CoreCodec/H.264 Codec "CoreAVC" [Archive] - Page 75

17th March 2008, 11:55

a decoder will output the same image as the next
its just DXVA doesn't allow for post processing
You mean inloop deblocking, since H.264 decoders rarely do any actual postprocessing? If a decoder doesn't do inloop deblocking, it will produce broken output, which is not the same as output from a proper H.264 decoder.

However, the latest GPU generations from NVIDIA and ATI should have no problems with deblocking, even through DXVA. On earlier GPUs, some decoders skipped deblocking because it slowed down hardware decoding too much (http://forum.doom9.org/showthread.php?p=1007475&highlight=cyberlink#post1007475) (either because of drivers or the hardware itself).

CruNcher

17th March 2008, 12:53

I think he means the Visually different Representation of the Colorimetry and Brightness Levels compared to other Decoders (without render dependency on either TV or PC Source VMR7/VMR9). (It seems (not sure here) to set it from the Streams VUI information and will always set it accordingly, you gonna never see compression artifacts in the luminance with CoreAVC but wrongly calibrated with another Decoder you will). You could say the Decoder doesn't need any calibration from the Renderer it calibrates the renderer to the source or itself to the renderer for optimal source presentation :)

Shinigami-Sama

17th March 2008, 20:20

You mean inloop deblocking, since H.264 decoders rarely do any actual postprocessing? If a decoder doesn't do inloop deblocking, it will produce broken output, which is not the same as output from a proper H.264 decoder.

However, the latest GPU generations from NVIDIA and ATI should have no problems with deblocking, even through DXVA. On earlier GPUs, some decoders skipped deblocking because it slowed down hardware decoding too much (http://forum.doom9.org/showthread.php?p=1007475&highlight=cyberlink#post1007475) (either because of drivers or the hardware itself).

mpc-hc uses full(bitstream) acceleration
so I'd assume the UVD does the in-loop deblocking, because I can't see them being able to brand it as an AVC accellerator without being able to decode properly

now DXVA2 sure, 'cause you can break up the decoding into different parts I can see parts being missed there

but not in bitstream mode

Jay Bee

18th March 2008, 00:37

because I can't see them being able to brand it as an AVC accellerator without being able to decode properly

That's exactly what they do, both ATI and Nvidia. Not only for cards that don't do deblocking, even for cards that don't do any acceleration at all (x1950 vc-1).

ADude

18th March 2008, 20:21

FYI, I just upgraded to Vista SP1 (downloaded from Microsoft Download Center) and no problems whatsoever. Definitely a lot faster, the lag in opening a folder has gone.

CoreAVC Professional 1.6.5 still working fine.

madshi

21st March 2008, 22:24

lol whut?

a decoder will output the same image as the next
its just DXVA doesn't allow for post processing
Get your facts straight before laughing at other people. You'll get 100% the same results only when you're talking about lossless compression (and bug free decoders, obviously). With lossy decoders it's absolutely usual that different decoders are outputting slightly different results. Granted, the differences are not very big, but there are differences. You can compare for yourself: Try different decoders, make screenshots and then run a hash (e.g. MD5) over the screenshots. You can even do it multiple times. The screenshots of the same decoder will always have the same hash. But screenshots of other decoders usually differ.

Dark Shikari

21st March 2008, 22:46

With lossy decoders it's absolutely usual that different decoders are outputting slightly different results.At least for H.264, any decoder that does so is in blatant violation of the standard and should not be used.

What you are saying is only true of standards that do not specify an exact DCT approximation, such as MPEG-4 ASP and MPEG-2.

Rectal Prolapse

22nd March 2008, 00:42

I would love to run CoreAVC (and other decoders) through an AVC compliance checker - there are several such tools available (for $$$) that can be used to evaluate decoder quality.

I recall reading somewhere that all AVC decoders have to have identical output compared to the official reference decoder, in order to be allowed the AVC label.

