[PATCH] D143056: [RFC][IR] llvm.minimum/maximum NaN propagation.

[Dan Gohman via Phabricator via llvm-commits]

sunfish added inline comments.


================
Comment at: llvm/docs/LangRef.rst:14809
+-0.0 is considered to be less than +0.0 for this intrinsic. Note that these
+are the semantics specified in IEEE 754-2019.
 
----------------
Requiring the output NaN to be one of the input NaNs is stronger than what IEEE 754 requires (it describes NaN propagation as a recommendation but not as a requirement). And, it prevents implementing these intrinsics efficiently on ARM CPUs configured for [Default NaN mode](https://developer.arm.com/documentation/ddi0274/h/introduction/modes-of-operation/default-nan-mode), RISC-V (as far I can tell from the manual), and Wasm.



Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D143056/new/

https://reviews.llvm.org/D143056