Preliminary patch for https://github.com/llvm/llvm-project/pull/83285 to
turn polymorphism on by default.

Will give a few days warning for people to remove the flag from their
workflow using flang-new.

Also easier to revert if issues with the gfortran test suites.

Picolibc does not provide the clock_gettime function nor the "rt" library.
check_library_exists was invalidly detecting the "rt" library due to cmake issue 
present, when cross-compiling[1]. This resulted with "chrono.cpp" trying to link
to the "rt" library and following error:
    unable to find library from dependent library specifier: rt

[1] https://gitlab.kitware.com/cmake/cmake/-/issues/18121

- Make `andi` cost 1 in SK_Broadcast
- Query the cost of VID_V, VRSUB_VX/VRSUB_VI which would scale with LMUL

This patch fixes the #67002 ([OpenMP][Clang] Scan Directive not
supported for Generic types). It disables the Sema checks/analysis that
are run on the helper arrays which go into the implementation of the
`omp scan` directive until the template instantiation happens.
Grateful to @alexey-bataev for suggesting these changes.

Modifying clang/lib/CodeGen/CGStmtOpenMP.cpp to accept multiple `init`
clauses with `interop` directive.

---------

Co-authored-by: Sunil Kuravinakop

Document that we don't use the double compare and swap instructions due
to ABI concerns.

Also combine (X != 0) | (Y != 0) -> (X | Y) != 0

This reverts commit 79330098 as pexpect
is not available on any of the GreenDragon bots.

This executed Alex' idea [1] of adding pexpect to the list of "strict
test requirements" as we're planning to stop vendoring it. This will
ensure all the bots have the package before we toggle the default.

[1] https://github.com/llvm/llvm-project/pull/83191

This patch adds smin/smax/umin/umax/sadd_sat/ssub_sat/uadd_sat/usub_sat
into canSplatOperand. It can help llvm fold vv instructions with one
splat operand to vx instructions.

Add an external interface to register a branch in a function that is in
disassembled state. Allows to make custom modifications to the
disassembler. E.g., a pre-CFG pass can add an instruction and register a
branch that will later be used during the CFG construction.

Move code that sorts TakenBranches right before the branches are used.
We can populate TakenBranches in pre-CFG post-processing and hence have
to postpone the sorting to a later point in the processing pipeline.
Will add such a pass later. For now it's NFC.

Extended VMX instructions and 8 bit apx extended instructions don't need
W bit, they are marked as W ignored in spec.
RFC:
https://discourse.llvm.org/t/rfc-design-for-apx-feature-egpr-and-ndd-support/73031/4

I'm not sure the f64 fma cases are correct.

In addition to being rather hard to follow, there isn't a good reason
why FatLTO shouldn't just share the same code for setting module flags
for (Thin)LTO. This patch simplifies the logic and makes sure we use set
these flags in a consistent way, independent of FatLTO.

Additionally, we now test that output in the .llvm.lto section actually
matches the output from Full and Thin LTO compilation.

Originally, when `EnableImportMetadata` enabled, `SourceFileName` will
be recorded as `thinlto_src_module`. Now `SourceFileName` will be
recorded as `thinlto_src_file` and `ModuleIdentifier` will be recorded
as `thinlto_src_module`.

There was a think-o in a previous commit that made us only able to
define 1 line commands when using command script add interactively.

There was also no test for this feature, so I fixed the think-o and
added a test.

WaveSizePredicate for DS_Reaf and FLAT_Real
  OtherPredicates for MTBUF_Real

My global find+replace was overzealous and broke post submit unit tests.

Link: #83345

When removing only lines that are global value CHECK lines, a related
CHECK-SAME line could be left dangling without a previous line to belong
to.

Resolves #78517

`isContiguousAccess` is an important affine analysis utility but is only
tested very indirectly via passes like vectorization and is not exposed.
Expose it and add a test pass for it that'll make it easier/feasible to
write test cases.  This is especially needed since the utility can be
significantly enhanced in power, and we need a test pass to exercise it
directly.

This pass can in the future be used to test the utility of invariant
accesses as well.

…fer] op).

This specifically addresses the warnings:
$LLVM/lldb/include/lldb/API/SBCommandReturnObject.h:119: Warning 509:
Overloaded method lldb::SBCommandReturnObject::PutCString(char const *)
effectively ignored,
$LLVM/lldb/include/lldb/API/SBCommandReturnObject.h:119: Warning 509: as
it is shadowed by lldb::SBCommandReturnObject::PutCString(char const
*,int).

