[all-commits] [llvm/llvm-project] 18d9ec: [AMDGPU] Mark address space cast from private to f...

[Shilei Tian via All-commits]

  Branch: refs/heads/users/shiltian/as-cast-from-0-to-5-not-uniform
  Home:   https://github.com/llvm/llvm-project
  Commit: 18d9ec996b56a41533b8e428820e7cfdbfaf1bae
      https://github.com/llvm/llvm-project/commit/18d9ec996b56a41533b8e428820e7cfdbfaf1bae
  Author: Shilei Tian <i at tianshilei.me>
  Date:   2025-08-06 (Wed, 06 Aug 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    M llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    A llvm/test/Analysis/UniformityAnalysis/AMDGPU/MIR/addrspacecast.mir
    A llvm/test/Analysis/UniformityAnalysis/AMDGPU/addrspacecast.ll

  Log Message:
  -----------
  [AMDGPU] Mark address space cast from private to flat as divergent if target supports globally addressable scratch

Globally addressable scratch is a new feature introduced in gfx1250. However, this feature changes how scratch space is mapped into the flat aperture, making address space casts from private to flat no longer uniform.



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