PATCH] Improve generating FMA by adding a widening_mul pass
>
> As GCC's reassociation pass does not have knowledge of FMA, when
> transforming expression lists to parallel, it reduces the
> opportunities to generate FMAs. Currently there's a workaround
> on AArch64 (
As GCC's reassociation pass does not have knowledge of FMA, when
transforming expression lists to parallel, it reduces the
opportunities to generate FMAs. Currently there's a workaround
on AArch64 (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84114),
that is, to disable the parallelization with flo
