[FFmpeg-devel] [PATCH] avutil/x86/emms: Document the emms_c() vs alloc/free relation.

[Henrik Gramner]

On Mon, Oct 24, 2016 at 8:09 PM, wm4 <nfxjfg at googlemail.com> wrote:
> B) add a compile time option that runs emms() unconditionally at the end
>    of each mmx asm block
>
> Since musl intentionally evades detection, neither can be enabled
> automatically, probably. It would be interesting to see what the speed
> impact of B) actually is.

Depends on the function, but in small/simple DSP functions (which most
relevant MMX functions are) adding emms at the end will make the
entire function twice as slow on most Intel CPUs.

So it's certainly not negligible.