[XviD-devel] [RFC] Decoder speedup

Michael Militzer michael at xvid.org
Mon Jun 14 18:01:27 CEST 2004


Hi,

Quoting skal <skal at planet-d.net>:

> On Sun, 2004-06-13 at 22:32, Edouard Gomez wrote:
> 
> > BENCHMARKs: VC: 255,768s VO:   0,189s A: 0,000s Sys:  12,763s =  268,720s
> <-- 1.0.1
> > BENCHMARKs: VC: 249,709s VO:   0,193s A: 0,000s Sys:  13,287s =  263,189s
> <-- 1.0.1 + get_coeff optimization
> > BENCHMARKs: VC: 222,033s VO:   0,196s A: 0,000s Sys:  12,269s =  234,498s
> <-- head
> > BENCHMARKs: VC: 217,613s VO:   0,187s A: 0,000s Sys:  12,141s =  229,940s
> <-- head + get_coeff optimization
> > 
> > So now the reason why i call for comments...
> > 
> 
> 	What is the average quant of your bitstream??
> 
> 	Because there's a big difference between ffmpeg and xvid:
> 	ffmpeg is dequantizing the coeffs on-the-fly, so that
> 	this operation scales as the number of non-zero coeffs.
> 	XviD, contrary, dequantizes every 64 coeffs with SIMD.
> 	It's faster, but inefficient if only few coeffs are
> 	non-zero.

Skal is right - this may be espcially useful for clips with b-frames as
these usually have a bit larger quants. Speaking of that: what features did
your test clip had? B-frames? Qpel? GMC? Which Quant? What resolution?

Michael


More information about the XviD-devel mailing list