[all-commits] [llvm/llvm-project] be653f: [X86] combineX86ShuffleChain - don't fold to trunc...

[Jonas Paulsson via All-commits]

  Branch: refs/heads/release/14.x
  Home:   https://github.com/llvm/llvm-project
  Commit: be653f6292e76b82d3532e6fba0e15ce633bd80a
      https://github.com/llvm/llvm-project/commit/be653f6292e76b82d3532e6fba0e15ce633bd80a
  Author: Simon Pilgrim <llvm-dev at redking.me.uk>
  Date:   2022-05-18 (Wed, 18 May 2022)

  Changed paths:
    M llvm/lib/Target/X86/X86ISelLowering.cpp
    M llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll

  Log Message:
  -----------
  [X86] combineX86ShuffleChain - don't fold to truncate(concat(V1,V2)) if it was already a PACK op

Fixes #55050

(cherry picked from commit e8305c0b8f492272446d517e67d451f390d35dbe)


  Commit: c56415735060cc71e58a12e9518569a0b3f7d6b2
      https://github.com/llvm/llvm-project/commit/c56415735060cc71e58a12e9518569a0b3f7d6b2
  Author: Jonas Paulsson <paulsson at linux.vnet.ibm.com>
  Date:   2022-05-18 (Wed, 18 May 2022)

  Changed paths:
    M llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
    M llvm/test/CodeGen/SystemZ/vec-move-23.ll

  Log Message:
  -----------
  [SystemZ] Bugfix in SystemZTargetLowering::combineINT_TO_FP()

Make sure to also handle extended value types to avoid crashing.

Resulting integers greater than 64 bits are not optimized (i128 is not a
legal type), and vectorizing seems to result in libcalls instead of just
scalarization.

Other extended vector types like <10 x float> are however now handled and
should result in vectorized conversions.

Reviewed By: Ulrich Weigand

Differential Revision: https://reviews.llvm.org/D125881

(cherry picked from commit 4273e616e573f4854e9c659c2e6d8a31f3e3c595)


Compare: https://github.com/llvm/llvm-project/compare/52528806579b...c56415735060