`__unconstrained_reverse_iterator` has outlived its usefullness, since
the standard and subsequently the compilers have been fixed.

Resolves #85065

---------

Co-authored-by: Matt Arsenault <arsenm2@gmail.com>

GCC 13 introduced the type trait `__remove_reference`. We can simplify
the implementation of `remove_reference` a bit by using it.

This patch includes a test that causes an assertion failure without the
other
changes in this patch.

This pull request fixes #77601 where using the `bitand` operator with
boolean operands should not trigger the warning, as it would indicate an
intentional use of bitwise AND rather than a typo or error.

Fixes #77601

Hi @joker-eph, This PR adds XeGPU 2D block operators. It contains:
1. `TensorDescType` and `TensorDescAttr` definitions
2. `MemoryScopeAttr` and `CacheHintAttr` definitions which are used by
`TensorDescAttr`.
3. `CreateNdDescOp`, `PrefetchNdOp`, `LoadNdOp`, and `StoreNdOp`
definitions, and their corresponding testcases for illustration.

---------

Co-authored-by: Mehdi Amini <joker.eph@gmail.com>

e.g. https://lab.llvm.org/buildbot/#/builders/77/builds/35540

On Windows you can not create symlinks without elevated privileges
unless you have Windows developer mode enabled. There's ~67 libcxx tests
that run into failures on windows if your environment is not set up
correctly (Go to windows settings and enable "developer mode").

This change:
- Adds a feature check for whether the host can create symlinks. (see
libcxx/utils/libcxx/test/features.py)
- Mark the feature as required for the 67 tests that hit failures on
windows due to this. This will allow lit to correctly mark these tests
as unsupported instead of unexpectedly failed (this is helpful since
then you know you didn't break something with your change, it's just
that it's not supported with your environment).

Follow on from #84739, which updates the DIBuilder class.

All the functions that have been added are temporary and will be
deprecated in the future. The intention is that they'll help downstream
projects adapt during the transition period.

```
New functions (all to be deprecated)
------------------------------------
LLVMIsNewDbgInfoFormat                      # Returns true if the module is in the new non-instruction mode.
LLVMSetIsNewDbgInfoFormat                   # Convert to the requested debug info format.

LLVMDIBuilderInsertDeclareIntrinsicBefore   # Insert a debug intrinsic (old debug info format). 
LLVMDIBuilderInsertDeclareIntrinsicAtEnd    # Same as above.
LLVMDIBuilderInsertDbgValueIntrinsicBefore  # Same as above.
LLVMDIBuilderInsertDbgValueIntrinsicAtEnd   # Same as above.

LLVMDIBuilderInsertDeclareRecordBefore      # Insert a debug record (new debug info format). 
LLVMDIBuilderInsertDeclareRecordAtEnd       # Same as above.
LLVMDIBuilderInsertDbgValueRecordBefore     # Same as above.
LLVMDIBuilderInsertDbgValueRecordAtEnd      # Same as above.
```

The existing `LLVMDIBuilderInsert...` functions call through to the
intrinsic versions (old debug info format) currently.

In the next patch, I'll swap them to call the debug records versions
(new debug info format). Downstream users of this API can query and
change the current format using the first two functions above, or can
instead opt to temporarily use intrinsics or records explicitly.

The pass uses the TTI hook getIntImmCostIntrin that only supports scalar
integer types. Whilst hoisting expensive vector constant is likely
worthwhile, this is new behaviour and so I've followed the path taken by
the GEP variant of collectConstantCandidates and simply bail for vector
types.

This changs the way the assembly matcher works for Aarch32 parsing.
Previously there was a pile of hacks which dictated whether the CC,
CCOut, and VCC operands should be present which de-facto chose if the
wide/narrow (or thumb1/thumb2/arm) instruction version were chosen.

This meant much of the TableGen machinery present for the assembly
matching was effectively being bypassed and worked around.

This patch makes the CC and CCOut operands optional which allows the ASM
matcher operate as it was designed and means we can avoid doing some of
the hacks done previously. This also adds the option for the target to
allow the prioritizing the smaller instruction encodings as is required
for Aarch32.

This patch fixes problems that pop up when clang emits DbgRecords
instead of debug intrinsics.

Note: this doesn't mean clang is emitting DbgRecords yet, because the
modules it creates are still always in the old debug mode. That will
come in a future patch.

Depends on #84739

This PR:
* adds Lifetime intrinsics/instructions
* fixes how the binary header is emitted (correct version and better
approximation of Bound)
* add validation into more test cases

In commit
https://github.com/llvm/llvm-project/commit/2b582440c16c72b6b021ea5c212ceda3bdfb2b9b,
we canonicalize the isInf/isNanOrInf idiom into fabs+fcmp for better
analysis/codegen (See also the discussion in
https://github.com/llvm/llvm-project/pull/76338).

This patch reverses the fabs+fcmp to `is.fpclass`. If the `is.fpclass`
is not supported by the target, it will be expanded by TLI.

Fixes the regression introduced by
https://github.com/llvm/llvm-project/commit/2b582440c16c72b6b021ea5c212ceda3bdfb2b9b
and
https://github.com/llvm/llvm-project/pull/80414#issuecomment-1936374206.

These includes were removed in 426e6945

This reverts commit 9b98692e.

This reverts commit 6f60ad7e.

Buildbots: https://lab.llvm.org/buildbot/#/builders/196/builds/47206

This reverts commit b097b3dc.

Buildbots: https://lab.llvm.org/buildbot/#/builders/196/builds/47206

This patch fixes problems that pop up when clang emits DbgRecords
instead of debug intrinsics.

Note: this doesn't mean clang is emitting DbgRecords yet, because the
modules it creates are still always in the old debug mode. That will
come in a future patch.

Depends on #84739

One more step in extending support for delayed privatization. This diff
adds support for scalar allocatables and pointers.

For each call that changes the streaming-mode ISel inserts a
COALESCER_BARRIER node for the FP and (non-scalable) vector arguments to
the callee.

When calling a non-streaming function from a streaming-compatible
function, it's not required to have +sme (in case the SME code-path is
not actually executed at runtime). The patterns to match the
COALESCER_BARRIER however were still predicated with `HasSME`, which is
incorrect. This patch tries to fix that.

Follow-up from https://github.com/llvm/llvm-project/pull/85604, this
change also fixes the ArithDialect target.

This option means that in assembly matching instructions with smaller
encodings will be preferred.

This will be used for the ARM instruction set where this is the correct
behavior after some other refactoring.

This fixes tied operand resolution in cases where there are optional operands before the tied operand.

Emits `2.0e+00f` instead of `(float)2.0e+00`.

This helps consumers of the emitted code, especially when there are
large numbers of floating point literals, to have a simple AST.

SinkCandidate is a VPSingleDefRecipe now, so no cast is needed to access
getUnderlyingInstr directly.

Adds a separate target for bufferization interfaces.

`//mlir:ArithDialect` would need to depend on
`//mlir:ValueBoundsOpInterface` as well, but that's creating a circular
dependency.

This change allows us to use `--lto-partitions` in some cases (not at
all guaranteed it works perfectly), as LDS is lowered before the module
is split for parallel codegen.

We must run LowerLDS before splitting modules as it needs to see all
callers of functions with LDS to properly lower them.

This commit changes MLIR's SROA implementation back from being pattern
based into a full pass. This is beneficial for upcoming changes that
rely more heavily on the datalayout.

Unfortunately, this change required substantial test changes, as the
IRBuilder no cleans up the IR.