Media Player Classic Home Cinema (MPC-HC) - DXVA!

[Spec-Chum]

All of this sounds like horrible development methodology. :eek:

PS Note that Beliyaal has posted in this thread since three or four of us asked for a post of the instructions for the vsync changes... and still nothing...

Whoa, steady on boy! Last time I checked this was an open source project, coded by very talented VOLUNTEERS, who give up THEIR FREE TIME so that YOU can use the player for FREE!

I'm not going to apologise for sounding harsh, I've got the greatest respect for all the devs and I think you should have too.

Spec

[krbo]

Of course there is and must be a respect for developers, this is enormous and fantastic work and they gave many great hours spent watching beautiful HD videos.

This thread serves mostly as a bug report for them, something they can fix, something can't but it's our obligation to report anything strange.

Like, that I'm now in a situation that there are no more MPC HC for me after v. 1043 so I will stick with it.

Maybe future SVN or Catalyst version would remove my problem, who knows

(oh, btw there is an intersting situation that 1043, last one before beliyaal has 3 times more "Просмотров" ("views" I guess as I'm from Slavic language group) then other versions at xvideo.ru)

And hardest work for any software developer is to write a manual or help file. I know that - been there (databases were my field) so I sympathize with Beliyaal :)

great, great work. Thank you devs !

[krbo]

Where did you get the "changes" in the SVN cause your 1050 it's different from what I see in the SVN http://mpc-hc.svn.sourceforge.net/viewvc/mpc-hc/?view=log :confused:

www.xvidvideo.ru , always downloaded from there as it's not my playgorund to compile such things by my own :)

edit: sorry for the typo in link, it was too late (1am) and I went to bed immediately , thanks for correction !

[cwl7454]

krbo, kindly remove your link, not fitting for this forum or can a moderator remove please

[flanger216]

I think he meant www.xvidvideo.ru :rolleyes: ... but yes, a quick edit by the poster or a moderator might avert tons of accidental NSFW disasters :D...

[alexins]

Media Player Classic HomeCinema (x86/x64), svn 1065 (http://www.xvidvideo.ru/content/view/726/1/)

Supported languages: BY, CN, CZ, DE, EN, ES, FR, HU, IT, KR, PL, RU, SK, TR, UA

Changes log (http://www.xvidvideo.ru/content/view/7/17/)

(1061-1065)
Change : increased the limit ref_num up to 16 for Nvidia;
Italian resources updated; translated new "TODO" lines;
Updated Hungarian translation;
Updated Russian translation;
Updated Turkish translation.

[Snowknight26]

r1065 crashes for me when DXVA tries to kick in when I play interlaced H.264 videos. MPC-HC x64, Vista, on a Radeon HD 4850, Catalyst 9.4. r1060 does not.

It seems like there are a lot of outstanding issues that aren't being taken care of or that new issues ones that shouldn't crop up, are. If any of the devs needs a testing environment somehow, be it through remote desktop or what have you, feel free to let me know - always willing to help the cause.

[marc99]

"Change : increased the limit ref_num up to 16 for Nvidia"

Why? Internal decoder doesnt't fix.

[Aleksoid1978]

"Change : increased the limit ref_num up to 16 for Nvidia"

Why? Internal decoder doesnt't fix.

I specialy test some .mkv(about 5) 1920, 1280 with level 5.1 and ref_num = 16(made in megui short sample) and all decoding with DXVA fine.

if you have file, that not working with DXVA - make short sample and upload for test.

[marc99]

Sample: http://www.sendspace.com/file/fjci8t

See this: http://forum.doom9.org/showthread.php?p=1262947#post1262947

[Mercury_22]

Sounds like a good idea to me.

----------
The recent changes should have fixed many bugs, and also introduced some new ones.

In order to get a clear and accurate view i invite everyone to verify that bugs can or cannot be reproduced or might even be fixed. (multiple verifications with different hardware/files are very welcome)

Also any new problems need to be reported in the bug tracker.

When reporting or confirming bugs please provide the following data:

-Windows and servicepack versions
-MPC-HC build version
-If appropriate FFdshow build number
-Videocard and driver version
-Renderer used and options selected in the output settings
-A sample so we can test it on our systems


We are unable to reproduce many of the bugs currently reported, from my limited testing i could close over half of them.

