Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Capturing and Editing Video > Avisynth Usage

Reply
 
Thread Tools Search this Thread Display Modes
Old 9th June 2005, 11:33   #21  |  Link
Leak
ffdshow/AviSynth wrangler
 
Leak's Avatar
 
Join Date: Feb 2003
Location: Austria
Posts: 2,441
Quote:
Originally Posted by psme
Edit: sorry, just saw others are already running Decomb with MT but seems not much performance gain...
Ummm... I really wouldn't use MT with plugins that make important decisions by analyzing the whole image - if you're unlucky, you get one half of the image deinterlaced and the other weaved, or one halfs is matched forward and one matched backward, or one half is considered video and the other film etc...
Leak is offline   Reply With Quote
Old 10th June 2005, 06:56   #22  |  Link
mg262
Clouded
 
mg262's Avatar
 
Join Date: Jul 2003
Location: Cambridge, UK
Posts: 1,148
Quote:
Originally Posted by tsp
Also found a bug? in pixiedust. When using dynamic_cast Pixiedust throws __non_rtti_object exception. And it always complaines about "First-chance exception at 0x015327e2 in virtualdubmod.exe: 0xC000001D: Illegal Instruction" in the debugview.
I believe you can't call the dust filters more than once in a script, which your filter presumably implicitly does... one workaround is to load the DLL twice and use the DLLName_FilterName syntax.
mg262 is offline   Reply With Quote
Old 10th June 2005, 17:18   #23  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
Quote:
Originally Posted by mg262
I believe you can't call the dust filters more than once in a script, which your filter presumably implicitly does... one workaround is to load the DLL twice and use the DLLName_FilterName syntax.
yes that explains why it produces garbage but it doesn't explain why it crash when the dll is loaded the first time when dynamic_cast is used because at that time the constructor is not called yet so there is no way dust knows it is called many times. It's really a shame such a good filter is programmed so "bad" (and the source code is not available).

Quote:
Originally Posted by Leak
Ummm... I really wouldn't use MT with plugins that make important decisions by analyzing the whole image - if you're unlucky, you get one half of the image deinterlaced and the other weaved, or one halfs is matched forward and one matched backward, or one half is considered video and the other film etc...
You're right about that. If/when I succeed in making avisynth.dll threadsafe I will try to make an option to process two or more frames in parallel. So such filters can be used.
tsp is offline   Reply With Quote
Old 14th June 2005, 00:01   #24  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
new version ready that includes a custom build of avisynth 2.56 that should speed this filter up.
tsp is offline   Reply With Quote
Old 14th June 2005, 02:17   #25  |  Link
psme
Registered User
 
Join Date: Mar 2005
Posts: 68
Thanks for the great filter. Will try it tonight.

Will it speed up Didee's LimitedSharpen? I use it for realtime DVD playback using Avisynth in FFDShow. LimitedSharpen uses up most of my P4 3G Northwood and I can only do good flag 24fps 480p NTSC DVD. 25fps 576p PAL DVD gets stutter with 100% CPU. I'll get a dual core CPU if the filter help.

Thanks in advance.

regards,

Li On
psme is offline   Reply With Quote
Old 14th June 2005, 03:24   #26  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
Sorry for the late update; had problems.

Test was done reading from one drive and writing to another.
System is a Gigabyte GA-7DPXDWP, modified xp2400+ processors,
Tekram U160 SCSI Adaptor, Maxtor 73GB and 147GB u360 drives.

Test avi is an analog capture at 640x480.


--
Script:

lloadplugin("d:\plugins\DeComb521.dll")
#loadplugin("d:\plugins\MT_0.10.dll")
#loadplugin("d:\plugins\MT_0.29.1.dll")
loadplugin("d:\plugins\MT_0.30.dll")
loadplugin("d:\plugins\TBilateral-096.dll")

Avisource("d:\test-mt.avi").ConvertToYV12

Telecide(order=1,Post=0,Guide=1)
Decimate(Cycle=5,Mode=0,Quality=3)


#TBilateral(diameterL=3,diameterC=3,sDevL=2.0,sDevC=2.0,iDevL=6.0,iDevC=7.0,d2=true,gui=false)

mt("TBilateral(diameterL=3,diameterC=3,sDevL=2.0,sDevC=2.0,iDevL=6.0,iDevC=7.0,d2=true,gui=false)")

--



No MT = 522 seconds

MT 0.10 =
An out-of-bounds memory access (access violation) occurred in module 'TBilateral-096'.

MT 0.29.1 = 410 seconds

MT 0.30 = 534 seconds
hartford is offline   Reply With Quote
Old 14th June 2005, 03:26   #27  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
@tsp

Can't download v0.40
hartford is offline   Reply With Quote
Old 14th June 2005, 07:48   #28  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
sorry I uploaded it compresssed as a rar file instead of zip but that should be fixed now.

hartford: thanks for the test. I did fear that version 0.30 would be slower than without mt but when using version 0.40 and the custom avisynth.dll the version should be more like version 0.10 just without the crashing

psme: It should be able to speed Limitsharpen up if you use version 0.40 and your cpu utilization isn't 100%. I don't know how well MT will work with hyperthreading but please try and report back.
tsp is offline   Reply With Quote
Old 15th June 2005, 00:25   #29  |  Link
Shadowfax3000
Registered User
 
Join Date: Feb 2003
Posts: 16
Wow, 100% CPU load. This plugin is sweet! Before it I could only get CCE to use 50% of my CPU's power with the filters I'm using. I'm using a Pentium 4 3GHz Northwood (Hyperthreading). I'm using this crazy series of filters called TimeDenoise that makes it take days to encode a movie. Before this plugin I was only able to achieve a speed of 0.046 (about 72 hours to finish). Now I have 0.053 and climbing (41 hours to finish). Will shave 31 hours off the encode, can't complain about that.
Shadowfax3000 is offline   Reply With Quote
Old 15th June 2005, 13:39   #30  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
Shadowfax3000: that is impresive. I didn't think HT would make such a big difference. Wonder what a dualcore Pentium 4 EE would do.

psme: I did a little test on my dual celeron 400 MHz and using MT speed limitedsharpen up by ~12%(version 0.40) to 35%(version 0.41) using the default setting.

