[Ffmpeg-cvslog] r6602 - trunk/libavcodec/cabac.h

[Michael Niedermayer]

Hi

On Mon, Oct 09, 2006 at 06:20:32PM +0300, Uoti Urpala wrote:
> On Mon, 2006-10-09 at 16:15 +0200, michael wrote:
> > first try of a handwritten get_cabac() for x86, this is 10-20% faster on P3 depening on if you try to subtract the START/STOP_TIMER overhead
> 
> With -O3, new asm gives about same speed as the C version with redundant
> state setting removed. With -O2 it gives about 2% faster overall speed.

good and thanks for the benchmarking ...


> 
> Timer results with default -O4:
> 420 dezicycles in get_cabac, 134149491 runs, 68237 skips
> 
> with -O2:
> 322 dezicycles in get_cabac, 67101846 runs, 7018 skips
> 
> 
> The inlined versions with -O4 have extra instructions between rdtsc and
> get_cabac asm. The C versions didn't have such differences in timer
> results depending on -O2 / -O3.

ill reconsider how to benchmark get_cabac() ...

[...]
-- 
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