We need detailed information on how to reproduce them

I have made batches, following Mercury's advice i would like to ask everyone to start testing batch 1:

https://sourceforge.net/tracker/?func=browse&group_id=170561&atid=854651&status=1&artgroup=898922

Tested batch 1 except "1833718 Eastern Europe font in Subtitles problem with G15 " and I CAN'T reproduce any of the bugs so NO confirmation for batch 1 with MPC-HC svn 1065 (and FFDShow svn 2896)

Edit: also except "1850145 MPC not reloading Elemend title from multi-segment file"

[tetsuo55]

@ Everyone:

The new code has fixed (a lot) of bugs/problems/feature request.

Unfortunately it has also revealed some bugs in driver code and caused a few regressions.
The dev's are spending all the time available now to work on the regressions and fix bugs.

We do need to provide them with detailed information including samples, this is why i have asked everyone to try and reproduce the bugs on the buglist on sourceforge and tell is if you could or could with detailed hardware/software information.

Also new bugs and regressions need to be added.

I have already closed a few bugs that are definately fixed, and i have many more on my list awaiting confirmation.

Once the buglist is clean we can focus all attention on the remaining bugs.

[tetsuo55]

Tested batch 1 except "1833718 Eastern Europe font in Subtitles problem with G15 " and I CAN'T reproduce any of the bugs so NO confirmation for batch 1 with MPC-HC svn 1065 (and FFDShow svn 2896)

Could you post your testing hardware/software and used samples on as a reply to the bugs?

[Mercury_22]

Could you post your testing hardware/software and used samples on as a reply to the bugs?

See my signature ! as for the samples is it really necesary ? (tested with various / multiple files)

[tetsuo55]

See my signature ! as for the samples is it really necesary ? (tested with various / multiple files)

A sample could be interesting for reports like: files with this "special feature" do not work, and you're sample with the same "special feature" does.

Information in the thread tends to get lost (just check out the number of posts), thats why we need confirmation in the bug list. it won't get lost there

[hoborg]

Hi.
Nobody care about problem i reported?
http://forum.doom9.org/showthread.php?p=1276859#post1276859

It is no longer possible to use standalone MPC-HC decoder becouse of broken SW h.264 decoding.

[tetsuo55]

Hi.
Nobody care about problem i reported?
http://forum.doom9.org/showthread.php?p=1276859#post1276859

It is no longer possible to use standalone MPC-HC decoder becouse of broken SW h.264 decoding.

Probably related to the change in 1053, casimir is working on it.

[hoborg]

Probably related to the change in 1053, casimir is working on it.

Thanks for info.

[Mercury_22]

A sample could be interesting for reports like: files with this "special feature" do not work, and you're sample with the same "special feature" does.

Information in the thread tends to get lost (just check out the number of posts), thats why we need confirmation in the bug list. it won't get lost there

Ok I'll try ! but no go on samples since they don't bother to post one ! (I'm trying to concentrate on testing not cutting & uploading samples if that's not relly necessary)
Also i'm thinking it will be better to ask / post in the bug page a request to the initial poster to retest with latest MPC for bugs reports OLDER THEN November 29, 2008 = latest "official MPC-HC" And if no respond in let's say 1 week you can close the bug report
Cause many of this bugs seems abandoned & fixed to me ! just that the initial poster never took time to report solved !
This way we can concentrate on real bugs !

P.S. I'll take the time to post samples for newer bugs (if necessary) Those old bugs I'm almost sure they are 99% fixed and not reported as fixed so I don't want to lose more time with them than a quick test !

[Phaser]

I noticed a regression since the Beliyaal merge, that's preventing me from upgrading to latest version. When watching a video fullscreen (1680x1050), random lines appear if I try to zoom in. This happens only when I use the Bicubic resizer with VMR9 renderless and EVR custom (I tested on both XP and Vista), with every video format SD & HD.

Here's an example:

Before
http://imgur.com/1v90s.png (http://imgur.com/1v90.jpg)

After
http://imgur.com/1vNKs.png (http://imgur.com/1vNK.jpg)


This is my configuration:
- Windows XP SP3 Pro
- works with svn 1043 / broken since svn 1048
- Ati Radeon HD 3650 / Catalyst 9.4
- Renderer: VMR 9 renderless / other output options: 3D surfaces, VMR9 mixer mode, lock back-buffer

[tetsuo55]

Yeah i agree, but i still need multiple verfications of not being able to reproduce.

