[PATCH] D26855: New unsafe-fp-math implementation for X86 target

[Sanjay Patel via Phabricator via llvm-commits]

spatel added a comment.

I just re-read some of the comments in this review, and back on Nov 30 I explained the possible compile-time hit. I requested a note to llvm-dev at that time. Based on the confusion in the subsequent comments (people think this patch will affect execution perf), I am making the suggestion to post to llvm-dev again. Moving from DAG heuristics to MI-scheduler-based transforms is a change in strategy for all targets and explaining that to a wider audience is an opportunity to get good feedback.


https://reviews.llvm.org/D26855