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 > Video Encoding > New and alternative video codecs

Reply
 
Thread Tools Search this Thread Display Modes
Old 19th January 2007, 05:10   #61  |  Link
BlindWanderer
Brain Dead
 
BlindWanderer's Avatar
 
Join Date: Nov 2003
Posts: 87
In the "Resize & aspect" page, the 'Multiply by' button text field is 1 pixel too long and overlaps with the text box.
__________________
I'll sell my soul for a cabbit.
BlindWanderer is offline   Reply With Quote
Old 19th January 2007, 14:22   #62  |  Link
haruhiko_yamagata
Registered User
 
Join Date: Feb 2006
Location: Japan
Posts: 1,560
Quote:
Originally Posted by Jeremy Duncan View Post
libavcodec decode:

Code:
int numthreads=deci->getParam2(IDFF_numLAVCdecThreads);
 if (numthreads>1 && mpeg12_codec(codecId))
  libavcodec->avcodec_thread_init(avctx,threadcount=numthreads);
 else
  threadcount=0;
-- omit --

Is this tweak being used ?
Does it improve multithreading ?
I think it's not working.

Code:
Thread=1 : User: 29s, kernel: 0s, total: 29s, real: 30s, fps:  41.2, dfps: 40.6
Thread=2 : User:  7s, kernel: 0s, total:  7s, real: 30s, fps: 162.1, dfps: 40.5
Both result "real = 30s". What does this mean?

By the way, what is the merit of using libavcodec-mpeg1/2?
With libmpeg2, HW deinterlacing was not working and it was crashing (or causing artifacts) on seek formerly.
Both of the bugs have been fixed untill now. Libmpeg2 has better performance in most cases. Will it change if the bug(?) of multithreading of libavcodec-mpeg2 is fixed?
Seeking in libavcodec-mpeg2 has issue. It causes artifacts untill key frame comes.

Last edited by haruhiko_yamagata; 19th January 2007 at 14:32.
haruhiko_yamagata is offline   Reply With Quote
Old 19th January 2007, 15:56   #63  |  Link
Jeremy Duncan
Didée Fan
 
Jeremy Duncan's Avatar
 
Join Date: Feb 2006
Location: Canada
Posts: 1,079
haruhiko_yamagata,

Please see post # 3537, # 3541 in this thread Link.

Here's a link to a FFDshow that uses the tweak I asked about.
Link
Jeremy Duncan is offline   Reply With Quote
Old 19th January 2007, 16:13   #64  |  Link
_xxl
ffdshow user
 
_xxl's Avatar
 
Join Date: Oct 2005
Location: Romania
Posts: 818
ffdshow

Multithreaded libavcodec mpeg1/2 de(en)coding is working for me.
1 thread:
Quote:
User: 7s, kernel: 0s, total: 7s, real: 7s, fps: 639.9, dfps: 634.7
2 threads:
Quote:
User: 1s, kernel: 0s, total: 1s, real: 5s, fps: 3008.6, dfps: 912.1
Small encoding test:
1 thread:
Quote:
1:52s
2 threads:
Quote:
1:42s
10s faster.

Last edited by _xxl; 19th January 2007 at 17:38.
_xxl is offline   Reply With Quote
Old 20th January 2007, 00:00   #65  |  Link
Romario
Registered User
 
Romario's Avatar
 
Join Date: Dec 2005
Location: Qetchua mountains in Peru, and Klingon battlecruiser D'Mar
Posts: 393
Well, not much faster, only 10 seconds for 2 threads???

That's very strange, it should be at least 50 % faster then 1 thread.
__________________
Live long and prosperLive long and prosperLive long and prosper
Romario is offline   Reply With Quote
Old 20th January 2007, 00:32   #66  |  Link
fastplayer
Registered User
 
Join Date: Nov 2006
Posts: 799
Quote:
Originally Posted by Romario View Post
That's very strange, it should be at least 50 % faster then 1 thread.
Really? Since when is every line of code "multi-threadable"?
fastplayer is offline   Reply With Quote
Old 20th January 2007, 01:01   #67  |  Link
F_L_C
Fun Lovin' Criminal
 
Join Date: Aug 2006
Posts: 16
Quote:
Originally Posted by Spaceman-Spiff View Post
I'm curious if SSE or MMX really makes a difference compared to normal builds.

