[FFmpeg-devel] VP8 sliced threading

[Reimar Döffinger]

On Wed, Jul 11, 2012 at 11:26:14PM -0400, Daniel Kang wrote:
> My implementation no longer uses sched_yield().

I guess the preparational patches should be ok.
Also on a cosmetic note, I think FFmpeg nowhere uses
a newline between "}" and "else".
I believe that with the updated code, the compiler is
still allowed to move the update of thread_mb_pos whereever
it wants since it is not marked volatile (this currently
and particularly with x86 is the higher risk compared
to the CPU moving that write ahead of the relevant MB data
writes, though to be fully correct there need to be barriers
to make that impossible, too).