Similarly to how block masks are created up front and later only
retrieved also make sure masks are created in cases where edge masks are
needed, i.e. blend recipes.

Creating block-in masks for all blocks in the loop also ensures edge
masks for all relevant edges have been created. Later, the new
getEdgeMask can be used to look up cached edge masks.

This makes sure edge masks are available in all cases for
https://github.com/llvm/llvm-project/pull/76090.

Add lowering for i16 and i32 vectors for Shuffle Vector instructions
with REV mask

Legalize G_ABS for larger/smaller width vectors with legal element sizes

Fallsback for the smaller width vector tests because it is unable to
legalize for G_ANYEXT smaller width vectors

Noticed by inspection - as undef bits are seen as zero, we could be matching in places where we require a zero but end up referencing an undef that could then get changed in a later fold.

Creating a test case has proven difficult as multi-use limits usually saves us, but really we shouldn't be allowing partial undefs for anything but constant folding.

This reverts commit 32878c20 due to #79756 and #76202.

Suppress warnings like

WARNING: Prefix AVX had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX1 had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX2 had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX2-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX512 had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX512F had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX512F-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX512-FAST had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll
WARNING: Prefix AVX512DQ-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-store-i16-stride-7.ll

This fills out the fcmp handling to be more like the other instructions,
adding better support for fp16 and some larger vectors.

Select of f16 values is still not handled optimally in places as the
select is only legal for s32 values, not s16. This would be correct for
integer but not necessarily for fp. It is as if we need to do
legalization -> regbankselect -> extra legaliation -> selection.

It's only used once and only for its name, so replace that with
a string literal.

We are NOT expecting undefs in AND/ANDNP faux shuffle masks

[X86] getTargetConstantBitsFromNode - add explicit support for whole/partial undef elements to all constant folding uses.

A NFC step before I start restricting the use of undef elements in getTargetConstantBitsFromNode

KnownBits::usub_sat is already exhaustively tested in the unit tests

Instead of creating live-ins manually, use getOrAddLiveIn which
automatically takes care of adding them to VPLiveInsToFree. Also use it
to create the VPValue for the trip-count. This fixes a leak:
https://lab.llvm.org/buildbot/#/builders/168/builds/18308/steps/10/logs/stdio

Early out from getTargetConstantBitsFromNode call instead of just returning if an undef was found.

This fixes the crash reported in #76667 and adds an initial unit test
for isSpelledInSource().

Note that in that issue there was still some underlying corrupted AST,
but this at least makes isSpelledInSource() robust to it.

---------

Co-authored-by: Younan Zhang <zyn7109@gmail.com>

Suppress warnings like

WARNING: Prefix AVX had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX1 had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX2 had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX2-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX512 had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX512F had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX512F-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX512-FAST had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll
WARNING: Prefix AVX512DQ-ONLY had conflicting output from different RUN lines for all functions in test vector-interleaved-load-i16-stride-7.ll

These have been deprecated since:

  commit 36117cc4
  Author: serge-sans-paille <sguelton@mozilla.com>
  Date:   Tue Jan 10 11:49:15 2023 +0100

  commit c4b39cd0
  Author: Joe Loser <joeloser@fastmail.com>
  Date:   Mon Jan 16 14:52:16 2023 -0700

Fixes #79685.

This reverts commit 863b2c84.

A compile-time regression has been reported:

https://github.com/llvm/llvm-project/pull/79681#issuecomment-1913325915

Reverts llvm/llvm-project#76316

Buildbot test is broken.

This reverts commit 6953b367.

Couldn't reproduce cmake issues from any of the affected buildbots:
https://lab.llvm.org/buildbot/#/builders/139/builds/57628
https://lab.llvm.org/buildbot/#/builders/163/builds/50270
https://lab.llvm.org/buildbot/#/builders/223/builds/34823
https://lab.llvm.org/buildbot/#/builders/225/builds/29950
https://lab.llvm.org/buildbot/#/builders/245/builds/19468

A common trait is that these builders don't clean build directory.
Push again and monitor these builders.

Remove unneeded empty check and reundant copies. NFC

This is possible now that the free GitHub runners for Windows and Linux
have more disk space:

https://github.blog/2024-01-17-github-hosted-runners-double-the-power-for-open-source/

I also had to switch from macOS-11 to macOS-13 in order to prevent the
job from timing out. macOS-13 runners have 4 vCPUs and the macOS-11
runners only have 3.

fixes #79474.

It seems the automatic Value naming starts on %1 instead of %0 sometimes.