[PATCH] D68651: [InstCombine] Signed saturation patterns

[Hideki Saito via Phabricator via llvm-commits]

hsaito added a comment.

> I haven't looked at the patch in detail, but as author of at least part of the prior art cited here, I agree with the direction*. I also participated in some of the vector idioms discussions from a few years ago. There's overlap with the vector idiom problems, but as noted, these are generic (scalar too) math ops, so it's not exactly the same. We invested significantly in IR analysis and codegen for the math intrinsics, so that may have changed the thinking. I don't remember the sequence of events or if there was a dedicated llvm-dev thread for this, but the general idea is that if we have a generic intrinsic for the math and can easily invert the transform in the backend for targets/types that are not supported, try to canonicalize to the intrinsic.

Don't get me wrong. I'm not against the direction. I just want to make sure we have a general agreement with the rest of the community on where and how to draw the line ---- and get it documented (and the document to be kept up-to-date). Then, we can go back to the idiom list we created and determine which ones should be canonicalized, and which ones should go to better pattern matchers --- reflect that into the document. That'll make it easier for other interested people to actually work on those, w/o having to make a point on individual cases. That's the main purpose of raising this question.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D68651/new/

https://reviews.llvm.org/D68651