [Fri, 5 May 2023 16:24:58 +0000 (09:24 -0700)]
[libc++][PSTL] Implement std::copy{,_n}

Reviewed By: ldionne, #libc

Spies: jloser, libcxx-commits

Differential Revision: https://reviews.llvm.org/D149706

14 months agoEnable frame pointer for all non-leaf functions on riscv64 Android
AdityaK [Mon, 15 May 2023 17:17:39 +0000 (10:17 -0700)]
Enable frame pointer for all non-leaf functions on riscv64 Android

Bringing parity with aarch64-android https://github.com/google/android-riscv64/issues/9#issuecomment-1535454205

Reviewers: enh, danalbert, pirama, srhines

Differential Revision: https://reviews.llvm.org/D150490

14 months agoFix build failure caused by https://reviews.llvm.org/D150352
Amy Kwan [Mon, 15 May 2023 16:53:12 +0000 (11:53 -0500)]
Fix build failure caused by https://reviews.llvm.org/D150352

This patch fixes the following build error on the clang-ppc64le-rhel bot seen in
in https://lab.llvm.org/buildbot/#/builders/57/builds/26816/steps/5/logs/stdio:

FAILED: tools/clang/tools/extra/clang-tidy/bugprone/CMakeFiles/obj.clangTidyBugproneModule.dir/UncheckedOptionalAccessCheck.cpp.o
.../clang-ppc64le-rhel/llvm-project/clang-tools-extra/clang-tidy/bugprone/UncheckedOptionalAccessCheck.cpp:43:27: error: 'build' is deprecated: Use the version that takes a const Decl & instead [-Werror,-Wdeprecated-declarations]
      ControlFlowContext::build(&FuncDecl, *FuncDecl.getBody(), ASTCtx);
                          ^
.../ppc64le-clang-rhel-test/clang-ppc64le-rhel/llvm-project/clang/include/clang/Analysis/FlowSensitive/ControlFlowContext.h:41:3: note: 'build' has been explicitly marked deprecated here
  LLVM_DEPRECATED("Use the version that takes a const Decl & instead", "")
  ^
.../clang-ppc64le-rhel/llvm-project/llvm/include/llvm/Support/Compiler.h:143:50: note: expanded from macro 'LLVM_DEPRECATED'
#define LLVM_DEPRECATED(MSG, FIX) __attribute__((deprecated(MSG, FIX)))
                                                 ^
1 error generated.

14 months ago[clang] Convert a few OpenMP tests to use opaque pointers
Sergei Barannikov [Sun, 14 May 2023 17:46:59 +0000 (20:46 +0300)]
[clang] Convert a few OpenMP tests to use opaque pointers

Reviewed By: nikic

Differential Revision: https://reviews.llvm.org/D150530

14 months ago[SLP][NFC] Cleanup: Separate vectorization of Inserts and CmpInsts.
Vasileios Porpodas [Fri, 5 May 2023 20:23:21 +0000 (13:23 -0700)]
[SLP][NFC] Cleanup: Separate vectorization of Inserts and CmpInsts.

This deprecates `vectorizeSimpleInstructions()` and replaces it with separate
functions that vectorize CmpInsts and Inserts.

Differential Revision: https://reviews.llvm.org/D149993

14 months ago[mlir] Fix a warning
Kazu Hirata [Mon, 15 May 2023 17:06:15 +0000 (10:06 -0700)]
[mlir] Fix a warning

This patch fixes:

  mlir/lib/Dialect/MemRef/Utils/MemRefUtils.cpp:45:2: error: extra ';'
  outside of a function is incompatible with C++98
  [-Werror,-Wc++98-compat-extra-semi]

14 months ago[gn build] Port 6851d078c54e
LLVM GN Syncbot [Mon, 15 May 2023 16:58:34 +0000 (16:58 +0000)]
[gn build] Port 6851d078c54e

14 months ago[libc++][PSTL] Implement std::transform
Nikolas Klauser [Mon, 15 May 2023 14:07:45 +0000 (07:07 -0700)]
[libc++][PSTL] Implement std::transform

Reviewed By: ldionne, #libc

Spies: libcxx-commits

Differential Revision: https://reviews.llvm.org/D149615

14 months ago[test] Fix const-str-array-decay.cl failure on PowerPC
Sergei Barannikov [Mon, 15 May 2023 16:54:40 +0000 (19:54 +0300)]
[test] Fix const-str-array-decay.cl failure on PowerPC

D150520 converted the test to use opaque pointers. The update version
fails on PowerPC because of different return type of the function.
This patch resolves the failure by removing the return type check;
it also makes the test look more like it was before the conversion to
prevent other potential issues caused by ABI differences across targets.

