[all-commits] [llvm/llvm-project] b79b68: [NVPTX] Add commutativity to SETP instructions to ...

[modiking via All-commits]

  Branch: refs/heads/users/modiking/nvptx-setp-predicate-inversion
  Home:   https://github.com/llvm/llvm-project
  Commit: b79b68fefb1537d273e5e9779cc0fbb5ce4baf6e
      https://github.com/llvm/llvm-project/commit/b79b68fefb1537d273e5e9779cc0fbb5ce4baf6e
  Author: root <mmo at nvidia.com>
  Date:   2026-04-13 (Mon, 13 Apr 2026)

  Changed paths:
    M llvm/lib/Target/NVPTX/NVPTXInstrInfo.cpp
    M llvm/lib/Target/NVPTX/NVPTXInstrInfo.h
    M llvm/lib/Target/NVPTX/NVPTXInstrInfo.td
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-bfloat16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-float16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-float32.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-float64.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-int16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-int32.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-int64.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-multiple-users.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-inversion-vector-float.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-bfloat16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-float16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-float32.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-float64.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-int16.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-int32.ll
    A llvm/test/CodeGen/NVPTX/machine-cse-predicate-no-inversion-int64.ll

  Log Message:
  -----------
  [NVPTX] Add commutativity to SETP instructions to enable MachineCSE of inverted predicates

Inverted predicates can be used freely in PTX. If we can invert a
predicate and CSE the generating instruction we can save calculating
the inverse.

Teach the NVPTX commuteInstructionImpl that SETP instructions can be
inverted to allow CSEing with previous SETP that match the inverted
form. This also inverts the branch users of the predicate to maintain
correctness.

Currently only allow the SETP inversion if all users are branches.
Future work can extend this to sel and not instructions.

Made-with: Cursor



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications