[X86] combineToExtendBoolVectorInReg - don't use changeVectorElementType to create the bool vector type

Converting a (simple) vXf32 type to a vXi1 type isn't guaranteed to be simple, causing the MVT type to be invalid.

Fixes #64627

This reverts commit a5fe6c7f.

This change is causing problems with Windows build bots due to a hanging zombie llvm-symbolizer.exe process.

RuntimeDyld has implemented more relocations than JITLink for ppc64. This patch adds relocations missing from JITLink for ppc64.

Reviewed By: lhames

Differential Revision: https://reviews.llvm.org/D157366

This would mean allowing vector type conversion involving sub-byte
element types.

Reviewed By: kuhar

Differential Revision: https://reviews.llvm.org/D157791

For integer extension or truncation with type emulation, we need
to make sure we perform masking or shifting to discard unwanted
bits to avoid pollute consumer ops.

Reviewed By: kuhar

Differential Revision: https://reviews.llvm.org/D157788

Fix a build failure in GCC 7 caused by the function:
`std::optional<SmallVector<std::unique_ptr<llvm::Module>>> loadBitcodeFiles` ,
in `NVVMTarget` & `ROCDLTarget`.
The failure is caused because GCC fails to use the move constructor in
`std::optional` for constructing the return value, which prompts a call to the
deleted copy constructor in `std::unique_ptr`, resulting in a failure.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D157804

This makes the code easier to search and read.

Reviewed By: kuhar

Differential Revision: https://reviews.llvm.org/D157782

Add support for DONE, RETRY, SAVED, and RESTORED (v9 Section A.11 & Section A.47).

Those instructions are used for low-level interrupt handling and register window
management by OS kernels.

Reviewed By: barannikov88

Differential Revision: https://reviews.llvm.org/D144936

According to the RISC-V ELF psABI, the relocations perform additions,
and this is also what the code implements.

The address-space intrinsics need detecting, and all have very similar
processing.  Refactor to do just that.

Differential Revision: https://reviews.llvm.org/D157763

Reviewed By: jroelofs

We need a pseudo for each scalar FP register class. Previously
we distinquished the pseudos by naming them with F16, F32, F64, or
BF16 in place of the F in the normal instruction name.

Because these strings can appear in other parts of the name we had
to do things like matching "_VBF16" to "_VF".

This patch replaces the F16, F32, F64 strings with FPR16, FPR32, and
FPR64. We also use FPR16 for BF16 since that is the scalar register
class for bf16.

Since the FPR16/32/64 string does not anywhere else in the pseudo
names, we can use this to simplify the string replacements. This
also allows us to simplify some BF16 related code.

Reviewed By: wangpc

Differential Revision: https://reviews.llvm.org/D157749

This avoids needing an FP value to represent LMUL.

Reviewed By: 4vtomat

Differential Revision: https://reviews.llvm.org/D157651

`MachineInstr.h` is a commonly included file and this includes
`llvm/ADT/SmallSet.h` for one function `getUsedDebugRegs()`, which is
used only in one place.

According to `ClangBuildAnalyzer` (run solely on building LLVM, no other
projects) the second most expensive template to instantiate is the
`SmallSet::insert` method used in the `inline` implementation in
`getUsedDebugRegs()`:

```
**** Templates that took longest to instantiate:
554239 ms: std::unordered_map<int, int> (2826 times, avg 196 ms)
521187 ms: llvm::SmallSet<llvm::Register, 4>::insert (930 times, avg 560
       ms)
...
```

By removing this method and putting its implementation in the one call
site we greatly reduce the template instantiation time and reduce the
number of includes.

When copying the implementation, I removed a check on `MO.getReg()` as
this is checked within `MO.isVirtual()`.

Differential Revision: https://reviews.llvm.org/D157720

Differential https://reviews.llvm.org/D145477 removed the check for `(yaml and args.export_fixes)` in line 303 to skip looking for the `clang-apply-replacements` binary. However, the `tmpdir` variable was set in this true branch when exporting fixes and therefore is `None` when invoking run-clang-tidy with `run-clang-tidy -p . -export-fixes fixes.yaml`.

Reviewed By: PiotrZSL

Differential Revision: https://reviews.llvm.org/D157773

- Changed parameters type in `std::rotr` and `std::rorl` functions from `unsigned int` to `int`.
- Implemented behaviour for negative parameter values.

Fixes #64544

Reviewed By: #libc, philnik

Spies: arichardson, philnik, libcxx-commits

Differential Revision: https://reviews.llvm.org/D157569

This work introduces `WGMMATypes` attributes for the `WgmmaMmaSyncOp`. This op, having been recently added to MLIR, previously used `MMATypes`. However, there arises a disparity in supported types between `MmaOp` and `WgmmaMmaSyncOp`. To address this discrepancy more effectively, a new set of attributes is introduced.