[Edit]
I just released version 0.41 it might be faster than version 0.40 at least it is on my computer but please try it and report back if it is slower or faster

Last edited by tsp; 16th June 2005 at 01:24.
tsp is offline   Reply With Quote
Old 16th June 2005, 11:08   #31  |  Link
psme
Registered User
 
Join Date: Mar 2005
Posts: 68
Thanks tsp! I may get a DFI 865PE-TAG board $90 (the only 865 with dual core support) and a Pentium D 820 $250 in the weekend.

I once had a BP6 with 2 Celeron 500M too!

regards,

Li On
psme is offline   Reply With Quote
Old 17th June 2005, 05:35   #32  |  Link
Shadowfax3000
Registered User
 
Join Date: Feb 2003
Posts: 16
No improvement for hyperthreading it seems like. After 20 minutes of encoding the speed reached 0.057 for both v.0.40 and v.0.41. Through testing I've noticed CCE 2.50 is faster with a Pentium 4 than any later version. I've tried 2.67 and 2.70, neither are as fast as 2.50. People have been saying that later versions are faster with the P4, I think they might be wrong. Or it's because of the filters and plugins I'm using. Anyways, I'll be on the look-out for new versions of this plugin, so consider me your hyperthreading man.
Shadowfax3000 is offline   Reply With Quote
Old 21st June 2005, 02:31   #33  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
Script and test avi the same as before.
Huffy Compression; read one drive, write to another
Audio included
Tests run back-to-back.


No MT: 539 seconds CPU time: 54%

MT 0.10 431 seconds CPU time: 65% to 74%

MT 0.41: 540 seconds CPU time: 54%


MT 0.41-1 with special build avisynth.dll

541 seconds CPU time: 50% to 58%



MT v0.10 isn't stable, but worked this time. It will error on loading
or it will work for the test.

If I have time, I will test on a longer capture (28000-30000 frames, or about 20 min)

Slight differences probably due to read drive being a bit full (slower read on inner tracks).

[Edit:] Clarification: Differences from June 13 test.

Last edited by hartford; 21st June 2005 at 02:38.
hartford is offline   Reply With Quote
Old 21st June 2005, 14:56   #34  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
hartford: did you use avisynthTS=true when you tested MT 0.41 with the custom avisynth.dll and also if you have the time could you test version 0.40 with the custom avisynth.dll and avisynthTS=true compaired to version 0.41.

Sligtly off-topic but try take a look at my dual processor test computer:
here and
here
tsp is offline   Reply With Quote
Old 23rd June 2005, 04:37   #35  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
Quote:
Originally Posted by tsp
hartford: did you use avisynthTS=true when you tested MT 0.41 with the custom avisynth.dll
No. I wasn't aware of that, um, toggle?


Quote:
Originally Posted by tsp
and also if you have the time could you test version 0.40 with the custom avisynth.dll and avisynthTS=true compaired to version 0.41.
I'll do that if you can tell me where to put "avisynthTS=true"




Interesting machine that you have there

[edit]
I need to report an error that I have using the custom DLL with this script:

vid=Avisource("01-Girl2.avi").converttorgb32()
logo=AVISource("d:\masks\mask-TVpg1-2.avi").ConvertToRGB32
vid2=vid.bilinearResize(147,110)
masklogo=mask(vid2,logo)
Layer(vid,logo,"mul",128,69,40)

I get an "undefined" access error from AviSynth. The script works ok with 2.55

[/edit]

Ok, I found "avisynthTS" explination in version 0.41.

Will report soonest.

Last edited by hartford; 23rd June 2005 at 04:59.
hartford is offline   Reply With Quote
Old 23rd June 2005, 21:37   #36  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
Quote:
Originally Posted by hartford
I get an "undefined" access error from AviSynth. The script works ok with 2.55
ups forgot a & in the code. Will fix that.

I'm working on integrating MT in the sourcecode so that's why there hasn't been a new version yet.
tsp is offline   Reply With Quote
Old 24th June 2005, 16:05   #37  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
Same script as above except added "AvisynthTS=true".

No MT = 519 seconds
MT v0.40a = 422 seconds
hartford is offline   Reply With Quote
Old 24th June 2005, 16:24   #38  |  Link
tsp
Registered User
 
tsp's Avatar
 
Join Date: Aug 2004
Location: Denmark
Posts: 807
and MT 0.41 with AvisynthTS=true?
tsp is offline   Reply With Quote
Old 25th June 2005, 04:23   #39  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
No Mt 519 seconds

MT v0.41 415 seconds
hartford is offline   Reply With Quote
Old 29th June 2005, 03:16   #40  |  Link
hartford
Registered User
 
hartford's Avatar
 
Join Date: Nov 2003
Posts: 324
@TSP

Is this near release?
hartford is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 14:25.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.