[llvm-branch-commits] [clang] [llvm] [AMDGPU][Attributor] Rework update of `AAAMDWavesPerEU` (PR #123995)

Matt Arsenault via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Thu May 1 02:42:09 PDT 2025


================
@@ -1333,6 +1308,56 @@ static void addPreloadKernArgHint(Function &F, TargetMachine &TM) {
   }
 }
 
+/// The final check and update of the attribute 'amdgpu-waves-per-eu' based on
+/// the determined 'amdgpu-flat-work-group-size' attribute. We can't do this
+/// during attributor run because the two attributes grow in opposite direction,
+/// we should not use any intermediate value to calculate waves per eu until we
+/// have a determined flat workgroup size.
+static void updateWavesPerEU(Module &M, TargetMachine &TM) {
----------------
arsenm wrote:

It just occurred to me that AMDGPUAttributor is only a Module pass. Should we add or replace it with a CGSCC pass?

https://github.com/llvm/llvm-project/pull/123995


More information about the llvm-branch-commits mailing list