Also improve the implementation of `findCommonDominator` (skip duplicate
blocks) and extract it from `BufferPlacementTransformationBase` (so that
`BufferPlacementTransformationBase` can be retired eventually).

`BufferPlacementTransformationBase::isLoop` checks if there a loop in
the region branching graph of an operation. This algorithm is similar to
`isRegionReachable` in the `RegionBranchOpInterface`. To avoid duplicate
code, `isRegionReachable` is generalized, so that it can be used to
detect region loops. A helper function
`RegionBranchOpInterface::hasLoop` is added.

This change also turns a recursive implementation into an iterative one,
which is the preferred implementation strategy in LLVM.

Also move the `isLoop` to `BufferOptimizations.cpp`, so that we can
gradually retire `BufferPlacementTransformationBase`. (This is so that
proper error handling can be added to `BufferViewFlowAnalysis`.)

This patch relaxes the one-use constraints for `icmp pred (binop X, Z),
(binop Y, Z)`. It will enable more optimizations with pointer
arithmetic.
One example in `boost::match_results::set_size`:

```
declare void @use(i64)
define i1 @src(ptr %a1, ptr %a2, ptr %add.ptr.i66, i64 %sub.ptr.rhs.cast.i) {
  %sub.ptr.lhs.cast.i = ptrtoint ptr %a1 to i64
  %sub.ptr.rhs.cast.i = ptrtoint ptr %a2 to i64
  %sub.ptr.sub.i = sub i64 %sub.ptr.lhs.cast.i, %sub.ptr.rhs.cast.i
  %sub.ptr.div.i = sdiv exact i64 %sub.ptr.sub.i, 24
  call void @use(i64 %sub.ptr.div.i)
  %sub.ptr.lhs.cast.i.i = ptrtoint ptr %add.ptr.i66 to i64
  %sub.ptr.sub.i.i = sub i64 %sub.ptr.lhs.cast.i.i, %sub.ptr.rhs.cast.i
  %sub.ptr.div.i.i = sdiv exact i64 %sub.ptr.sub.i.i, 24
  %cmp.i.not.i.i = icmp eq i64 %sub.ptr.div.i.i, %sub.ptr.div.i
  ret i1 %cmp.i.not.i.i
}
define i1 @tgt(ptr %a1, ptr %a2, ptr %add.ptr.i66, i64 %sub.ptr.rhs.cast.i) {
  %sub.ptr.lhs.cast.i = ptrtoint ptr %a1 to i64
  %sub.ptr.rhs.cast.i = ptrtoint ptr %a2 to i64
  %sub.ptr.sub.i = sub i64 %sub.ptr.lhs.cast.i, %sub.ptr.rhs.cast.i
  %sub.ptr.div.i = sdiv exact i64 %sub.ptr.sub.i, 24
  call void @use(i64 %sub.ptr.div.i)
  %cmp.i.not.i.i = icmp eq i64 %sub.ptr.sub.i.i, %sub.ptr.sub.i
  ret i1 %cmp.i.not.i.i
}
```

This reverts commit 0903d992.

This is causing all non-Visual Studio builds fail.

Visual Studio needs the class template stuff.

C:\llvm\include\llvm/Analysis/MemoryBuiltins.h(217): error C2990: 'llvm::SizeOffsetType': non-class template has already been declared as a class template
C:\llvm\include\llvm/Analysis/MemoryBuiltins.h(193): note: see declaration of 'llvm::SizeOffsetType'

The ParamPoint datatype has each column representing an affine function.
The code for generating functions is modified to reflect this.

We add some basic type aliases and function definitions relating to
cones for Barvinok's algorithm.
These include functions to get the dual of a cone and find its index.

This should have been checking that the transform was valid, but used
incorrect conditions letting through invalid combinations of lo/hi
extracts.

Hopefully fixes #76769

see #73359

Declarative assemblyFormat ODS is more concise and requires less
boilerplate than filling out CPP interfaces.

Changes:
* updates the Ops defined in `SPIRVAtomicOps.td` to use assemblyFormat.
* Removes print/parse from`AtomcOps.cpp` which is now generated by
assemblyFormat
* Adds `Trait` to verify that a pointer operand `foo`'s pointee type
matches operand `bar`'s type
* * Updates error message expected in tests from new Trait
* Updates tests to updated format (largely using <operand> in place of
"operand")

This reverts commit b5de136e.

Based on post commit feedback, I need to some other work before
this makes sense.

Closes #76106

---------

Co-authored-by: Lei Zhang <antiagainst@gmail.com>

-Rename sub_32_hi to sub_gpr_odd
-Add dedicated sub_gpr_even.
-Rename sub_32 and sub_16 to sub_fpr32 and sub_fpr16.
-Remove start offset from sub_gpr_odd. AArch64 doesn't use non-zero offset for GPR
tuples so I don't think we need to.