I wish it were true - it seems everyone gets away with making an AVC decoder that breaks the rules. :)

Ranguvar

22nd March 2008, 00:43

Lossy decoders?? Thank dog I have not run up against such a malformed piece of code yet ;)

me7

22nd March 2008, 04:34

Dark Shikari

22nd March 2008, 04:41

I used the CoreCodec on my Vista SP1 system and I'm pleased with the decoding speed. Today I installed PowerDVD for Blu-Ray playback and now Windows Media Player uses the CyberLink decoder instead of CoreCodec. The CyberLink codec has issues with some of my mkv files and stutters extremely, so I'd like to deactivate it and use CoreCodec instead. The problem is I can't find any option to turn it off. Neither in WMP, nor in the codec settings. Google didn't come up with anything helpful either (only a tip for XP that doesn't work anymore).1. Download the CCCP Insurgent (http://www.cccp-project.net/).

2. Run a Test Render on the problem file.

3. Find the problematic Cyberlink codec in the test render results.

4. Find the same codec on the list of installed DirectShow filters on your system.

5. Go to advanced mode, and unregister the problematic Cyberlink filter.

clsid

22nd March 2008, 15:28

Is it save to disable the Cyberlink filter without breaking PowerDVD?

The latest version of CoreAVC (1.6.5) has an option to increase its merit. Use that and WMP should start using CoreAVC again.

Dark Shikari

22nd March 2008, 15:29

clsid

22nd March 2008, 15:55

Yes, I believe it was added specifically for this Cyberlink problem.

me7

22nd March 2008, 18:39

Is it save to disable the Cyberlink filter without breaking PowerDVD?

The latest version of CoreAVC (1.6.5) has an option to increase its merit. Use that and WMP should start using CoreAVC again.

I guess you mean the "Preferred decoder" setting? I just found out that it works, but only if you run the configurator in admin mode (right-click on app\run as admin). Yesterday I tried the same option in normal mode but nothing happened. I guess it's a bug that should be fixed in the next version.

clsid

22nd March 2008, 19:55

That setting simply requires administrator privileges for the registry changes that it needs to make.

ADude

23rd March 2008, 00:16

In Vista, you can't change anything unless you are in Administrator mode.

I have changed things in User mode, and then the next day, Vista decided that an evil unauthorized force had made the changes, and reverted them to their original values.

me7

23rd March 2008, 12:09

madshi

24th March 2008, 17:37

At least for H.264, any decoder that does so is in blatant violation of the standard and should not be used.

What you are saying is only true of standards that do not specify an exact DCT approximation, such as MPEG-4 ASP and MPEG-2.
Ok, I didn't know that the h264 specification was stricter than other comparable specifictions. Thanks for the information.

However, I read in another thread (on another forum, IIRC) that people ran an MD5 over screenshots produced by different h264 decoders and the MD5 was different with each decoder - except for those decoders which were based on the same source code.

Lossy decoders?? Thank dog I have not run up against such a malformed piece of code yet ;)
So why do the open source AC3 decoders output different results compared to what Dolby reference AC3 decoders output? Same with DTS. Also the MPEG2 decoders output different results. As I said, the differences are not very big, but they are there.

DigitalDeviant

24th March 2008, 19:03

Ok, I didn't know that the h264 specification was stricter than other comparable specifictions. Thanks for the information.

However, I read in another thread (on another forum, IIRC) that people ran an MD5 over screenshots produced by different h264 decoders and the MD5 was different with each decoder - except for those decoders which were based on the same source code.

Probably the decoders were doing some sort of postprocessing, like TV->PC levels conversion.

ADude

25th March 2008, 01:18

I am loged in as admin, but apps still need permission from you to change sysem critical data like the registry. Normally a UAC dialog comes up and asks you whether you want to allow the app to mess with your system.
In this case, no UAC dialog came up. The CoreCodec configurator didn't ask for the admin's permission, it just ignored the setting I've set. I had to give it my permission manually via "right-click\run as admin". I'd say this is a feature that roughly 1% of windows users are aware of, so something needs to be done here.

