[PATCH] D139902: IR: Add nofpclass parameter attribute

[Joshua Cranmer via Phabricator via llvm-commits]

jcranmer-intel added inline comments.


================
Comment at: llvm/docs/LangRef.rst:1424
+   <denormal_fp_math>` indicating zero treatment of input denormals,
+   it cannot be assumed the value is 0; it may still be a denormal.
+
----------------
nikic wrote:
> I don't really get what this whole paragraph is trying to say.
> 
> Please at least write out the 0xf0.
I think this is trying to say that even if we know that the function is in DAZ mode, a subnormal value is not turned into poison by `nofpclass(zero)`. But I agree with the wording being weird.


================
Comment at: llvm/docs/LangRef.rst:1426
+
+   In textual IR, various string names are supported for readability.
+
----------------
Maybe mention how to combine two kinds of things into one mask, like `nofpclass(sub zero)`?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D139902/new/

https://reviews.llvm.org/D139902