On Fri, 3 Jul 2020, James Almer wrote:
On 7/3/2020 9:00 AM, Lynne wrote:
Maybe they've fixed the FATE failures.
We have done this at least two times before, and all of them limited to
the newest compilers which supposedly were good. It always resulted in
compilations issues being reported sooner or later, and us having to
remove it again.
If there's issues with the latest compiler versions, it would be good if
the issues would be reproduced, reduced and reported - otherwise the bugs
probably won't get fixed ever.
In general, autovectorization does help quite notably in cases when
there's no SIMD yet. At least on dav1d's DSP routines on aarch64, GCC's
autovectorization does a better job than clang (without misoptimizations).
// Martin
_______________________________________________
ffmpeg-devel mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".
