@aachrisg Same is true even on aarch64:
https://godbolt.org/z/TWenG8vbh
Conversation
Notices
-
Embed this notice
pinskia (pinskia@hachyderm.io)'s status on Sunday, 24-Nov-2024 07:02:14 JST pinskia -
Embed this notice
Chris Green (aachrisg@mastodon.gamedev.place)'s status on Sunday, 24-Nov-2024 07:02:19 JST Chris Green D-oh:
std:fma will generate the fused multiply add instruction in MSVC and GCC, regardless of the setting of fp:precise as you would hope..
But MSVC doesn't know how to auto vectorize it, which kind of defeats the purpose :-( :-(. GCC does though (as does Clang):
In conversation permalink
-
Embed this notice