[PATCH] D53157: Teach the IRBuilder about constrained fadd and friends

[Cameron McInally via Phabricator via cfe-commits]

cameron.mcinally added a comment.

In https://reviews.llvm.org/D53157#1304900, @andrew.w.kaylor wrote:

> I think at this point we're all on the same page in this regard. I just wanted to make sure we also understand why we're on that page. I still believe it was the correct choice.


Yup, we're on the same page.

To explain where I'm coming from, our in-house compiler has tracked down *many* of the trap-unsafe transformations in LLVM and can toggle them as desired. The execution penalty for running trap-safe code is (much) less than 5% with that compiler. My goal for this project is to make Clang competitive with that compiler.


https://reviews.llvm.org/D53157