[PATCH] D153151: [EarlyCSE] Do not CSE convergent readnone calls

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/test/Transforms/EarlyCSE/AMDGPU/convergent-call.ll:22
   ; %x = ballot operation over all lanes.
   %x1 = call i32 @llvm.read_register.i32(metadata !{!"exec_lo"}) convergent
   %cmp = icmp eq i32 %cond, 0
----------------
I think something is wrong here but you should test with something normal and not read_register


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D153151/new/

https://reviews.llvm.org/D153151