[flang-commits] [flang] [flang] Respect math flags during conversions (PR #214311)

[via flang-commits]

jon-gibney wrote:

> Can you please explain which targets are affected by "expensive overflow/NAN checks"? In addition: what is the motivating example for this change and how other compilers handle FP-to-INT conversions (with and w/o fast math)?

I've noticed this with x86 Genoa; haven't investigated others. Motivated by differences in performance between our HPE Cray compiler and Flang with the PERFECT/BDNA benchmark. Would it be helpful if I looked into what happens with other targets/compilers?

https://github.com/llvm/llvm-project/pull/214311