Yes, I find that Vista requires applications to be aware of new methods, and to use them in installation. For example, Zoom Player never really grasped that or researched it (the developer did not even acquire Vista), and that was one of the reasons I changed to Media Player Classic - Home Cinema.

By the way, in Vista, when you are running an account that quote "is an Administrator", you still don't have full priveleges - do a web search for Vista Super Administrator.

BetaBoy

3rd April 2008, 01:37

All.... a quick update.... we are on QA for CoreAVC 1.7 atm and found a few small bugs and one major bug (related to the NG Filter SMP code) that testers have come across.

I'll keep everyone posted on where were at later in the week.

Sasovics

3rd April 2008, 01:43

Is 1.7 version fixing our long-lasting AviSynth compatibility problem ?

BetaBoy

3rd April 2008, 03:05

Yes... all pin connector bugs are fixed... we also found a bug in Avisynth that was related to CoreAVC connecting to it. We were however able to create a work around for it in 1.7... and we sent off a report to them about it.

BetaBoy

4th April 2008, 07:37

Major bug fixed.... also SMP core load distribution should be improved against i-frames.... moving on to the last of the smaller issues.

chrislynch

4th April 2008, 19:19

Major bug fixed.... also SMP core load distribution should be improved against i-frames.... moving on to the last of the smaller issues.

I am eagerly waiting for this release! :thanks:

Ice =A=

5th April 2008, 13:35

Thanks for the info!

Sasovics

14th April 2008, 20:40

Any ETA for releasing 1.7 version ? Can't wait to have it ;)

HowlerX

14th April 2008, 22:55

Please don't give an ETA. This thread is long enough. Don't wanna read useless posts regarding long, long overdue ETAs.

Just wait till it's done.

Disabled

16th April 2008, 17:59

1.7 is out:
CoreAVC H.264 Video Codec - Version 1.7.0.0 (20080415)
- Add: Support for Mainconcept and ArcSoft demuxers.
- Add: Workaround for broken DirectShowSource in AviSynth
- Add: Installer improved
- Add: Better multiple CPUs/Cores balance
- Fix: Others internal fixes

Haali Media Splitter (20080329)
- Add: Added support for muxing FLAC audio as A_FLAC to the muxer
- Add: Added support for parsing recovery point SEIs in H.264 streams in PS and TS containers
- Add: Added an automatic colorspace selection option to the renderer, it switches to BT.709 when video width is 1024 or more
- Fix: Added a workaround for some broken matroska files with trailing 0s
- Fix: Disabled the PS3 upscaler since it doesn't work properly on all hardware
- Fix: Fixed handling of VC-1 with changing sequence headers in PS and TS containers
- Fix: Fixed VC-1 aspect ratio detection
- Fix: Added support for PCM audio with 7 and 8 channels
- Fix: Fixed channel mapping for PCM in M2TS container
- Fix: Added a workaround to make MS VC-1 DMO decoder accept formats with included aspect

ratio information
- Fix: Fixed the misbehaving Apply button in color planes shift dialog in the renderer

BlackSun

16th April 2008, 18:00

omg, you guys are too fast for me !

BetaBoy

16th April 2008, 18:00

We have just released our Milestone build for CoreAVC 1.7 Standard and Professional Editions. This is by far our fastest, and most polished version to date. We are sending update notifications now.

We have fixed all reported issues here at D9 but we are also aware of 'other' things we are looking into now. The biggest of which is in Vista the 'preferred decoder' does not register properly but can be manually set if the config panel is allowed to be run by an admin.

Pls post anything else you find... Thanx!

BetaBoy

16th April 2008, 18:05

Also... as promised with 1.7 we have now posted the CoreAVC Professional Edition 14 day Trial Version @ http://www.coreavc.com

Jay Bee

16th April 2008, 18:18

I'll save my energy, no opinion, just facts:

-EVR on Win XP doesn't work. The image freezes and has the wrong Aspect Ratio. Worked in v1.5. Seems to work when YV12 is disabled.

-The field order is wrong when hardware deinterlacing is on, no matter which renderer (do I really need to link to the sample clips at x264.nl AGAIN?).

