This allows users who include `llvm-project` as a subrepository to
access the `LLVM_VERSION_MAJOR` variable on par with if they were
relying on an installed LLVM and `FindLLVM.cmake`. They just need to do
something like:

```
  get_directory_property(LLVM_VERSION_MAJOR DIRECTORY "third_party/llvm-project/llvm" LLVM_VERSION_MAJOR)
```

Context: https://github.com/openxla/iree/pull/16606 -- like other
projects with similar needs that I found by some googling, our
work-around had been to rely on the CMake cached variable
`CLANG_EXECUTABLE_VERSION`. Being cached, it over time inevitably ended
up having a wrong value.

Added support for runtime strides.

Reviewers: preames, RKSimon

Reviewed By: preames

Pull Request: https://github.com/llvm/llvm-project/pull/81517

This is a follow up of #83004, which made the same change for
`MLIR_CUDA_CONVERSIONS_ENABLED`. As the previous PR, this PR commit
exposes mentioned CMake variable through `mlir-config.h` and uses the
macro that is introduced with the same name. This replaces the macro
`MLIR_ROCM_CONVERSIONS_ENABLED`, which the CMake files previously
defined manually.

… AAPCS frame chain fix (#82801)"

This reverts commit 00e4a419. This patch
was found to cause miscompilations and compilation failures.

https://github.com/llvm/llvm-project/blob/762f762504967efbe159db5c737154b989afc9bb/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp#L394-L407

Comment from @topperc:
> This transforms assumes the mask is a non-zero splat. We only know its
a splat and not provably all 0s. The mask is a constexpr that includes
the address of the global variable. We can't resolve the constant
expression to an exact value.

Fixes #83947.

Buildbot failure:
https://lab.llvm.org/buildbot/#/builders/5/builds/41530/steps/9/logs/stdio

The promote alloca to vector transformation assumes that the
vector index is a constant value. If it is not a constant, then
either an assert occurs or the tranformation generates an
incorrect index.

Parenthesized list intializers are sequenced operations, see C++20
[decl.init]p16.5 and [decl.init]p16.6.2.2 for more details.

Fixes #83474

In PR #83677 I was surprised to see that outdated checker callback
signatures are a problem. It turns out, we need the `registerChecker...`
function to invoke the `Mgr.registerChecker<>()` which would instantiate
the `_register` calls, that would take the address of the defined
checker callbacks. Consequently, if the expected signatures mismatch, it
won't compile from now on, so we have static guarantee that this issue
never pops up again.

Given we need the `register` call, at this point we could just hook this
checker into the `debug` package and make it never registered. It
shouldn't hurt anyone :)

[Clang][Sema] Fix crash when using name of UnresolvedUsingValueDecl with template arguments (#83842)

The following snippet causes a crash:
```
template<typename T>
struct A : T {
  using T::f;
  void f();

  void g() {
    f<int>(); // crash here
  }
};
```
This happens because we cast the result of `getAsTemplateNameDecl` as a
`TemplateDecl` in `Sema::ClassifyName`, which we cannot do for an
`UnresolvedUsingValueDecl`. This patch fixes the crash by considering a name
to be that of a template if _any_ function declaration is found per [temp.names] p3.3.

Define BUF Real instructions with this general pattern for all
architectures (not just GFX11):

  multiclass Something_Real_gfx11<...> {
    defvar ps = !cast<Pseudo>(NAME);
    def _gfx11 : ...;
  }

This allows removing a huge amount of repetition in the definitions of
individual Real instructions, where they would have to !cast their own
name to a Pseudo and pass that in as a class argument.

When inlining across functions with different target features, we
perform roughly two checks:
 1. The caller features must be a superset of the callee features.
2. Calls in the callee cannot use types where the target features would
change the call ABI (e.g. by changing whether something is passed in a
zmm or two ymm registers). The latter check is very crude right now.

The latter check currently also catches inline asm "calls". I believe
that inline asm should be excluded from this check, as it is independent
from the usual call ABI, and instead governed by the inline asm
constraint string.

Fixes https://github.com/llvm/llvm-project/issues/67054.