Furthermore, this patch refines and optimizing the verification mechanisms of `WgmmaMmaSyncOp` Op.

It also adds support for f8 types, including `e4m3` and `e5m2`, within the `WgmmaMmaSyncOp`.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D157695

If llvm-symbolizer finds a malformed command, it echoes it to the
standard output. New versions of binutils (starting from 2.39) allow to
specify an address by a symbols. Implementation of this feature in
llvm-symbolizer makes the current reaction on invalid input
inappropriate. Almost any invalid command may be treated as a symbol
name, so the right reaction should be "symbol not found" in such case.

The exception are commands that are recognized but have incorrect
syntax, like "FILE:FILE:". The utility must produce descriptive
diagnostic for such input and route it to the stderr.

This change implements the new reaction on invalid input and is a
prerequisite for implementation of symbol lookup in llvm-symbolizer.

Differential Revision: https://reviews.llvm.org/D157210

WgmmaMmaSyncOp is asynchronous operation. There was a typo named op. This work fixes that.

Reviewed By: nicolasvasilache

Differential Revision: https://reviews.llvm.org/D157697

This patch skips images with an empty path or a 0 UUID from loading as a
SymbolFileJSON.

rdar://112107986

Differential Revision: https://reviews.llvm.org/D157137



Signed-off-by: Med Ismail Bennani <ismail@bennani.ma>

Prior to this patch, when a user loaded multiple crash report in lldb,
they could get in a situation where all the targets would keep the same
architecture and executable path as the first one that we've created.

The reason behind this was that even if we created a new CrashLog
object, which is derived from a Symbolicator class that has a newly
constructoted image list as a default argument, because that default
argument is only created once when the function is defined, every CrashLog
object would share the same list.

That will cause use to append newly parsed  images to the same
Symbolicator image list accross multiple CrashLog objects.

To address this, this patch changes the default argument value for the
image parameter to `None` and only initialize it as an empty list when
no argument was passed.

This also removes the image list stored in each CrashLog parsers since
they shouldn't have any state and should be re-usable. So now, the only
source of truth is stored in the CrashLog object.

rdar://84984949

Differential Revision: https://reviews.llvm.org/D157044



Signed-off-by: Med Ismail Bennani <ismail@bennani.ma>

This patch moves CrashLogParseMode into TextCrashLogParser since it's
the only class using it.

Signed-off-by: Med Ismail Bennani <ismail@bennani.ma>

This patch changes the parsing logic for the legacy crash report format
to avoid interrupting the parsing if there are new lines in the middle
of a section.

To do, the parser starts to skip all consecutive empty lines. If the
number of lines skipped is greater than 1, the parser considers that it
reached a new setion of the report and should reset the parsing mode to
back to normal.

Otherwise, it tries to parse the next line in the current parsing mode.
If it succeeds, the parser will also skip that line since it has already
been parsed and continue the parsing.

rdar://107022595

Differential Revision: https://reviews.llvm.org/D157043



Signed-off-by: Med Ismail Bennani <ismail@bennani.ma>

This patch changes the way we dump the registers from the legacy
crashlog command to make sure that the ordering matches the one from lldb.

rdar://109172073

Differential Revision: https://reviews.llvm.org/D156919



Signed-off-by: Med Ismail Bennani <ismail@bennani.ma>

add `owner` to OpResultsList. this is useful for when the list is empty and an element can't be used to fetch the owner.

Differential Revision: https://reviews.llvm.org/D157769

Constructs such as inline variables, #line, and #include can create
lexical blocks with a different filename.
GCOVProfiling and llvm-cov gcov currently don't handle such cases (see
GCOVLines::writeOut and GCOVFile::readGCNO) and would incorrectly
attribute the line number to the current file.

For now, ignore such blocks. Missing line execution counts is better
than wrong ones.

Adds the passes `nvvm-attach-target` & `rocdl-attach-target for attaching `nvvm.target` & `rocdl.target` attributes to GPU Modules.

These passes search GPU Modules in the immediate region of the Op being acted on, attaching the target attribute to the module.
Modules can be selected using a regex string, allowing fine grain attachment of targets, see the test `attach-target.mlir` for an example.

Depends on D154153

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D157351

Adds documentation to the GPU dialect docs giving a general overview of the new
compilation mechanism introduced in the patch series ending in D154153.

Reviewed By: mehdi_amini

Differential Revision: https://reviews.llvm.org/D157461