[all-commits] [llvm/llvm-project] 8963be: [AMDGPU] Model the cost of the expanded integer to...

[Dmitry Sidorov via All-commits]

  Branch: refs/heads/users/mrsidims/fix-amdgpu-expanded-int-fp-cast-cost
  Home:   https://github.com/llvm/llvm-project
  Commit: 8963be584d8a246c986e5ec800c55c0200d2cbe8
      https://github.com/llvm/llvm-project/commit/8963be584d8a246c986e5ec800c55c0200d2cbe8
  Author: Dmitry Sidorov <Dmitry.Sidorov at amd.com>
  Date:   2026-09-15 (Tue, 15 Sep 2026)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/test/Analysis/CostModel/AMDGPU/cast.ll
    M llvm/test/Analysis/CostModel/AMDGPU/expanded-int-fp-casts.ll
    M llvm/test/Analysis/CostModel/AMDGPU/fptosi.ll
    M llvm/test/Analysis/CostModel/AMDGPU/fptoui.ll

  Log Message:
  -----------
  [AMDGPU] Model the cost of the expanded integer to/from floating point casts

No instruction converts to or from a 64 bit integer, and narrow vector
lanes are converted one at a time. Price these expansions by the FP64
rate, sdwa and 16 bit instruction support, and price i33 to i63 like
i64 and bf16 like f32 plus rounding.

Assisted-by: Claude Code Opus 5



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