[RESOLVED] DGMVCSource b22 and VC++2012

This is the home of QuickSync (aka Intel Media SDK) stuff.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Many thanks, r0lZ is being to test and we'll tell you if it works !
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Hi Don!

Unfortunately, the last release doesn't work at all, at least on my PC (without the magical Intel hardware).

I did 2 tests, both with "The Flying Swords of Dragon Gate 3D", because that BD has a few frames that cannot be properly decoded with DGMVCSource and FRIMSource and the Intel lib v5.14.4.28 (28/4/2014). The decoding problem is in the AVC stream at 1:23:35. Apparently, an i-frame cannot be decoded properly, and of course, the whole GOP is bad. The MVC frames are bad as well, since they depend of the AVC stream. As far as I know, it's a bug in the Intel library, not in the avisynth plugins. A similar bug is also present in Pacific Rim, as you know, but I don't have that BD3D and I can't test it.

The result of my test with your beta of GDMVCSource and the latest Intel lib (in SW mode) is extremely bad. All MVC frames are full of decoding problems. There is a sort of pixelisation, but also freezes and abrupt jumps. But there is at least one good point: the AVC stream is perfect everywhere, including in the Dragon Gate shot at 1:23:35. It's that fact that makes me think that that bug has been fixed by Intel. But I wonder why the MVC stream cannot be decoded properly.

I did another test, but this time I have replaced the new Intel lib with the old one (v5.14.4.28, 28/4/2014, that has been distributed with your old plugin and with FRIMSource). With that lib, the MVC stream is decoded properly, but of course, the Dragon Gate bug is still present.

The results of my tests are exactly identical than the results of the tests I did with the old version of the plugin and with the new version of the Intel lib. That means that your new plugin works exactly like the old one. Obviously, Intel has changed something that makes the new lib incompatible with the code you have used to load or call the lib. I know that it is possible to use it properly, because Pistacho's MVCSource works fine. It can decode the MVC stream properly, and the Dragon Gate bug doesn't happen. Pistacho has explained at D9 that he has modified his code to make it work with the new lib, but unfortunately, he has not explained exactly what he did. Perhaps you can contact him personally.

I can't help much more, but of course, I can test new betas if you wish. Anyway, thanks again for your plugin.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

OK, thank you for the test results. I'll look into it. But honestly, this is why I dropped DiAVC...they didn't like to bother maintaining backwards compatibility for existing applications. Now here we see the same. Compare it to nVidia's CUVID, where the only problem I ever had was that earlier cards were dropped for support. I shouldn't have to re-engineer my applications when a new library is released!

Is there anything my stuff is doing that pistacho is not? I ask because if there is no value added anymore, maybe there is no point.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Thanks for your interest about this plugin :)

The mainly difference between both plugin is the license : your plugin is free to use and Pistacho's plugin need to install his full application in order to use only his plugin. I think others points are similars...
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Yes, due to its license, Pistacho's plugin cannot be distributed with our applications, and the user needs to install a lot of other DLLs and decoders to use MVCSource. There is a risk of conflicting filters. Your plugin is much better for us. Also, I guess that you want to solve the problem for your more general decoding plugin as well. I understand how the lack of backward compatibility and support can be frustrating for you (and us), but I suppose that it should be possible to find the doc on how to adapt existing programs to the new version. Maybe on the Intel forums? I agree that nVidia does a better job, but afaik there is no free nVidia MVC decoder available.
Anyway, we'll grateful if you can have a look again. Thanks in advance. :-)
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Investigating...
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Everything is fine until the first call to m_pmfxDEC->DecodeFrameAsync(), which returns MFX_ERR_UNKNOWN.

I've posted a query about it at the INTEL forum.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Thanks for your investigation. I hope Intel will help.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Thanks Donald ! I've seen there is a response at the Intel forum, so it's better than last time :)

Perhaps you could find a solution with Intel forum help... I hope ;)

Edit : Not an issue of course but a solution ! Thanks r0lZ.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Oy. After posting the log I waited a few days for a reply but there was none. So I sent a private message to Tony to make sure he saw it and to ask what he thought. There has been no response. :?

So, I'm afraid things appear to be no different from last time.

Even the MPEG-2 decode error I found was never resolved. Finally, the reply was: "Yes, there does seem to be an error in the sample code." But they never followed up with a fix or other solution.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Damn! But the strange thing is that there is already a solution. The problem is to know what it is and how to implement it. Pistacho did it, but as you know, he is sparing with the information. But perhaps you can contact him yourself. Perhaps he will accept to help another programmer.
Anyway, thanks for having tried.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

I'll keep trying all available ways. Sending pistacho a PM now.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Thanks, I hope he'll respond to you !
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Me too, but don't get your hopes up. I'm hoping that Tony will come thru for us.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

OK, things are looking up. Tony posted this driver link:

https://downloadcenter.intel.com/Search ... rodId=3720

Use the 1-12-2015 version. Please also re-download b23 as I updated the build to INDE 2015 Update 1.

