https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110305
--- Comment #2 from tkisuki at tachyum dot com --- (In reply to Andrew Pinski from comment #1) > First off using signal nans but not caring about signed zeros seems a combo > that normally won't happen in real life. Did you find this by accident or > where you testing this combo on purpose with real code you are writing. . Thank you for reviewing. We are developing own architecture and gcc backend for it. My colleague found this case and reported it to me. So, it's more like accidental and we don't need this fix for existing architectures, like x86 or aarch64. I just wanted to share what my colleague reported in the upstream.