The first batch has a few bugs that might be feature request or not be bugs at all.
1850145 - MPC not reloading Elemend title from multi-segment file

Does anyone have a sample of such a file? does it not work with both internal and haali splitter? Does it work in other players like vlc and mplayer?
1805361 - Unable to open video file with MPC while it is being saved

I think this depends on the "downloading" application and iirc this happens in the older version too. Can other players open the file when MPC(-HC) can not?

[tetsuo55]

I noticed a regression since the Beliyaal merge, that's preventing me from upgrading to latest version. When watching a video fullscreen (1680x1050), random lines appear if I try to zoom in. This happens only when I use the Bicubic resizer with VMR9 renderless and EVR custom (I tested on both XP and Vista), with every video format SD & HD.

Here's an example:

Before
http://imgur.com/1v90s.png (http://imgur.com/1v90.jpg)

After
http://imgur.com/1vNKs.png (http://imgur.com/1vNK.jpg)


This is my configuration:
- Windows XP SP3 Pro
- works with svn 1043 / broken since svn 1048
- Ati Radeon HD 3650 / Catalyst 9.4
- Renderer: VMR 9 renderless / other output options: 3D surfaces, VMR9 mixer mode, lock back-buffer

This is a known bug, i assume you are using DXVA?

Please update the bug report here:
https://sourceforge.net/tracker/index.php?func=detail&aid=2018012&group_id=170561&atid=854651

[Mercury_22]

Yeah i agree, but i still need multiple verfications of not being able to reproduce.

The first batch has a few bugs that might be feature request or not be bugs at all.
1850145 - MPC not reloading Elemend title from multi-segment file

Does anyone have a sample of such a file? does it not work with both internal and haali splitter? Does it work in other players like vlc and mplayer?
1805361 - Unable to open video file with MPC while it is being saved

I think this depends on the "downloading" application and iirc this happens in the older version too. Can other players open the file when MPC(-HC) can not?

I think that bug repors wich are NOT providing your requested data When reporting or confirming bugs please provide the following data:

-Windows and servicepack versions
-MPC-HC build version
-If appropriate FFdshow build number
-Videocard and driver version
-Renderer used and options selected in the output settings
-A sample so we can test it on our systems

should be IGNORED / closed especially when that data (samples) it's really needed and not provided !

[tetsuo55]

There appear to be some serious compatibility problems with the G15 manager.

It appears to break some subtitles and some renderers.

So far reports are in that eastern european fonts get rendered incorrectly and that selecting VMR9 renderless causes artifacting.

Anyone here have a G15 and the manager, i think its a bug with the manager itself but it's worth having a look. (it probably uses vmr9 renderless to print to the keyboard LCD)

[tetsuo55]

I think that bug repors wich are NOT providing your requested data
should be IGNORED / closed especially when that data (samples) it's really needed and not provided !

The instructions for bug reporting have to be re-written to be more clear.
We should give everyone a chance of # days to update their report with the required information(and give others the chance to reproduce with the information available already)

I will write new and more detailed bug reporting instructions (these should be added to the wiki)

I will add information on creating samples and screenshots and where to host them

[Phaser]

This is a known bug, i assume you are using DXVA?

Please update the bug report here:
https://sourceforge.net/tracker/index.php?func=detail&aid=2018012&group_id=170561&atid=854651


I don't think it's related, in my example it was without DXVA (AVI XviD file). And the problem it's not the quality, but the "grid lines" that appear over the video.

Also I noticed this on the Beliyaal thread. Maybe he knows more about what's causing this:

Known Bugs
* Bicubic resizer sometimes renders incorrectly.

[Aleksoid1978]

Sample: http://www.sendspace.com/file/fjci8t

See this: http://forum.doom9.org/showthread.php?p=1262947#post1262947

Yea, at this sample playback freeze. I think ref_num=16 work normaly for 1920x and 1280x files.

Need some more testing ...

[marc99]

Yea, at this sample playback freeze. I think ref_num=16 work normaly for 1920x and 1280x files.

Need some more testing ...

If file has static picture, max number of bframes is use and freez.

