[PATCH] D156471: [mlir] Support fast-math friendly constants for identity value

Fri Jul 28 16:42:22 PDT 2023

dcaballe accepted this revision.
dcaballe added a comment.
This revision is now accepted and ready to land.

Thanks, Quentin! LGTM. I mostly agree with Renato's feedback. We also have the fact that fmin/fmax semantics <https://mlir.llvm.org/docs/Dialects/ArithOps/#arithminf-arithminfop> in MLIR match LLVM's fmaximum/fminimum <https://llvm.org/docs/LangRef.html#llvm-vector-reduce-fminimum-intrinsic>. We have started <https://reviews.llvm.org/D155877> to align MLIR with LLVM, including the reduction flavors, but it's going to take some time...

================
Comment at: mlir/include/mlir/Dialect/Arith/IR/Arith.h:125
 /// Returns the identity value attribute associated with an AtomicRMWKind op.
+/// \p useOnlyFiniteValue defines whether the identity value should steer away
+/// from infinity representations or anything that is not a proper finite
----------------
We use markdown style instead of `\p` in MLIR

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D156471/new/

https://reviews.llvm.org/D156471