uManiac's instant build 03.26.2002

[wing1]

Interlace support is here!!!! whoohooo :D

-h u did it heh? download and going to test it for capture to see how it fairs.

[kastro68]

Where can i go to read up on the interlace support.

I would like to find out theory behind, what it does and when to use it.

thx

[uManiac]

According to h on the XviD board (http://www.xvid.org/forum/viewtopic.php?topic=413&forum=3&0) , the code is there, but it can't be used yet (to buggy). He says he might have it fixed within a week. We'll have to see. :)

uManiac

[Koepi]

Interlacing support just works for I frames up to now. (-h told me that on a chat).
P-Frame support is still buggy and thus disabled.

So you can use it when encoding i-frame only....

Regards,
Koepi

[wing1]

@kastro68

get it at uManiacs's site. Please refer to xvid Q&A thread to link to his site or Nic's and Koepi's urls. Currently uManiac's link has it in the instant build section.

Preliminary tests came back with the following observations :

1. 640x480 capture max out at 100% on my 1.467Ghz AMD. Frame doesn't drop but it is running at maximum at all time. Not so good for CPU's temp :D . Reduced to 544 x 400 with crop enable (-32x & -16y) = 512x384 CPU is running at 24%.

2. You still can see the interlace lines at playback on PC monitor. Wondering if this is going to look good on interlaced devices :) Quality is quite good. At 900 1-pass CBR data rate flows steady at 125-150K/s (not bad) and 1000+ (155-225K/s).

3. Nic's DSfilter bombed WMP6.4. You have to use the included DSfilter if you wanna play back on computer :D. Furthermore, WMP crashed immediately after you shut it down :(

more testings....

[Nic]

Unfortunatly I wont be able to update things for a little while :(
May have to wait till easter sunday :(

-Nic

[lar1r]

Hey Umaniac,

Are your builds optimized by your compiler?
Just curious to know if there is that much of a speed difference b/w Nics builds.

Note: I love the way you set up Instant Builds on your site. Is the process automatic - if it is then that is impressive.

[uManiac]

The builds in the instant build section are built using the release profile (configuration) in XviD. This tries to maximize speed, but I think it's possible to squeeze more speed out of it as Nic has sayd.

The process is completly automated. It's a very simple combination of Windows Task scheduling and using command line interfaces for everything. :)

uManiac

[-h]

Well I-frame interlacing is working fine, but is disabled unless you change xvid.h when compiling. Did you do this wing1? If I've slowed down "normal" encoding, this is very bad..

I still have this weird error with P-frames (think it's ac prediction) - I'd rather work that out than commit mmx/xmm optimisations for code which could be buggy. Hopefully tonight, as I'll have some time.

Then field motion estimation tomorrow night.

-h

[wing1]

@-h

I download the binaries from uManiacs's page and ran the test only. Sorry for the uproar ( the release just got me so excited :D ). If it is indeed disabled in the binaries, then I should be able to play the capture back using Nic's DSfilter? However, the clip crashed immediately and only uManiac's xvid.ax will open up the clip. Even then it crashes sporatically.

Quality wise, it is much better then previous xvid builds ( It is the first test I ran with every builds for interlace capture :D ).

Once again, I appologized for the mis-information ::rolleyes:

[avih]

it doesn't seems a filter issue, although there could be a bug since they behave slightly different.

i think it should play well with nic's filter, even when interlaced and b-frames are enabled.

cheers
avi

[wing1]

@avih

I was using Nic's latest DSfilter ( 03.23.2002 xvid.ax 188K ) and the clip crashed in WMP6.4 immediately. I removed it and replaced it with the installed xvid.ax from uManiac's same build and it works fine then. Still I do get a crash here and there afterward. I can't explain it at all.

Removed uManiac's 03.26.2002 build and installed his 03.25.2002 and everything is back to normal once again.

[avih]

i still think the filter shouldn't be affected by the changes, and u might have found a bug.

nevertheless, nic would be able to give better answer.