This is preparation for a RV64 GPRPair for Zacas.

This reverts commit 16875555.

BTW, we relax the condition for EVEX compression from
ST.hasAVX512() to ST.hasEGPR() || ST.hasAVX512(). It does not have any
effect now b/c no APX instruction is in the EVEX compression table so
far.

This patch is to extract NFC in #77065 into a separate commit.

This fixes the bug introduced by

https://github.com/llvm/llvm-project/commit/6db007a0654ed7a6ed5c3aa3b61a937c19a6bc6b.

We construct placeholder template arguments for template-template
parameters to avoid mismatching argument substitution since they have
different depths with their corresponding template arguments. In this
case,

```cpp
template <template <Concept C> class T> void foo(T<int>);
```

T lies at the depth 0, and C lies at 1. The corresponding argument, of
which there is exactly one, int, is at depth 0. If we consider the
argument as the outermost one, then we would end up substituting 'int'
into the wrong parameter T.

We used to perform such placeholder construction during the context
walk-up. In the previous patch, we slipped through that inadvertently
because we would walk up to the parent, which is precisely a FileContext
for template-template parameters, after adding innermost arguments.

Besides, this patch moves the sanity check up to the context switch.
That way, we avoid dereferencing null pointers if ND is unspecified.

Closes https://github.com/llvm/llvm-project/issues/57410.
Closes https://github.com/llvm/llvm-project/issues/76604. (The case is
slightly different than that in #57410. We should *not* assume the
surrounding context to be a file-scope one.)

Those "_REV" instructions should not appear before encoding
optimization, while macro fusion and flag-copy lowering are before
encoding optimization.

1. Simplify getValueFromBitsInit about cast and return type
2. Remove out-of-date comments and allow memory ops in function
   object `IsMatch` so that we can reuse it for EVEX2Legacy compression.

This patch is to extract NFC in #77065 into a separate commit.

With #70253 landed, selects for reduction results are explicitly used by
ComputeReductionResult and Selects can be marked as not having
side-effects again.

This reverts the revert commit 17303290.

Parameter variable which is forwarded in lambda capture list or in body
by reference is reasonable and current version of this check produces
false positive on these cases. This patch try to fix the
[issue](https://github.com/llvm/llvm-project/issues/68105

)

Co-authored-by: huqizhi <836744285@qq.com>

We shift the GeneratingFunction.h header file to the include/ directory
and wrap it in a `detail` namespace.

This patch is to address my review comments in #77065 to simplify the
implemention of EVEX2Legacy compression.

Remove these two classes and put all the entries in X86 EVEX compression tables
that need special handling in .def file.

PR #77065 tries to add entries that need special handling for APX in
.def file. Compared to setting fields in td files, that method looks
cleaner. This patch is to unify the addition of manual entries.

This is important to aid development of Python type annotations in the
libclang binding.
See https://github.com/llvm/llvm-project/issues/76664 for more details.

* Run on all pull requests and direct pushes.
* This makes use of the existing llvm-project-tests.yml recipe, which
will preload ccache from previous runs.
* Building libclang currently takes about 9mins when ccache is warm and
about an 1hr 20mins if it is cold using the standard GitHub ubuntu
runner.
* In the future, this could be broken into the following discrete steps
for clarity:
   1. Build libclang dependency.
       ninja -C build libclang
   2. Run Python unit tests.
       ninja -C build check-clang-python
* Followup changes will bring testing on older python versions and
static type checking.

Issue https://github.com/llvm/llvm-project/issues/76601.

Remove the redundant sccache cache key prefix.
This prefix is already added by the ccache action, which results in
cache keys like "sccache-sccache-ubuntu-...".

See the following source lines as proof:

https://github.com/hendrikmuhs/ccache-action/blob/2a51777f6f64b7b7bea213601acba8f5f4fdbe03/src/restore.ts#L22-L23

The "Setup ccache" step on macOS-11 builds takes between 15 to 20 mins,
whereas this step takes a less than 10 seconds on other runners. The
bulk of this time is spent at the "Install sccache" step, where brew
emits warnings like "Warning: You are using macOS 11." and "We (and
Apple) do not provide support for this old version...".

Bumping the version of macOS greatly decreases this cache setup time to
about 20 seconds. Furthermore, it seems like it is speeding up general
build times, too.

It appears that
https://github.com/actions/virtual-environments/issues/5900 has been
resolved or obsoleted, so I do not believe
we need to lock macOS to 11 anymore.

This patch gets the code model from global variable attribute if it has,
otherwise the target's will be used.

---------

Signed-off-by: WANG Rui <wangrui@loongson.cn>