After I installed both everything is working for me. Can you all try that and report your results? Thank you.
DAE avatar
Nico83500
Posts: 78
Joined: Thu Jan 02, 2014 9:28 am

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Nico83500 »

Hi Donald,
Thanks to you for this update. I'll try it tomorrow and let you know.
I don't have an Intel CPU but an AMD Phenom II X6 so I don't use Intel driver.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

I've just updated b23 again, so all please re-download. Be sure to get the new libmfxsw32.dll from there too.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Always the same link http://rationalqm.us/dgmvcsource/dgmvcsource100b23.zip?
I'll test it tomorrow. To late here do do it now.
Thanks anyway.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Yes, it's the same link.
DAE avatar
Sharc
Posts: 233
Joined: Thu Sep 23, 2010 1:53 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by Sharc »

I tested the new version with Pacific Rim, SW decoding. The glitch has gone. (I can't test HW decoding with my PC).
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

I did also a quick test, and encoded only a few minutes of the beginning of Pacific Rim (SW decoding). No problem so far. :) I am now encoding it up to the "famous" glitch at 1:06:14, but I'm confident it will work fine. Then, I'll do another test with Flying Swords of Dragon Gate to verify the glitch at 1:23:35, but again I'm pretty sure it will work fine. I hope some peoples will test it with HW acceleration.

However, there have been reports that Pistacho's MVCSource with the latest DLLs from Intel work usually fine but fails on some specific machines (notably a Dell computer, with or without HW decoding, I don't know). That must be tested too, to be sure. But if it passes that test, your DGMVCSource will be the first fully working MVC decoder based on the Intel lib. (In addition to the problem with the Dell, we have also found 2 terrible bugs in Pistacho's MVCSource, not directly related to the Intel lib, that makes it totally unusable in practice.)
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

Test of Pacific Rim finished. No glitches any more. Perfect with my Intel Core 2 Quad Q8200 CPU. :)
I will now do the test of Dragon Gate and I'll report the result in 3 hours or so...
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Good news, guys. Thank you for your testing.
r0lZ wrote:we have also found 2 terrible bugs in Pistacho's MVCSource, not directly related to the Intel lib, that makes it totally unusable in practice.)
I'm curious to know what those bugs are.
DAE avatar
r0lZ
Posts: 30
Joined: Fri Feb 06, 2015 1:36 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by r0lZ »

admin wrote:I'm curious to know what those bugs are.
They have been discovered by frank and confirmed by me.

1. Its avisynth plugin manipulates the PWD of avisynth (probably so that it can find the Intel lib even if it is not in the Windows PATH). Therefore, some avisynth functions do not work as expected. For example, BD3D2MK3D generates SupTitle() commands when the user wants to hardcode subtitles on the video. The subtitle is in the same directory than the AVS script and therefore its file name is specified without its path in the SupTitle() command. That allows the user to move the entire project to another disc if he thinks that it may require more disc space. That works pretty well with your DGMVCSource and with FRIMSource, but not with Pistacho's MVCSource: SupTitle() issues a warning telling that it cannot find the subtitle file. Obviously, the current directory is not where it should be when MVCSource is used.

2. When a 3DBD with inverted left and right views is converted, it is necessary to change the "mode" parameter of MVCSource to specify to invert the two views (to build a combined SBS or T&B stereo file with the views in the common order: left view first). That works well for the first 45 minutes, but after that, the views are abruptly changed, and you end up with a final movie with a mix of left and right view first! Aaargh!

And I consider also as a (sort of) bug the fact that the DLL cannot be used independently of his BDtoAVCHD program. It installs a lot of things, including DirectShow codecs, and it makes hundreds of changes in the registry. I don't think it's a good thing. But that's another story.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: [RESOLVED] DGMVCSource b22 and VC++2012

Post by admin »

Interesting. Thanks for the explanation. They sound like bad bugs. Doesn't pistacho want to fix them?

Isn't FRIMSource() OK now too, in which case that would be the first. ;)

Lastly, do y'all use Avisynth 2.5.8 or does it work with later versions? In my help file I specify 2.5.8 but I can't remember why I have that there. Update: I just tested with AVS+ and things seem fine.
Post Reply