[PATCH] D157966: [AMDGPU] Autogenerate the fmuladd.f16.ll and llvm.fmuladd.f16.ll codegen tests.

[Ivan Kosarev via Phabricator via llvm-commits]

kosarev added a comment.

> Do we really need both? Are they testing different things?

fmuladd.f16.ll seems to test negated operands, which llvm.fmuladd.f16.ll doesn't test whereas llvm.fmuladd.f16.ll tests `@llvm.fmuladd.v2f16()`, which fmuladd.f16.ll doesn't. So looks like these can be combined, but I'm not sure about the differences in the RUN lines; they use different combinations of `-denormal-fp-math` and `-denormal-fp-math-f32`.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D157966/new/

https://reviews.llvm.org/D157966