[Ffmpeg-devel] [PATCH] SIMD accelerated SNOW decoding

[Guillaume POIRIER]

Hi,


On 12/3/05, Reimar D?ffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> Hi,
> On Wed, Nov 30, 2005 at 11:24:52PM +0100, Guillaume POIRIER wrote:
> > Please find in attachment the fixed files, which should be tested, if
> > possible, on IA-32 and AMD64.
>
> Works fine for me, but no speedup at all with my sample.

That's unfortunate. Robert Swain on IRC reported a speed-up around
5-7% on IA32. I did not measure it on my machine though as I can't
access it right now. Maybe there's no speed-up at all because the
relevant code is not called?


> > Now I'm off to find a solution to fix 'inner_add_yblock_mmx' (which
> > still segfault when it executes the instruction 'movdqa
> > (%rdi),%xmm3').
>
> Activating this again finally gives a speedup, around 10%, and no
> crashes with my sample. Maybe you can make yours available?
> If it doesn't crash either I suspect either compiler bug (I used gcc
> 4.0.0 from gentoo) or a missing/wrong clobber list.
> Or did I miss anything else? I didn't have too close a look at it.

I confirm that inner_add_yblock_mmx was crashing on my machine. I
don't remember the compiler I was using though.
But now it's even better, it's not compilable anymore here since last
commits that have been done in lavc. It's too late for me right now to
try to fix it. What's more, given the more than few posts that this
thread received, it looks like not many ppl seem to care about having
a faster Snow implementation or care about Snow altogether. Bummer.

My test sample is available here:
http://tuxrip.free.fr/transperl/MPlayer/test.avi

You can find in attachment an updated version of the patch that can be
applied against latest CVS.

Guillaume

--
MPlayer's doc is offline. Find some fresh one here:
http://tuxrip.free.fr//MPlayer-DOCS-HTML/en/
http://tuxrip.free.fr//MPlayer-DOCS-HTML/fr/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: new_snow.patch
Type: text/x-patch
Size: 32222 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20051218/f66bcd45/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: snow_mmx_sse2.h
Type: text/x-chdr
Size: 47856 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20051218/f66bcd45/attachment.h>