jyknight wrote: > You'll get different results depending on whether the input is implicitly > flushed in fcmp vs. not in the is.fpclass
This sounds intuitively like the sort of semantics-breaking optimization which is expected from `-ffast-math`. If the only issues are things like getting a slightly-wrong answer from is.fpclass if you somehow pass a denormal to it, when you've disabled denormal handling...that just doesn't seem so bad. > Arguably, the only safe value is dynamic (compiler doesn't know what it's set > to). Yes, we could change the default to "dynamic" for _all_ builds (except on platforms where FTZ/DAZ flags don't exist). Potentially doing so is justified regardless of `-ffast-math`, since users can of course use `fesetenv` or `_mm_setcsr` to enable these flags themselves. (Sidenote: "dynamic" isn't even [documented](https://clang.llvm.org/docs/UsersManual.html#cmdoption-fdenormal-fp-math)). > So it's a matter of which behavior is going to be the least likely to be > incorrect in practice. So, alternatively...we could just go with the simplest solution, and use "ieee" as the default even under `-ffast-math`. https://github.com/llvm/llvm-project/pull/80475 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
