[Ffmpeg-devel] benchmark of different CABAC routines

[Michael Niedermayer]

Hi

On Wed, Oct 11, 2006 at 10:30:15PM +0300, Uoti Urpala wrote:
> On Wed, 2006-10-11 at 21:30 +0300, Uoti Urpala wrote:
> > branchless cmov: 4890
> > branchless no-cmov: 4834
> > non-branchless: 4553
> > non-branchless C: 5058
> > non-branchless C, modified: 4996
> > branchless C: 5440
> 
> Results with h264.c compiled with -O2 are somewhat different:
> branchless cmov: 5054
> branchless no-cmov: 5185
> non-branchless: 5080
> non-branchless C: 5350
> non-branchless C, modified: 5435
> branchless C: 6067
> 
> Either this means that inlining effects are significant or that
> "irrelevant" code changes affect the time noticeably. 

yes ill bet on the later, gcc is a great random number generator ive also
seem large changes in speed for small changes in c code for example i tried
several times to remove the bit variable, after all just s^=1 in the second
branch of the loop and return s&1 will work just fine but the code gets 
slower ...

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

In the past you could go to a library and read, borrow or copy any book
Today you'd get arrested for mere telling someone where the library is