Revert "[X86] X86LowerTileCopy: Find dead register to use to prevent save-reload of tile register (#83628)"

This reverts commit 34acbb38.

This change causes major compile-time regressions.

To accomodate for recent changes in array index calculations.

foldFDivPowDivisor can address A / powi(x, y) to A * powi(x, -y),
while for small const value y, for example y=2, the instcombine will
transform powi(x, 2) to fmul x, x, so it is not optimal for A / powi(x, 2).

Fix https://github.com/llvm/llvm-project/issues/77171

Since PR86428, foldPowiReassoc is called by both FMul and FDiv,
as the optimization of FDiv is placed after the FMul, so now
it is correct we don't add the checking of FDiv for powi(X, Y) / X.
But, we may add more matching scenarios later, so add the checking opcode
explicitly is easier to understand.

The call stack table has a constant key length, so we don't need to
serialize or deserialize it for every key-data pair.  Omitting the key
length saves 0.64% of the indexed MemProf file size.

Note that it's OK to change the format because Version2 is still under
development.

The Python bindings generated for "async" dialect didn't include any of
the "async" dialect ops. This PR fixes issues with generation of Python
bindings for "async" dialect and adds a test case to use them.

This reverts commit f2931182.

This was again causing buildbot failures. #83962 has been updated with the new
failures, notifying the buildbot maintainers that they need to update their
bots.

Original commit message: "

Clang's CodeGen is designed to work with a single llvm::Module. In many cases
for convenience various CodeGen parts have a reference to the llvm::Module
(TheModule or Module) which does not change when a new module is pushed.
However, the execution engine wants to take ownership of the module which does
not map well to CodeGen's design. To work this around we clone the module and
pass it down.

With some effort it is possible to teach CodeGen to ask the CodeGenModule for
its current module and that would have an overall positive impact on CodeGen
improving the encapsulation of various parts but that's not resilient to future
regression.

This patch takes a more conservative approach and keeps the first llvm::Module
empty intentionally and does not pass it to the Jit. That's also not bullet
proof because we have to guarantee that CodeGen does not write on the
blueprint. However, we have inserted some assertions to catch accidental
additions to that canary module.

This change will fixes a long-standing invalid memory access reported by
valgrind when we enable the TBAA optimization passes. It also unblock progress
on https://github.com/llvm/llvm-project/pull/84758.
"

This patch reverts adc4f623 and removes
the check of `named_metadata_empty` of the first llvm::Module because on darwin
clang inserts some harmless metadata which we can ignore.

This reverts commit ca090452 and
1faf3148 because it broke a darwin bot.

This patch fixes:

  clang/lib/Interpreter/IncrementalParser.cpp:214:29: error: moving a
  temporary object prevents copy elision [-Werror,-Wpessimizing-move]

  clang/lib/Interpreter/IncrementalParser.cpp:232:22: error: moving a
  temporary object prevents copy elision [-Werror,-Wpessimizing-move]

The comment being deleted in this patch is not correct.  We already
construct an instance of RecordWriterTrait with Version.

This patch teaches the constructor of RecordWriterTrait to accept
Schema.  While I am at it, this patch makes Version a private
variable.

Clang's CodeGen is designed to work with a single llvm::Module. In many
cases
for convenience various CodeGen parts have a reference to the
llvm::Module
(TheModule or Module) which does not change when a new module is pushed.
However, the execution engine wants to take ownership of the module
which does
not map well to CodeGen's design. To work this around we clone the
module and
pass it down.

With some effort it is possible to teach CodeGen to ask the
CodeGenModule for
its current module and that would have an overall positive impact on
CodeGen
improving the encapsulation of various parts but that's not resilient to
future
regression.

This patch takes a more conservative approach and keeps the first
llvm::Module
empty intentionally and does not pass it to the Jit. That's also not
bullet
proof because we have to guarantee that CodeGen does not write on the
blueprint. However, we have inserted some assertions to catch accidental
additions to that canary module.

This change will fixes a long-standing invalid memory access reported by
valgrind when we enable the TBAA optimization passes. It also unblock
progress
on https://github.com/llvm/llvm-project/pull/84758.

This is closer to that the current interpreter does. It also fixes
diagnostics in a case I was looking into. Unfortunately, this is not
possible to test right now since it requires a large array and we
don't implement array fillers yet.

- `clang/lib/CodeGen/CGBuiltin.cpp` - Add a generic mull add
implementation. Make DXIL implementation tied to target.

resolves #88944

InstCombine may invert branch condition and profile metadata. In such
case, BranchProbabilityAnalysis should also be updated.

These tests just don't check the output written to the current directory. The
current directory may be write protected e.g. in a sandboxed environment.

The Testcases that use -emit-llvm and -verify only care about stdout/stderr
and are in this patch changed to use -emit-llvm-only to avoid writing to an
output file. The verify-inlineasmbr.mir testcase that also only care about
stdout/stderr is in this patch changed to throw away the output file and just
write to /dev/null.

`TemplateArgumentLocInventIterator` default constructor should not
exists

https://github.com/llvm/llvm-project/blob/3496927edcd0685807351ba88a7e2cfb006e1c0d/clang/lib/Sema/TreeTransform.h#L4742
because it doesn't and couldn't initialize `Self` member that is
reference:

https://github.com/llvm/llvm-project/blob/3496927edcd0685807351ba88a7e2cfb006e1c0d/clang/lib/Sema/TreeTransform.h#L4721-L4723

Instantiation of this constructor is always a compile-time error. 

Please note, that I didn't run any tests, because cannot imagine
situation where this constructor can be properly used. There are no new
tests for this fix for the same reason.

The new name uses experimental which better conveys what it means.

[clang] Marshallers.h - use move semantics for 'NodeKinds' and update possible callers to use it  (#87273)

Fixes: https://github.com/llvm/llvm-project/issues/87248



Signed-off-by: amila <amila.15@cse.mrt.ac.lk>

This implements two kinds of optimizations. Specifically
- `char_traits<char8_t>` uses `char` code paths; these are heavily
  optimized and the operations are equivalent
- `char16_t` and `char32_t` `find` uses `std::find` to forward to
  `wmemchr` if they have the same size

Resolve #85250 
Alive2: https://alive2.llvm.org/ce/z/7DMRCy

Also fix a bug in annotating TrailingAnnotation.

Closes #61233.

Matches up with the clang setting

Lambda without trailing auto could have return type declared inside the
body too.

Fixes #68775

Avoid passing them around.

They are not optional.

This reverts commit 2dfa30d0.

This relands commit 0a6c74e2.

This patch originally caused a host of buildbot failures due to several
buildbots not having the necessary python version. That was tracked in
issue #83962, and all bots that failed in the first round have now been
updated. This is an attempt to reland the patch to see if it sticks or
if there are a number of long-running bots where this patch will cause
failures.