Compiling with -DCMAKE_CXX_STANDARD=20 produces 228 warnings from this
file due to:
```
LVOptions.h:515:16: warning: implicit capture of 'this' with a capture default of '=' is deprecated [-Wdeprecated-this-capture]
```

So I've changed these to explicitly list the captures, including `this`.

As llvm requires at least c++17, I think we could use `[=, *this]`
instead. However when I did so I got a lot of errors about const. So on
balance, explicitly listing the captures seems better than adding some
kind of const cast to each one.

These and other warnings can be seen on the c++20 buildbot
https://lab.llvm.org/buildbot/#/builders/249.

Currently if the merged string is used by metadata, its metadata uses
are not replaced if the string is merged. This is to add code support
for the metadata use replacement.

The intrinsic is defined as a GNU extension here:
https://gcc.gnu.org/onlinedocs/gfortran/SIGNAL.html

And as an IBM extension here:
https://www.ibm.com/docs/en/xffbg/121.141?topic=procedures-signali-proc-extension

The IBM version provides a compatible subset of the functionality
offered by the GNU version. This patch supports most of the GNU
features, but not calling SIGNAL as a function. We don't currently
support intrinsics being both subroutines AND functions and this changed
seemed too large to be justified by a non-standard intrinsic.

I cannot point to open source code Fortran using this intrinsic. This is
needed for a proprietary code base.

This patch provides specific test macros to deal with `errno`.
This will help abstract away the differences between unit test and integration/hermetic tests in #79319.
In one case we use `libc_errno` which is a struct, in the other case we deal directly with `errno`.

The LegalizerHelper only has support to lower G_SELECT with
vector operands. The approach is the same for scalar arguments,
which this PR adds.

clang's -Wdtor name is correct, but the standard may have not
intended that meaning, according to https://bugs.llvm.org/show_bug.cgi?id=46979#c1.
Some of the wording may have changed in 20/23, but we of course
need to support c++17 as well as that's our default.

One workaround would be to explicitly open the namespaces,
then declare the destructor inside that.

Another as shown in the bug report is to repeat the class name, without
the template arguments, before the ::~. For example `Bar::Foo<T>::Foo::~Foo`.
(this extra Foo is the injected class name
https://en.cppreference.com/w/cpp/language/injected-class-name)

I chose to do this because it's the smallest change. It works
with gcc-13 and clang in c++17 and 20 modes (https://godbolt.org/z/fqs4fGE7T).

GEPArg can only be constructed from int32_t and mlir::Value. Explicitly
cast other types (e.g. unsigned, size_t) to int32_t to avoid narrowing
conversion warnings on MSVC. Some recent examples of such are:

```
mlir\lib\Dialect\LLVMIR\Transforms\TypeConsistency.cpp: error C2398:
Element '1': conversion from 'size_t' to 'T' requires a narrowing
conversion
    with
    [
        T=mlir::LLVM::GEPArg
    ]

mlir\lib\Dialect\LLVMIR\Transforms\TypeConsistency.cpp: error C2398:
Element '1': conversion from 'unsigned int' to 'T' requires a narrowing
conversion
    with
    [
        T=mlir::LLVM::GEPArg
    ]
```

Co-authored-by: Nikita Kudriavtsev <nikita.kudriavtsev@intel.com>

This fixes a miscompile from #79072 where we were taking the wrong SrcVec to do
the M1 shuffle. E.g. if the SrcVecIdx was 2 and we had 2 VRegsPerSrc, we ended
up taking it from V1 instead of V2.

After 0ab539fd, the canonical IV in the
epilogue vector loop may be used by a trunc. Relax the corresponding
assert.

This should fix some build-bot failures, including
    https://lab.llvm.org/buildbot/#/builders/187/builds/14113
    https://lab.llvm.org/buildbot/#/builders/98/builds/32350
    https://lab.llvm.org/buildbot/#/builders/239/builds/5473

This PR adds a check for IsTerminator trait to prevent deletion of ops
like gpu.terminator as a "simple op" by RemoveDeadValues pass.

Fix assert in `MemProfContextDisambiguation::applyImport` when exists
direct recursion.

