[PATCH] D127115: [DAGCombine] Make sure combined nodes are added back to the worklist in topological order.

Thu Jun 1 07:24:22 PDT 2023

foad added a comment.

> FYI I ran this patch on an internal corpus of AMDGPU graphics shaders and didn't see anything alarming. There was a slight change in the way fmul+fadd are combined into fma, but I don't think it was consistently worse - just different. Anyway I will work on that, but I do not think it should block this patch.

I added a test case `@fma_vs_output_modifier` in `test/CodeGen/AMDGPU/dagcombine-fma-fmad.ll` so if you rebase now you will see a new regression. D151890 <https://reviews.llvm.org/D151890> fixes it.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D127115/new/

https://reviews.llvm.org/D127115