[FFmpeg-devel] [PATCH] Transpose filter

[Reimar Döffinger]

On Mon, Oct 18, 2010 at 11:51:45PM +0200, Sebastian Vater wrote:
> Thanks for the benchmark, this is *VERY* interesting for me too, since I
> use lots of pointer arithmetic stuff like that instead of array indexing
> in my ported code (TuComposer => AVSequencer), too.
> 
> Strange enough, I had almost exactly the opposite results ~ 10 years ago
> when developed the original TuComposer code.

Usually the difference is due to gcc recognizing something as a loop
and something else indexed by the loop counter and then doing some
optimizations (like comparing the pointer address against &in[max]
and not keeping the x variable around at all).
Unfortunately gcc's detection of such a case is very brittle and
will usually break whenever you do something that is not 100% standard.