[tetsuo55]

If file has static picture, max number of bframes is use and freez.

Appears on the surface to be a bug in the nvidia decoder, does this sample work with CUDA/Coreavc ?

[marc99]

Appears on the surface to be a bug in the nvidia decoder, does this sample work with CUDA/Coreavc ?

No (macroblocking, not freez).

[tetsuo55]

No (macroblocking, not freez).

Though so. Then it's probably a combination of mpc decoder/nvidia decoder

-----------

I rewrote the bug reporting rules, does anyone have feedback/corrections/additions?

How to report bugs

1. Provide as much system information as possible:

-Windows and servicepack versions

-MPC-HC build version, when reporting regressions please name the last working build, also comparisons to other players (even MPC regular) are very welcome

-Try with both internal and haali splitter and report the difference in results

-Try with both internal and external codecs and report the difference in results

-If appropriate FFdshow build number

-Videocard and driver version

-Renderer used and options selected in the output settings (a screenshot of the output menu, rightmouseclick filters and rightmousclick renderer settings). Also try with different settings/renderers.

2. Provide a Sample and screenshots

-A sample can be cut ....

-A screenshot can be made ....

Please upload screenshots to http://imageshack.us/

Please upload samples to http://sharebee.com and post all the mirrors

3. Write a detailed explanation of how to reproduce the problem

[Mixer73]

Guys

I'm using the 1060 build in 32bit and I notice with EVR that I can no longer zoom up the picture either with the number keys in the keypad, or with the "Pan/Scan" menu options. Can somebody else test this?

in EVR CP it works but EVR CP is unusably slow on my system. It is less than half speed but if you look at the stats it claims it is not dropping any frames.

EDIT: I increased the buffers to 10 and I have playback now but if I select Vsync I get the "playing frames as fast as you can decode" problem others have reported.

Vista x64, SP1
MPC-HC build 1060
Codecs don't seem to matter, tested in this case with MPEG2 Microsoft and internal decoder

EDIT 2:
Above testing was with MPEG2 footage, both SD & HD, with a HD AVI file in XVID EVR CP is again totally unusable. Its my understanding that all Beliyaal's work is around the EVR CP functionality.

I got this working better by using D3D fullscreen and unticking every box under Renderer Settings but as soon as you pause and exit fullscreen MPC hangs:

Problem Event Name: AppHangB1
Application Name: mplayerc.exe
Application Version: 1.2.1060.0
Application Timestamp: 49ee2ae8
Hang Signature: fe19
Hang Type: 2048
OS Version: 6.0.6001.2.1.0.256.1
Locale ID: 3081
Additional Hang Signature 1: 7335330df957fa0be2a823efc5d2b9e8
Additional Hang Signature 2: a2c2
Additional Hang Signature 3: ad53d77355753997dcdb8b97c3c2208e
Additional Hang Signature 4: fe19
Additional Hang Signature 5: 7335330df957fa0be2a823efc5d2b9e8
Additional Hang Signature 6: a2c2
Additional Hang Signature 7: ad53d77355753997dcdb8b97c3c2208e

I'm pretty sure others have got this same symptom so I think this does formally need to be entered as a bug

[gngn]

Yea, at this sample playback freeze. I think ref_num=16 work normaly for 1920x and 1280x files.

Need some more testing ...

on XP the old ReFrames =< 11 rule is still valid, even with the latest nvidia WHQL drivers 182.50 , so the change "increased the limit ref_num up to 16 for Nvidia" should'nt be implemented for XP users at least.

there is another problem since version 1053: "Fixed : MpcVidecDec output MediaType incorrectly filled to 1088 instead of 1080 for software decoding of 1080 material" , true 1080p material is shown as 1088 pixels high, as Snowknight26 reported here (http://forum.doom9.org/showthread.php?p=1275945#post1275945) . sideeffect is that on 1080p videos the first 2-3 seconds have horrible macroblocking with DXVA on:
http://i266.photobucket.com/albums/ii259/gngnmumu/dxva/th_1088p.png (http://i266.photobucket.com/albums/ii259/gngnmumu/dxva/1088p.png)

here (http://www.mediafire.com/download.php?nfzjnzzdt3y) is a sample file

[squid_80]

Appears on the surface to be a bug in the nvidia decoder, does this sample work with CUDA/Coreavc ?
The sample uses 16 reference frames. The current version of CoreAVC will not use CUDA decoding for 16 reference frames (<16 is fine), since the nvidia decoder can only output decoded frames by storing them in the DPB.

