[FFmpeg-devel] Idea about speedup of startcode search

[Thorsten Jordan]

Michael Niedermayer schrieb:
> On Fri, Feb 08, 2008 at 02:35:11PM +0100, Thorsten Jordan wrote:
>> Hello,

ok, just one last comment, ff_avc_find_startcode is good for startcodes,
but what about the "emulation prevention 3 byte" filter code in h264.c?

even if gcc manages to boil it down to 3 cycles, with that mmx code one
can get even faster speeds... for 20mbit data streams (HDVT) that would
be 2.5 million bytes to scan per second. Not much for a 2Ghz core, but
would such an optimization wouldn't it be worth?

Yeah i know, if i would send a patch ... ;)

>>
>> 			     "packsswb %%mm0, %%mm0	\n\t"
>> 			     "packsswb %%mm1, %%mm1	\n\t"
>> 			     "por %%mm1, %%mm0		\n\t"
> 
> movq (%0), %%mm0
> por  1(%0), %%mm0
> pcmpeqb %%mm2, %%mm0
> packsswb %%mm0, %%mm0
> 
> and this is not a patch ...
it wasn't meant as one... ;)

-- 
Regards, Thorsten