[PATCH] D136918: [AMDGPU] Scheduler: fix RP calculation for a MBB with one successor

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D136918#3893989 <https://reviews.llvm.org/D136918#3893989>, @vpykhtin wrote:

> I'm not sure if the undef definition with subreg should be treated as a whole reg def because LIS doesn't think so. However there is a different point of view:
>
>   llvm/lib/CodeGen/RegisterPressure.cpp: line 541
>   
>   void collectOperandLanes(const MachineOperand &MO) const {
>   ...
>         // Treat read-undef subreg defs as definitions of the whole register.
>         if (MO.isUndef())
>           SubRegIdx = 0;

The custom pressure tracking's treatment of sub registers always seemed broken to me. The register tuples do have a stronger constraint than the same number of individual registers. Pressure is always going to be a fuzzier metric than exact liveness


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D136918/new/

https://reviews.llvm.org/D136918