On Wed, 23 Jun 2004, Jonathan Woithe wrote:
> I have encountered something rather odd about yuvdenoise from mjpegtools
> 1.6.2 and the way it works on two different machines. In short, I encoded
> a video in two 90 minute sections using the same command line - one was on a
> (PII-vintage) Celeron 466 and one on a 3.0GHz P4 with hyperthreading. The
Wow - you must have the patience of a saint to do encoding on a
Celeron 466! :)
> Has anyone else noticed this effect? It doesn't seem right that the output
No, can't say I've seen the effect but then I've been experimenting
with y4mdenoise (from cvs) or working with digital data that doesn't
require denoising at all.
> of yuvdenoise differs depending on what CPU you run the program on. Any
> thoughts?
Makes perfect sense - if there's a bug in the SSE routines ;)
There's also a chance that the bug has been fixed in the cvs version -
don't suppose you could give that a go and see if the problem's
still lurking about?
The vintage Celeron doesn't have SSE capability (I think the newer
ones do have it) so it uses only the MMX routines. The P4 on the
other hand does have SSE (and SSE2) and will use the SSE based routines.
You can probably speed things up a little on the P4 by disabling the
hyperthreading. I tried it and on a good day the speed hit was only
10 or 15% but some job mixes the system slowed down by closer to 30%.
When Intel comes out with their new true dual core cpus then that'll
be nice but till then I haven't had much success with the
hyperthreading's quasi 2nd cpu.
Cheers,
Steven Schultz
-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 -
digital self defense, top technical experts, no vendor pitches,
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
