[all-commits] [llvm/llvm-project] 82113a: [LLVM][NVPTX] Remove nonexistent ftz ops (#106100)

[Billy Zhu via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 82113a432c5bffe026682ea117a3e2cd67a2fed0
      https://github.com/llvm/llvm-project/commit/82113a432c5bffe026682ea117a3e2cd67a2fed0
  Author: Billy Zhu <billyzhu at modular.com>
  Date:   2024-08-28 (Wed, 28 Aug 2024)

  Changed paths:
    M llvm/lib/Target/NVPTX/NVPTXInstrInfo.td
    M llvm/test/CodeGen/NVPTX/bf16-instructions.ll

  Log Message:
  -----------
  [LLVM][NVPTX] Remove nonexistent ftz ops (#106100)

According to the PTX
[spec](https://docs.nvidia.com/cuda/parallel-thread-execution/#half-precision-floating-point-instructions-max),
max & min instructions do not support the `ftz` modifier for `bf16` &
`bf16x2` types. This PR removes them from instr info, and the non-ftz
legal versions will be emitted instead.



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