14 months ago[clang][AIX] Remove Newly Added Target Dependent Test Case
Qiongsi Wu [Mon, 15 May 2023 16:37:32 +0000 (12:37 -0400)]
[clang][AIX] Remove Newly Added Target Dependent Test Case

https://reviews.llvm.org/D144190 added a test case that is target dependent and requires assembly code generation, which fails on x64 and aarch64 buildbots. This patch removes the test case. We have test cases for code generation added in https://reviews.llvm.org/D144189 already and this removed case was nice to have.

Differential Revision: https://reviews.llvm.org/D150586

14 months ago[flang][hlfir] Fixed copy-in for polymorphic arguments.
Slava Zakharin [Mon, 15 May 2023 16:02:14 +0000 (09:02 -0700)]
[flang][hlfir] Fixed copy-in for polymorphic arguments.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D150502

14 months ago[flang][hlfir] Fixed lowering for intrinsic calls with null() box argument.
Slava Zakharin [Mon, 15 May 2023 16:02:14 +0000 (09:02 -0700)]
[flang][hlfir] Fixed lowering for intrinsic calls with null() box argument.

Reviewed By: jeanPerier

Differential Revision: https://reviews.llvm.org/D150501

14 months agoFix test from b763d6a4ed4650c74c6846d743156468563b0e31
Erich Keane [Mon, 15 May 2023 16:43:25 +0000 (09:43 -0700)]
Fix test from b763d6a4ed4650c74c6846d743156468563b0e31

14 months ago[Mips] Remove MipsRegisterInfo::requiresRegisterScavenging. NFC.
Jay Foad [Mon, 15 May 2023 16:29:00 +0000 (17:29 +0100)]
[Mips] Remove MipsRegisterInfo::requiresRegisterScavenging. NFC.

This method is unused since MipsRegisterInfo is abstract and it is
overridden in both concrete subclasses.

14 months ago[libc][NFC] Clean up the memory buffer handling for RPC
Joseph Huber [Mon, 15 May 2023 14:46:56 +0000 (09:46 -0500)]
[libc][NFC] Clean up the memory buffer handling for RPC

We do a lot of arithmetic on void pointers here, so include a helper and
make some more consistent names. Changes no functionality.

Reviewed By: JonChesterfield

Differential Revision: https://reviews.llvm.org/D150576

14 months ago[AMDGPU] Trim zero components from buffer and image stores
Mateja Marjanovic [Mon, 15 May 2023 16:20:50 +0000 (18:20 +0200)]
[AMDGPU] Trim zero components from buffer and image stores

For image and buffer stores the default behaviour on GFX11 and
older is to set all unset components to zero. So if we pass
only X component it will be the same as X000, or XY same as XY00.

This patch simplifies the passed vector of components in InstCombine
by removing zero components from the end.

For image stores it also trims DMask if necessary.

Reviewed By: foad, arsenm
Differential Revision: https://reviews.llvm.org/D146737

14 months ago[AArch64][CostModel] Add costs for fixed operations when using fixed vectors over...
Dinar Temirbulatov [Mon, 15 May 2023 16:18:45 +0000 (16:18 +0000)]
[AArch64][CostModel] Add costs for fixed operations when using fixed vectors over SVE.

Currently any cast operation with fixed length vectors uses NEON costs,
If those operations are end up using SVE instruction then we estimate
those operations based upon SVE costs.

Differential Revision: https://reviews.llvm.org/D133955

14 months ago[clang][USR] Prevent crashes on incomplete FunctionDecls
Kadir Cetinkaya [Wed, 3 May 2023 08:50:46 +0000 (10:50 +0200)]
[clang][USR] Prevent crashes on incomplete FunctionDecls

FunctionDecls can be created with null types (D124351 added such a new
code path), to be filled in later. But parsing can stop before
completing the Decl (e.g. if code completion
point is reached).
Unfortunately most of the methods in FunctionDecl and its derived
classes assume a complete decl and don't perform null-checks.
Since we're not encountring crashes in the wild along other code paths
today introducing extra checks into quite a lot of places didn't feel
right (due to extra complexity && run time checks).
I believe another alternative would be to change Parser & Sema to never
create decls with invalid types, but I can't really see an easy way of
doing that, as most of the pieces are structured around filling that
information as parsing proceeds.

Differential Revision: https://reviews.llvm.org/D149733

14 months agoAdd C++26 compile flags.
Erich Keane [Fri, 12 May 2023 14:30:21 +0000 (07:30 -0700)]
Add C++26 compile flags.

Now that we've updated to C++23, we need to add C++26/C++2c command line
flags, as discussed in
https://discourse.llvm.org/t/rfc-lets-just-call-it-c-26-and-forget-about-the-c-2c-business-at-least-internally/70383

Differential Revision: https://reviews.llvm.org/D150450

14 months agoRevert "[X86] Use the CFA as the DWARF frame base for better variable locations aroun...
J. Ryan Stinnett [Mon, 15 May 2023 15:52:43 +0000 (16:52 +0100)]
Revert "[X86] Use the CFA as the DWARF frame base for better variable locations around calls."

This reverts commit d421f5226048e4a5d88aab157d0f4d434c43f208.

LLDB tests are failing as shown in
https://green.lab.llvm.org/green/view/LLDB/job/lldb-cmake/55133/testReport/

14 months ago[mlir][sparse][gpu] first implementation of the GPU libgen approach
Aart Bik [Fri, 12 May 2023 19:43:35 +0000 (12:43 -0700)]
[mlir][sparse][gpu] first implementation of the GPU libgen approach

The sparse compiler now has two prototype strategies for GPU acceleration:

* CUDA codegen: this converts sparsified code to CUDA threads
* CUDA libgen: this converts pre-sparsified code to cuSPARSE library calls

This revision introduces the first steps required for the second approach.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D150170

14 months ago[AIX][clang] Storage Locations for Constant Pointers
Qiongsi Wu [Mon, 15 May 2023 15:14:05 +0000 (11:14 -0400)]
[AIX][clang] Storage Locations for Constant Pointers

This patch adds clang options `-mxcoff-roptr` and `-mno-xcoff-roptr` to specify storage locations for constant pointers on AIX.

When the `-mxcoff-roptr` option is in effect, constant pointers, virtual function tables, and virtual type tables are placed in read-only storage. When the `-mno-xcoff-roptr` option is in effect, pointers, virtual function tables, and virtual type tables are placed are placed in read/write storage.

This patch depends on https://reviews.llvm.org/D144189.

Reviewed By: hubert.reinterpretcast, stephenpeckham

Differential Revision: https://reviews.llvm.org/D144190

14 months ago[KnownBitsTest] Remove stray semicolons
Jay Foad [Mon, 15 May 2023 15:20:06 +0000 (16:20 +0100)]
[KnownBitsTest] Remove stray semicolons

14 months ago[mlir][memref] Extract isStaticShapeAndContiguousRowMajor as a util function.
Oleg Shyshkov [Mon, 15 May 2023 15:04:03 +0000 (17:04 +0200)]
[mlir][memref] Extract isStaticShapeAndContiguousRowMajor as a util function.

Differential Revision: https://reviews.llvm.org/D150543

14 months ago[OpenMP] Implement task record and replay mechanism
Chenle Yu [Mon, 15 May 2023 14:56:48 +0000 (09:56 -0500)]
[OpenMP] Implement task record and replay mechanism

This patch implements the "task record and replay" mechanism.  The idea is to be able to store tasks and their dependencies in the runtime so that we do not pay the cost of task creation and dependency resolution for future executions. The objective is to improve fine-grained task performance, both for those from "omp task" and "taskloop".

The entry point of the recording phase is __kmpc_start_record_task, and the end of record is triggered by __kmpc_end_record_task.

Tasks encapsulated between a record start and a record end are saved, meaning that the runtime stores their dependencies and structures, referred to as TDG, in order to replay them in subsequent executions. In these TDG replays, we start the execution by scheduling all root tasks (tasks that do not have input dependencies), and there will be no involvement of a hash table to track the dependencies, yet tasks do not need to be created again.

At the beginning of __kmpc_start_record_task, we must check if a TDG has already been recorded. If yes, the function returns 0 and starts to replay the TDG by calling __kmp_exec_tdg; if not, we start to record, and the function returns 1.

An integer uniquely identifies TDGs. Currently, this identifier needs to be incremented manually in the source code. Still, depending on how this feature would eventually be used in the library, the caller function must do it; also, the caller function needs to implement a mechanism to skip the associated region, according to the return value of __kmpc_start_record_task.

Reviewed By: tianshilei1992

Differential Revision: https://reviews.llvm.org/D146642

14 months ago[KnownBitsTest] Align with ConstantRange test infrastructure (NFC)
Nikita Popov [Mon, 15 May 2023 13:41:16 +0000 (15:41 +0200)]
[KnownBitsTest] Align with ConstantRange test infrastructure (NFC)

Align the way we perform exhaustive tests for KnownBits with what
we do for ConstantRange. Test each case separately by specifying
a function on KnownBits and one on APInts. Additionally, specify
a callback that determines which cases are supposed to be optimal,
rather than only correct. Unlike the ConstantRange case there is
a well-defined, unique notion of optimality for KnownBits.

If a failure occurs, print out the inputs, computed result and
exact result. Adjust the printing function to produce the output
in a format that is meaningful for KnownBits, i.e. print the
actual known bits, using ? to signify unknowns and ! to signify
conflicts.