[NFC] Rename TargetInstrInfo::FoldImmediate to TargetInstrInfo::foldImmediate and simplify implementation for X86

This reverts commit 5e9f0e37
because it creates a new warning from clang:
```
NSDictionary.cpp:1063:14: warning: ISO C++ requires the name after '::~' to be found in the same scope as the name before '::~' [-Wdtor-name]
    D32, D64>::~GenericNSDictionaryMSyntheticFrontEnd() {
    ~~~~~~~~~^~
             ::GenericNSDictionaryMSyntheticFrontEnd
```

If you remove the template arguments from before the `::`, you
then get:
```
NSDictionary.cpp:1062:27: error: use of class template 'lldb_private::formatters::GenericNSDictionaryMSyntheticFrontEnd' requires template arguments
lldb_private::formatters::GenericNSDictionaryMSyntheticFrontEnd::~GenericNSDictionaryMSyntheticFrontEnd() {
                          ^
```

And I'm not aware of a way to fix that.

When compiling with gcc 11+ and -DCMAKE_CXX_STANDARD=20, errors
like the following happened:
```
llvm-project/lldb/source/Plugins/Language/ObjC/NSDictionary.cpp:1063:5: error: template-id not allowed for destructor
 1063 |     ~GenericNSDictionaryMSyntheticFrontEnd<D32,D64>() {
      |     ^
```
This appears to be something only gcc enforces and only from 11 and beyond.

This changes fixes all the instances of this pattern by removing the
template arguments.

It is split off from #77342.
InstCombine transform min/max reduction with i1 into arithmetic
reduction,
so this patch reuses the cost logic in arithmetic reduction cost
function.

Currently when interleaving vector calls with linear arguments,
the Part is ignored and all vector calls use the initial value
from the first lane of the current iteration.

Fix this to extract from the correct part of the linear vector.

Add a new recipe to model scalar cast instructions, without relying on
an underlying instruction.

This allows creating scalar casts, without relying on an underlying
instruction (like the current VPReplicateRecipe). The new recipe is 
used to explicitly model both truncating the induction step and the
VPDerivedIVRecipe, thus simplifying both the recipe and code
needed to introduce it.

Truncating VPWidenIntOrFpInductionRecipes should also be modeled using
the new recipe, as follow-up.

PR: https://github.com/llvm/llvm-project/pull/78113

This intrinsic is a gnu extension. See
https://gcc.gnu.org/onlinedocs/gfortran/SLEEP.html

This intrinsic is used in minighost:
https://github.com/Mantevo/miniGhost/blob/c2102b521568a74862fa5abb074b1fc8041fc222/ref/MG_UTILS.F#L606

The utility functions this patch modifies are part of cleanly
transitioning from a context where we use dbg.value intrinsics to one
where we use DPValue objects to record debug-info, and back again.
However, this is a waste of time in non-debug builds (i.e. no -g on the
command line). We still have to call the function on all blocks though
to set the IsNewDbgInfoFormat flag.

To reduce the overhead of this, test whether there's any debug-info in
the function by checking whether the function has a DISubprogram, and
pass a flag down to the utility functions indicating whether they can
skip the scan.

It feels a bit dumb to me now that we're scanning and setting a flag in
a load of blocks when we don't have to -- however it's been really
valuable during development for working out where spurious dbg.value
intrinsics leak into a RemoveDIs context. Happily we'll be able to just
delete this flag entirely when RemoveDIs lands and sticks, and the
conversion routines will eventually be pushed down into the debug-info
autoupgrade path.

This change makes the callback consistent with
`notifyOperationInserted`: both now notify about IR insertion, not IR
creation. See also #78988.

This change also simplifies the dialect conversion: it is no longer
necessary to override the `inlineRegionBefore` method. All information
that is necessary for rollback is provided with the
`notifyBlockInserted` callback.

Previously, for masked tile loads/stores we directly used the dimension
size from the `vector.create_mask` operation as the upper bound of the
`scf.for` over the tile slices. This was not correct, as `create_mask`
allows operands to be greater than the size of the vector dimension, in
which case the for loop bounds should be clamped to the number of tile
slices.

Adds tests showing that we select function version according to the
highest feature priority. This will make the changes introduced by
#79316 more evident.

As of 5aea839a, this FIXME has been
resolved.

So we can remove this comment.

This patch avoids adding redundant vcreate_v intrinsics to the RISCV
IntrinsicList.
Since vcreate_v uses LFixedLog2LMUL, I believe we can simply set
Log2LMUL to the smallest value (-3) to prevent the creation of redundant
vcreate_v instances with the same intrinsic name and prototype in the
IntrinsicList when clang creates it.

This reland commit 0a3b5ece with fix.

This reverts commit 0a3b5ece.