[llvm] [AMDGPU] Use correct VGPR threshold for flagging ExcessRP regions in unified register file case (PR #85860)
Stanislav Mekhanoshin via llvm-commits
llvm-commits at lists.llvm.org
Thu Mar 21 14:30:35 PDT 2024
================
@@ -974,11 +974,13 @@ void GCNSchedStage::checkScheduling() {
<< DAG.MinOccupancy << ".\n");
}
- unsigned MaxVGPRs = ST.getMaxNumVGPRs(MF);
+ unsigned MaxVGPRs = ST.getMaxNumVGPRs(MF, /*WholeRegisterFile*/ true);
+ unsigned MaxArchVGPRs = ST.getMaxNumVGPRs(MF, /*WholeRegisterFile*/ false);
----------------
rampitec wrote:
Do I get this right, after the last change `MaxArchVGPRs` is initialized with the equivalent of `min(MaxVGPRs, getAddressableNumArchVGPRs)`? If so it is easier to do just this.
https://github.com/llvm/llvm-project/pull/85860
More information about the llvm-commits
mailing list