[FFmpeg-cvslog] r30965 - in trunk/libswscale: rgb2rgb.c rgb2rgb_template.c swscale.c swscale_template.c utils.c

Michael Niedermayer michaelni
Thu Apr 1 00:54:34 CEST 2010

On Wed, Mar 31, 2010 at 05:29:17PM -0400, Ronald S. Bultje wrote:
> Hi,
> On Wed, Mar 31, 2010 at 3:20 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> > On Wed, Mar 31, 2010 at 07:10:49PM +0200, Reimar D?ffinger wrote:
> >> You know, you could try to fix it yourself, there's nothing particularly
> >> difficult about those...
> >
> > yes, more asm coders are welcome
> I will try. gcc-asm syntax looks scary, especially for a newbie. Maybe
> you (Michael) should make a wiki page with some of the optimization
> tasks you've suggested in the past, sort of to serve as a quick-start,
> and maybe a way to test (START/STOP_TIMER, or whatever, and which
> samples to test, etc.) the effect. I'd like to get started on
> optimization-related work but other (easier-reachable, from my
> perspective) tasks tend to get my preference at the end of the day.

i leave it to others to collect my random disorganized notes and raise
them to the wiki

about someting easy to try to optim take:
void ff_scale_vector_to_given_sum_of_squares(float *out, const float *in,
                                             float sum_of_squares, const int n)
    int i;
    float scalefactor = ff_dot_productf(in, in, n);
    if (scalefactor)
        scalefactor = sqrt(sum_of_squares / scalefactor);
    for (i = 0; i < n; i++)
        out[i] = in[i] * scalefactor;

the for loop more precissely

about gcc asm i would assume there must be a good tutorial somewhere
and writing another would border on NIH syndrom. Anyone knows a good

Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

No great genius has ever existed without some touch of madness. -- Aristotle
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-cvslog/attachments/20100401/18b97921/attachment.pgp>

More information about the ffmpeg-cvslog mailing list