There is exactly one declaration of SBCommandReturnObject::PutCString.
The second parameter (of type `int`) has default value `-1`. Without
investigating why SWIG believes there are 2 method declarations, I
believe it is safe to ignore this warning. It does not appear to
actually impact functionality in any way.

rdar://117744660

This is first step (of many) cleaning up our tests to use the new and
exciting sparse_tensor.print operation instead of lengthy extraction +
print ops.

Change `image list -r` so that it actually shows the ref count and whether the image is in the shared cache. (#83341)

The help for the `-r` option to `image list` says:

       -r[<width>] ( --ref-count=[<width>] )
Display the reference count if the module is still in the shared module
cache.

but that's not what it actually does. It unconditionally shows the
use_count for all Module shared pointers, regardless of whether they are
still in the shared module cache or whether they are just in the
ModuleCollection and other entities are keeping them alive. That seems
like a more useful behavior, but then it is also useful to know what's
in the shared cache, so I changed this to:

       -r[<width>] ( --ref-count=[<width>] )
Display whether the module is still in the the shared module cache
(Y/N), and its shared pointer use_count.

So instead of just `{5}` you will see `{Y 5}` if it is in the shared
cache and `{N 5}` if not.

I didn't add tests for this because I'm not sure how much we want to fix
shared cache behavior in the testsuite.

The sparse structure buffers might not always be memrefs with rank == 1
with the presence of batch levels.

This reverts commit 2eb63982.

This caused verifier error
```
Instruction does not dominate all uses!
```
for some projects using Halide.
The verifier error happens inside `Halide::Internal::CodeGen_LLVM::optimize_module`
and looks like a genuine SROA issue.

The -d(ebug) option broke 5 years ago when I migrated the driver to
libOption. Since then, we were never check if the option is set. We were
incorrectly toggling the internal variable (m_debug_mode) based on
OPT_no_use_colors instead.

Given that the functionality doesn't seem particularly useful and nobody
noticed it has been broken for 5 years, I'm just removing the flag.

`Driver::ClangExecutable` is derived from:

* (-canonical-prefixes default): `realpath` on the executable path
* (-no-canonical-prefixes) argv[0] (consult PATH if argv[0] is a word)

`Dir` and `ResourceDir` are derived from `ClangExecutable`. Both
variables are used to derive certain include and library paths.

`InstalledDir` is a related concept used to derive certain other paths.
`InstalledDir` is weird in the -canonical-prefixes mode: Clang
calls `make_absolute` but does not follow symlinks
(FIXME from 9ade6a9a).
This causes some search and library paths to be mix-and-matched.

The "Do a PATH lookup, if there are no directory components." logic
makes things worse.
`InstalledDir` is different when you invoke it via `PATH`:
```
% which clang
/usr/bin/clang
% clang -v |& grep InstalledDir
InstalledDir: /usr/bin
% /usr/lib/llvm-16/bin/clang -v |& grep InstalledDir
InstalledDir: /usr/lib/llvm-16/bin
```

I believe `InstalledDir` was a partial solution to
`-no-canonical-prefixes` and should be eventually removed.

This patch removes `SetInstallDir` and relies on Driver::Driver to set
`InstalledDir` to `Dir`. The behavior for regular file `clang` or
`-no-canonical-prefixes` is unchanged.

If a user creates a symlink to the regular file `clang` and uses the
default `-canonical-prefixes`, they now consistently get search and
library paths relative to the regular file `clang`, not mix-and-match
paths.

If a user creates a symlink to the regular file `clang` and replaces
some directorys from the actual installation, they should change the
symlink to a wrapper that calls the underlying clang with
`-no-canonical-prefixes`.

I've seen cases where the cost per use increase the number of spills.
Disabling improves the codegen for #79918.

I propose adding this option to allow easier experimentation.

[SelectionDAG] Remove unused getIndexedStridedLoadVP/getIndexedStridedStoreVP functions. NFC (#82847)

These appear to have been copied from getIndexedLoadVP/getIndexedStoreVP
which in turn were copied from the non-VP versions.

Codify that we use lower_case for
readability-identifier-naming.ConstexprFunctionCase and then fix the 11
violations (rather than codify UPPER_CASE and have to fix the 170 violations).