[PATCH] D75576: [SDAG] simplify FP binops to undef

Sanjay Patel via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Mar 4 05:06:36 PST 2020


spatel added a comment.

In D75576#1904936 <https://reviews.llvm.org/D75576#1904936>, @lebedev.ri wrote:

> LGTM, thank you.
>  If we don't do this in middle-end, we should.


Yes, that would be a follow-on. There's less risk/reward in making the change here vs. IR, so I started here to be safer.

> I'm guessing this is already done for `FNEG`?
>  What about `FCMP`/`PHI`/`SELECT`?

FNEG - no. Similarly, we don't do much/anything with opcodes that exist in SDAG but not IR. 
FCMP is still in flux. We allow FMF on those currently, but I've been trying to transition away from that. So PHI/SELECT are relatively new to allow FMF in IR, and that has not been extended to SDAG yet.
In summary, we're a long way from complete in either layer.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75576/new/

https://reviews.llvm.org/D75576





More information about the llvm-commits mailing list