On Thu, May 17, 2001 at 09:06:04PM -0400, safemode wrote:
> > Lemme guess, you tried running aviplay at -20 and X at 0?
> > According to my experience, renicing X alone yields better results, if you
> > renice both aviplay and X everything is (almost) perfect.
> Actually the cvs of X runs itself at a sub 0 value by default. This is not a
> debian specific thing, it runs S<L
Aaaah will have to check it.
> I tried benchmark at -20, -5, -1, 0, 20 I found that with almost no cpu
> usage ( i still had konqueror open and kmail and freeamp and numerous Eterms,
> and gtk apps; they all dont use any cpu when not actually rendering or doing
> something active)
This may be true for tdfx and nvidia perhaps, but Xv on ATI 64 or 128 cause X
to take pretty much CPU. I dunno if it's a defficiency in hardware or ati
drivers, but renicing X helps.
I must admit I haven't compared voodoo Xv performance with various niceness
levels, my X runs at -20 defaultly cuz I previously had an ATI.
> 0 and 20 give you about the same numbers. 0 of course is
> your most optimal here. As for being perfect, it's not like i'm dropping
> frames on any movie at any size ( noting size doesn't matter here since no
> more cpu is used if it's at 1x or fullscreen), it's that it is being
> efficient.
At least for voodoo I am inclined to agree.
> I'd like to use the benchmark program to figure out where aviplay
> development can focus it's work on to increase efficiency by lessening sync
> time and such and the amount of cpu used for actual decompression.
Actually Zdenek done lots of improvements on lessining sync time after I
fucked it up with the caching stuff :-). As I posted before, on the voodoo I
got "Sync 0.000757 ms" for the 50 frames or so of Tiger/Dragon which is imho
pretty good.
> To really do that though you'd have to test it on different systems using
> the same clips and look at the general info, the wide range of systems
> should cancel out the extremeties and give you some real program performance
> data.
It ain't that easy as it looks, I speak from experience as I was tinkering
with it back in january/february.
> I think 2 clips should be sufficient unless data shows it to be unnecessary.
> one with a simple animation and another extremely complex. I think we all
> agree that the format should probably be in div3 and maybe asf since those
> are the most common media used with avifile. You really cant get as much of
> a performance increase as one would like though because the codecs are closed
> and that's that. As far as i can tell, the cvs is doing a good job compared
> to the stable version and hi fives to everyone in development.
Well currently decoding requires somewhat more CPU power than under windows
(at least for divx). This is (so I've heard) caused by not being able to use
faster colorspace decoder because it segfaults on initialization.
> aviplay beats MS Media player easily in avi playback and seeking, i mean
> obviously.
Good to know. I don't have windows so I can't compare. And the cool thing is
you can also seek asfs that are unseekable under windows (g).
> Not sure if it'll ever be possible to do the hardware filtering mediaplayer
> does to make edges less jagged, some kind of blurring, but that's fine.
Yeah I read about it somewhere. I may be wrong but I think the codecs do that
directly. Try using "--quality auto" parameter for aviplay, it seems to be
doing smoothing for me.
Bye,
Peter Surda (Shurdeek) <[EMAIL PROTECTED]>, ICQ 10236103, +436505122023
--
0 and 1. Now what could be so hard about that?
PGP signature
