On Sun, 21 Sep 2003, Richard Ellis wrote:
> For the numbers I posted, the q value was 7. It is interesting that
> -N would have differing effects depending on the -q value. That may
Initially I thought it unusual but after thinking about it a
bit it's to be expected. The lower -q is the "harder" the
encoder will push up against the bitrate limit - the increase
in the high frequency quantizers have less effect if a high
quality (low -q) has been given.
> depend on how/where -N is applied within the encoder. If it's
> applied after -q has had it's effect, then that might explain why
> changing -q changes -N's effect.
Probably a good idea to mention that -N affects the quanization
matrices. It's not that the two options are separate operations
applied in sequence. -q sets the quantization scale and -N
affects the hf entries in the matrices. I think that setting a
low qscale causes the slight "rolloff" from low -N to have less
effect - eventually, at larger -N, the effect finally becomes
more noticeable.
The difference between bitrate (which I tend to use instead of
filesize) for -N 0.0 and -N 0.1 at -q 5 is 7379200 vs 7378400
while for -q 8 it was 7048000 vs 6896400
It should be an interesting table but will take another day or so
to complete.
> reports that it's using the MMX and SSE extensions that are available
> for a P4. Are all the MMX/SSE optimizations the same for both copies
> of mpeg2enc? If not, that would imply that the optimizations aren't
Supposedly MMX/SSE is enabled for all instances of mpeg2enc.
But that has to be the case - have you ever tried running a
non-MMX enabled mpeg2enc? Oh my god it's slow - about a
factor of 5 or so if memory serves me correctly (few years back
there was a kernel bug that prohibited MMX from working correctly,
I could run mpeg2enc but it was painful - thankfully the bug was
fixed and instead of .3 frames/sec a P3-800 can get around 1.5
or a bit less).
I suppose there could be mathlibrary differences - I'm not sure
how much mpeg2enc uses -lm routines.
I should try a run on the G4 Powerbook and see what it comes up
with.
Steven Schultz
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Mjpeg-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/mjpeg-users
