[all-commits] [llvm/llvm-project] fda8b4: [InterleaveAccess] Copy fast math flags when adjus...

[David Green via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: fda8b4714e05c68deee469970cb6f7f7ea8b32b7
      https://github.com/llvm/llvm-project/commit/fda8b4714e05c68deee469970cb6f7f7ea8b32b7
  Author: David Green <david.green at arm.com>
  Date:   2021-06-17 (Thu, 17 Jun 2021)

  Changed paths:
    M llvm/include/llvm/IR/InstrTypes.h
    M llvm/lib/CodeGen/InterleavedAccessPass.cpp
    M llvm/test/Transforms/InterleavedAccess/AArch64/binopshuffles-inseltpoison.ll
    M llvm/test/Transforms/InterleavedAccess/AArch64/binopshuffles.ll

  Log Message:
  -----------
  [InterleaveAccess] Copy fast math flags when adjusting binary operators in interleave access pass

The Interleave Access pass will convert shuffle(binop(load, load)) to
binop(shuffle(load), shuffle(load)), in order to create more
interleaving load patterns (VLD2/3/4) that might have been messed up by
instcombine. As shown in D104247 we were missing copying IR flags to the
new instruction though, which should just be kept the same as the
original instruction.

Differential Revision: https://reviews.llvm.org/D104255