I saw this on Wikipedia:
A common misconception is that ICL SSE/SSE2 builds will decode video better than "generic" builds. In fact, the video decoders are always compiled in gcc and are usually hand-optimized; it's the ffdshow filters that benefit from ICL.
The developers have said that ICL builds may have better performance if you use any of the filters. Resize is not affected but denoise3D, sharpen, and the rest are. Decoding speed has been shown to be nearly identical to generic builds.
F_L_C is offline   Reply With Quote
Old 20th January 2007, 05:31   #68  |  Link
haruhiko_yamagata
Registered User
 
Join Date: Feb 2006
Location: Japan
Posts: 1,560
I'm writing faq.htm.
Here's the results.
Code:
libavcodec-DivX5 Decoding 720p CinderellaMan_HD.avi
result(sec)=builds 
42=ffdshow_rev752_20070108_xxl.exe 
41=ffdshow_rev752_20070108_sse_xxl.exe
42=ffdshow_rev752_20070108_sse2_xxl.exe
42=ffdshow_rev756_20070109_clsid_icl9.exe
Code:
CinderellaMan_HD.avi + xsharpen
result(sec)=builds
148=ffdshow_rev752_20070108_xxl.exe
90 =ffdshow_rev752_20070108_sse_xxl.exe 
147=ffdshow_rev752_20070108_sse2_xxl.exe 
102=ffdshow_rev756_20070109_clsid_icl9.exe
Code:
CinderellaMan_HD.avi + Denoise 3d
result(sec)=builds
131=ffdshow_rev752_20070108_xxl.exe
130=ffdshow_rev752_20070108_sse_xxl.exe
115=ffdshow_rev756_20070109_clsid_icl9.exe
haruhiko_yamagata is offline   Reply With Quote
Old 20th January 2007, 09:33   #69  |  Link
_xxl
ffdshow user
 
_xxl's Avatar
 
Join Date: Oct 2005
Location: Romania
Posts: 818
ffdshow

Quote:
Originally Posted by Romario View Post
Well, not much faster, only 10 seconds for 2 threads???
That's very strange, it should be at least 50 % faster then 1 thread.
Quote:
Originally Posted by fastplayer View Post
Really? Since when is every line of code "multi-threadable"?
What is multithreaded?
Quote:
(ffmpeg 2772)
multithreaded/SMP motion estimation
multithreaded/SMP encoding for MPEG1/MPEG2/MPEG4/H263
Quote:
(ffmpeg 2778)
multithreaded spatial complexity calculation for i frames
Quote:
(ffmpeg 2810)
multithreaded mpeg2 decoding
Quote:
* DV decoding/encoding now supports MultiThreading for up to 324 CPUs ;-)
Quote:
Originally Posted by haruhiko_yamagata View Post
Here's the results.
Code:
CinderellaMan_HD.avi + xsharpen
result(sec)=builds
148=ffdshow_rev752_20070108_xxl.exe
90 =ffdshow_rev752_20070108_sse_xxl.exe 
147=ffdshow_rev752_20070108_sse2_xxl.exe 
102=ffdshow_rev756_20070109_clsid_icl9.exe
Code:
CinderellaMan_HD.avi + Denoise 3d
result(sec)=builds
131=ffdshow_rev752_20070108_xxl.exe
130=ffdshow_rev752_20070108_sse_xxl.exe
115=ffdshow_rev756_20070109_clsid_icl9.exe
More tests are needed.If SSE(2) compiled by MSVC2003 are faster or not?

Last edited by _xxl; 20th January 2007 at 10:30.
_xxl is offline   Reply With Quote
Old 20th January 2007, 11:47   #70  |  Link
haruhiko_yamagata
Registered User
 
Join Date: Feb 2006
Location: Japan
Posts: 1,560
I have added FAQ to the web.
Please fix my English,html and anything.
__________________
[ Download ffdshow | Wiki ]
haruhiko_yamagata is offline   Reply With Quote
Old 20th January 2007, 12:00   #71  |  Link
fastplayer
Registered User
 
Join Date: Nov 2006
Posts: 799
Quote:
Originally Posted by haruhiko_yamagata View Post
I have added FAQ to the web.
Please fix my English,html and anything.
I'm on it.

FAQ & Changelog updated.