This patch adds support for canonicalization of icmp with a scalable
splat. Some optimizations assume that `icmp pred X, APInt C` is in
canonical form.

Fixes https://github.com/llvm/llvm-project/issues/83931.

Previously SelectionDAGBuilder used ABI alignment for
compressstore/expandload. This patch allows SelectionDAGBuilder to use
parameter alignment like vp intrinsics. This does not follow the
original code to default use vector type alignment, since it is possible
implemented to unaligned vector alignment.

Add functionality to APInt::toString() that allows it to insert
separators between groups of digits, using the C++ literal
separator ' between groups.

Fixes issue #58228

Reviewers:  @AaronBallman, @cjdb, @tbaederr

This allows the removal of FFR related psuedo nodes that only existed to
work round machine verifier failures.

Summary:
The other test locations only give these messages when we are in verbose
logging mode. The average user does not care about which tests are not
being built, and most platforms will have missing tests.

The Base_ prefix seems redundant on a class that is only used for GFX11.

This class only existed to set the dlc bit for GFX11 atomics. It is
simpler to set dlc for all loads/stores/atomics in the base class.

We should generate the `MCInstPredicate` twice, one with `FirstMI`
and another with `SecondMI`.

[InstCombine] Fix Failure to convert vector fp comparisons that can be represented as integers #82241 (#83274)

Resolve #82241

---------

Co-authored-by: SahilPatidar <patidarsahil@2001gmail.com>

According to the PTX ISA this doesn't exist (and ptxas rejects it)

See https://github.com/pytorch/pytorch/issues/118589

I was wondering about this when I recently used `Regex`, and I thought
it would
be nice to have a test documenting this behavior.

As of now, we only check if a class directly inherits from NSObject to
determine if said class has fixed offsets and can therefore "opt-out"
from the non-fragile ABI for ivars.

However, if an NSObject subclass has fixed offsets, then so must the
subclasses of that subclass, so this allows us to optimize instances of
subclasses of subclasses that inherit from NSObject and so on.

To determine this, we need to find that the compiler can see the
implementation of each intermediate class, as that means it is
statically linked.

Fixes: #81369

This allows removing a couple of MTBUF helper (multi)classes.

An EXTRACT_VECTOR_ELT can extend the element to the width of its result
type, leaving the high bits undefined. Previously if we attempted to
query the bytes in these high bits we would recurse and hit an
assertion. This fixes it by bailing if the index is outside of the
vector element size.

I think the assertion Index < ByteWidth may still be incorrect, since
ByteWidth is calculated from Op.getValueSizeInBits(). I believe this
should be Op.getScalarValueSizeInBits() whenever VectorIndex is set
since we're querying the element now, not the vector. But I couldn't
think of a test case to trigger it. It can be addressed in a follow-up
patch.

Fixes #83920

As a temporary solution, lower workshare to the single directive

Post-commit fixup patch for a request on
https://github.com/llvm/llvm-project/pull/81135

move constructors should be marked noexcept

Since svn r133708 visitPHINode no longer has to check basic block
argument types. Better 13 years late than never!

Remove the unused method `CoreEngine::ExecuteWorkListWithInitialState`.

Align attribute has already been used for masked.compress/expandload in
commit #83519, #83763 and #83516.

IMHO it would be productive to make a similar change for `typeid`, in
`ParseCXXTypeid`, in order to improve Clang's error message for
https://godbolt.org/z/oKKWxeYra
But that might be better done by adding a new DeclaratorContext
specifically for TypeidArg, instead of pretending that the argument to
`typeid` is a template argument, because I don't know what else that
change might affect.

Fixes #77585

Due to improper use of RecursiveASTVisitor.

Fixes #80630

This contains two updates:

* Generalize the "no loads stores of aggregates" to "no values of
aggregate type" in general, and be clearer about the exceptions where it
is okay to use them.
 * Mention that you should not load/store non-byte-size types.

Fixes: https://github.com/llvm/llvm-project/issues/64803

Removed unsupported branch command and changed Regex to
accept /cherry-pick: command.