14 months agoUpdate __cplusplus for C++23, add C++23 diag group alias.
Erich Keane [Mon, 15 May 2023 14:09:07 +0000 (07:09 -0700)]
Update __cplusplus for C++23, add C++23 diag group alias.

This came up during the C++26 flag discussion, so split this out into a
separate patch.

14 months ago[LLVM][Uniformity] Propagate temporal divergence explicitly
Sameer Sahasrabuddhe [Mon, 15 May 2023 11:36:06 +0000 (17:06 +0530)]
[LLVM][Uniformity] Propagate temporal divergence explicitly

At a cycle C with divergent exits, UA was using a naive traversal of the exiting
edges to locate blocks that may use values defined inside C. But this traversal
fails when it encounters a cycle. This is now replaced with a much simpler
propagation that iterates over every instruction in C and checks any uses that
are outside C. But such an iteration can be expensive when C is very large; the
original strategy may need to be reconsidered if there is a regression in
compilation times.

Also fixed lit tests that should have originally caught the missed propagation
of temporal divergence.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D149646

14 months ago[MLIR][ROCDL] add gpu to rocdl erf support
Manupa Karunaratne [Mon, 15 May 2023 14:41:49 +0000 (14:41 +0000)]
[MLIR][ROCDL] add gpu to rocdl erf support

This commit adds lowering of lib func
call to support erf in rocdl.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D150355

14 months ago[mlir] allow repeated payload in structured.fuse_into_containing
Alex Zinenko [Mon, 15 May 2023 12:28:21 +0000 (12:28 +0000)]
[mlir] allow repeated payload in structured.fuse_into_containing

Structured fusion proceeds by iteratively finding the next suitable
producer to be fused into the loop. Therefore, it shouldn't matter if
the same producer is listed multiple times (e.g., it is used as multiple
operands). Adjust the implementation of the transform op to support this
case.

Also fix the checking code in the interpreter to actually respect the
TransformOpInterface indication that repeated payload is allowed, it
seems to have been accidentally dropped in one of the refactorings.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D150561

14 months ago[X86] Use the CFA as the DWARF frame base for better variable locations around calls.
Kyle Huey [Mon, 15 May 2023 14:08:18 +0000 (15:08 +0100)]
[X86] Use the CFA as the DWARF frame base for better variable locations around calls.

Prior to this patch, for the DWARF frame base LLVM uses the frame pointer
register if available, otherwise the stack pointer register. If the stack
pointer register is being used and a call or other code modifies the stack
pointer during the body of the function this results in the locations being
wrong and the debugger displaying the wrong values for variables.

By using DW_OP_call_frame_cfa in these situations the emitted location for
the variable will automatically handle changes in the stack pointer.
The CFA needs to be adjusted for the offset between the frame pointer/stack
pointer to allow the variable locations themselves to remain unchanged by
this patch.

Reviewed By: #debug-info, scott.linder, jryans

Differential Revision: https://reviews.llvm.org/D143463

14 months ago[AArch64] Add test case where widening mull could be used.
Florian Hahn [Mon, 15 May 2023 14:05:10 +0000 (15:05 +0100)]
[AArch64] Add test case where widening mull could be used.

Extra test using mull for D150482.

14 months ago[ConstantFold] use StoreSize for VectorType folding
khei4 [Mon, 15 May 2023 13:33:15 +0000 (22:33 +0900)]
[ConstantFold] use StoreSize for VectorType folding
Differential Revision: https://reviews.llvm.org/D150515
Reviewed By: nikic

14 months agoRevert "[libc++][PSTL] Implement std::transform"
Nikolas Klauser [Mon, 15 May 2023 13:56:40 +0000 (06:56 -0700)]
Revert "[libc++][PSTL] Implement std::transform"

This reverts commit cbd9e5454741ebe6b39521fe1a8ed4eed5c2c801.

The wrong patch was landed.

14 months ago[unittests][llvm-exegesis] Remove build warnings [NFCI]
Francesco Petrogalli [Mon, 15 May 2023 13:26:32 +0000 (15:26 +0200)]
[unittests][llvm-exegesis] Remove build warnings [NFCI]

Remove the warning caused by a missing field initializer.

The field is `StartAtCycle` of `struct MCWriteProcResEntry`.

It has been set to the default `StartAtCycle = 0`.

Differential Revision: https://reviews.llvm.org/D150569

14 months ago[libc++][PSTL] Implement std::transform
Nikolas Klauser [Fri, 5 May 2023 16:16:05 +0000 (09:16 -0700)]
[libc++][PSTL] Implement std::transform

Reviewed By: ldionne, #libc

Spies: libcxx-commits

Differential Revision: https://reviews.llvm.org/D149615