Last edited by fastplayer; 20th January 2007 at 15:20.
fastplayer is offline   Reply With Quote
Old 20th January 2007, 16:19   #72  |  Link
cc979
Curious BetaTester
 
Join Date: Oct 2005
Posts: 430
Quote:
Originally Posted by fastplayer View Post
I'm on it.

FAQ & Changelog updated.
maybe you should add hali splitter has to be installed when using timecodec to faq
__________________
Asrock N68-S AMD Athlon(tm) II X4 620 Processor (2.6GHz) - Crucial 2GB PC6400 800MHz DDR2 - Nvidia 9600GT

Tools: ProcessExplorer & ProcessMonitor - BatchCompressor

Guide: MinGW Compiling GCC
cc979 is offline   Reply With Quote
Old 20th January 2007, 16:20   #73  |  Link
fastplayer
Registered User
 
Join Date: Nov 2006
Posts: 799
Will do.
fastplayer is offline   Reply With Quote
Old 20th January 2007, 18:39   #74  |  Link
clsid
*****
 
Join Date: Feb 2005
Posts: 5,642
Here are 6 builds for filter performance comparison.

download

MSVC2003: mmx/sse/sse2
ICL9: mmx/sse/sse2

Are there other filters besides xsharpen and Denoise3d that benefit from optimizations?
__________________
MPC-HC 2.1.7.2
clsid is offline   Reply With Quote
Old 20th January 2007, 19:15   #75  |  Link
fastplayer
Registered User
 
Join Date: Nov 2006
Posts: 799
From the FAQ:
Quote:
Generic builds (ICL 9.1) work on Pentium-MMX/AMD K6 or later.
Is this true? I thought SSE1 was minimum requirement for ICL.
fastplayer is offline   Reply With Quote
Old 20th January 2007, 20:18   #76  |  Link
clsid
*****
 
Join Date: Feb 2005
Posts: 5,642
ICL supports generic builds just fine. In fact, all my ffdshow ICL9 builds on sourceforge are generic (only MMX is required).
__________________
MPC-HC 2.1.7.2
clsid is offline   Reply With Quote
Old 20th January 2007, 22:07   #77  |  Link
F_L_C
Fun Lovin' Criminal
 
Join Date: Aug 2006
Posts: 16
This is what foxyshadis said about the builds and filters:

Quote:
Originally Posted by foxyshadis
Decoding and resizing do not benefit, but deinterlacing and many ffdshow filters (blur, sharpen, levels, properties, and the goofier less common filters) can show a noticeable boost, along with the non-lavc decoders (like libfaad and libdts). Like I said, it only means anything if you use ffdshow filters, instead of straight decoding.
Edit: and nice to see a new ICL build on the download page

Last edited by F_L_C; 20th January 2007 at 22:12.
F_L_C is offline   Reply With Quote
Old 20th January 2007, 22:24   #78  |  Link
fastplayer
Registered User
 
Join Date: Nov 2006
Posts: 799
Thanks, I'll add it to the FAQ later.

@clsid:
Please add links to the tryouts changelog and FAQ to the first post in this thread. Thanks!

Edit: FAQ updated. Please do a "spell check" ;-)

Last edited by fastplayer; 21st January 2007 at 13:47.
fastplayer is offline   Reply With Quote
Old 20th January 2007, 23:14   #79  |  Link
Inventive Software
Turkey Machine
 
Join Date: Jan 2005
Location: Lowestoft, UK (but visit lots of places with bribes [beer])
Posts: 1,953
Quick request while my mind's on it still: add MS WMV decoding for WMV7, WMV8, WMV9 as well as libavcodec...
__________________
On Discworld it is clearly recognized that million-to-one chances happen 9 times out of 10. If the hero did not overcome huge odds, what would be the point? Terry Pratchett - The Science Of Discworld
Inventive Software is offline   Reply With Quote
Old 21st January 2007, 01:06   #80  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
The heavy deinterlacers, kernel and tom's, and currently ICL-only in InnoSetup builds. They're pre-compiled, since they rarely change. So you get the benefit even with generic builds. =p

I could have sworn the WMV9 option was there before, I wonder what happened to it. I must have forgotten to add some dll.

drivel_xxl tells me that ICL can correctly compile lavc and might get a speed boost, but ICL hasn't worked right for me since I updated VC to SP1.
foxyshadis is offline   Reply With Quote
Reply

Tags
ffdshow, ffdshow tryouts, ffdshow-mt, ffplay, icl

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 10:03.


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