-Some videos (Blu Ray remuxes) that are smooth with v1.5 stutter severely.

-This is the third release in about six months that is so broken that I have to go back to an older version.

juGGaKNot

16th April 2008, 18:31

Also... as promised with 1.7 we have now posted the CoreAVC Professional Edition 14 day Trial Version @ http://www.coreavc.com

will give it a try ! THANKS.

Cheers.

juGGaKNot.

BetaBoy

16th April 2008, 19:01

Jay Bee... I'm not go into this again... EVR on XP is a hack.... and field order tests fine here but we'll look into it. As far as your BR 'remux'... Do you have a sample?

ADude

16th April 2008, 19:18

1.7 working fine on 1080p mkv files, two cores look very well balanced, thanks !

ADude

16th April 2008, 19:22

Since CoreAVC installs Haali Splitter, I thought I would ask here:

Is there any reason not to let Haali Splitter do AVI as well ?

CiNcH

16th April 2008, 19:23

Tried the trial with DVBViewer. Hardware deinterlacing operates at 50fps but VMR9 shows a jitter of over 10ms which manifests itself in clearly visible stuttering.

bob0r

16th April 2008, 19:36

Tried the trial with DVBViewer. Hardware deinterlacing operates at 50fps but VMR9 shows a jitter of over 10ms which manifests itself in clearly visible stuttering.

Dvbviewer coder was unable to reproduce our problems, maybe BlackSun or BetaBoy should contact him again, and give him coreavc 1.7 so he can try again and update dvbviewer.

Since i got my Quad core, coreavc works flawlessly with dvbviewer, ALL H.264 satellite channels (http://en.kingofsat.net/hdtv.php) work perfectly, so yeah issues issues issues, if only EVERYONE would work with FIXED standards, all our lives would be easier :D

CiNcH

16th April 2008, 20:06

CPU usage is not much below the MainConcept which requires quite some CPU time...

http://members.inode.at/762450/coreavc/coreavc_dvbv.jpg

MainConcept runs with 0ms jitter (also hardware deinterlacing).

Jay Bee

16th April 2008, 20:29

Tried the trial with DVBViewer. Hardware deinterlacing operates at 50fps but VMR9 shows a jitter of over 10ms which manifests itself in clearly visible stuttering.

Sounds like wrong field order. Does it go away when you set deinterlace to bob?

Beta Boy: Here's a Blu Ray sample: http://rapidshare.com/files/108034632/sample.ts.html

Inerestingly the stuttering begins after about 19 seconds of continuous play, even if I start playing at two different places! Another interesting thing is that it happens in both ZP and MPC.

I understand that EVR isn't officially supported under XP but maybe it's an easy fix as it works with v1.5, which uses YUY2 instead of YV12?

And about the field order problem. It's only with HW deinterlacing (not with bob) and my card is an ATI x1950 pro. v1.5 works correctly. Since x264.nl seems to be down, here is a sample file:
http://rapidshare.com/files/108028549/hd.forum.tf1.hd_1080i50.ts.html

gbates

16th April 2008, 22:27

"Download limit reached !" when trying to download trial version...

BlackSun

16th April 2008, 22:29

That's because you excedeed the download limit. Maybe you have used a download accelerator.

gbates

16th April 2008, 22:56

No download accelerators, I'm using Opera and I get this message after first click on download link...

BlackSun

16th April 2008, 23:52

Please pm the download url you got and I'll check if there is a bug.

BlackSun

17th April 2008, 10:03

Thanks for the two reports about the trial, the bug is fixed. You can ask for the trial again and you will get a working link :)

CiNcH

17th April 2008, 16:47

DVBViewer (VMR-Fix off) and BOB-Deinterlacing within CoreAVC mostly results in a framerate of about 40 fps. The renderer however does not skip frames. The DVB Source Filter frequently triggers a graph rebuild due to buffer overflow.

May this be caused by the nature of the DVBViewer graph being a push scenario instead of a pull scenario?