[tetsuo55]

The sample uses 16 reference frames. The current version of CoreAVC will not use CUDA decoding for 16 reference frames (<16 is fine), since the nvidia decoder can only output decoded frames by storing them in the DPB.

Every hardware decoder is limited to DPB, if it works with DXVA i think CUDA should too? (in other words, MPC also stores those 16 ref frames in DPB)
But the DPB needs to have room for ref frames + 1 if i'm not mistaken (that would mean that the nvidia DPB can store 17 ref frames)

[squid_80]

But the DPB needs to have room for ref frames + 1 if i'm not mistaken (that would mean that the nvidia DPB can store 17 ref frames)
That's right, and due to a windows limitation it's impossible to setup a DPB with more than 16 frames.

[tetsuo55]

That's right, and due to a windows limitation it's impossible to setup a DPB with more than 16 frames.

Then why does it work with MPC?? (excluding the sample posted above)

@Aleksoid1978, can you post the working sample's you made?

[squid_80]

Then why does it work with MPC?? (excluding the sample posted above)
Exactly my point: it seems it doesn't.

[73ChargerFan]

Those old bugs I'm almost sure they are 99% fixed and not reported as fixed so I don't want to lose more time with them than a quick test !
Have you tested on XP? Vista+EVR is different from XP+VMR9. Also the Beliyaal merge needs to be regression tested. Bugs should be tested on popular configurations.

DTS 1536 over spdif & the internal matroska splitter are still buggy, and have been known to be for years, going back to Gabest. Some code doesn't have an owner anymore, and no one to fix certain bugs.

[73ChargerFan]

I rewrote the bug reporting rules, does anyone have feedback/corrections/additions?
Please add links to the Changes Log (http://mpc-hc.svn.sourceforge.net/viewvc/mpc-hc/?view=log) and to the Archive of svn builds (http://www.xvidvideo.ru/content/category/1/1/2/).

If using ffdshow, then also test with internal transform filter if appropriate.

If the bug causes mpc-hc to crash, attach a minidump.

MPC-HC settings stored in the registry may be incompatible with earlier versions. The settings can be deleted in Windows XP: save the following code as "clean-registry.reg" and run it.
Windows Registry Editor Version 5.00

[-HKEY_CURRENT_USER\Software\Gabest]

[tetsuo55]

Thanks and updated, I would also like some help on the plain text formatting, it looks like crap now :(

----------------------
How to report bugs

Step One: Make sure you have followed the following steps to make sure it's not a configuration/Version problem:

**First download the newest build here:http://www.xvidvideo.ru/content/category/1/1/2/
(You can also find older builds here for regression testing)
**You can check out the changelog here, the problem might have been fixed in a newer version than the one you where using:http://mpc-hc.svn.sourceforge.net/vi...c-hc/?view=log
**MPC-HC settings stored in the registry may be incompatible with earlier versions. The settings can be deleted: save the following code as "clean-registry.reg" and run it(with admin rights in vista/win7).

Windows Registry Editor Version 5.00

[-HKEY_CURRENT_USER\Software\Gabest]The same goes for the settings.ini file created next to the mplayerc.exe, delete it.
(In both cases MPC will recreate the settings entries on the first run.)
**Make sure you're videocard drivers and directX version are up to date

If this does not solve the problem move on to step two

Step Two: Provide as much system information as possible:

**Windows and servicepack versions
**MPC-HC build version, when reporting regressions please name the last working build, also comparisons to other players (even MPC regular) are very welcome
**Try with both internal and haali splitter and report the difference in results
**Try with both internal and external codecs and report the difference in results
**If appropriate FFdshow build number, If using ffdshow, then also test with internal transform filter if appropriate.
**Videocard and driver version
**Renderer used and options selected in the output settings (a screenshot of the output menu, rightmouseclick filters and rightmousclick renderer settings). Also try with different settings/renderers.

Step Three: Provide sample(s) and screenshot(s)

**A sample can be cut with avidemux(http://fixounet.free.fr/avidemux/) or mkvmerge(http://www.bunkus.org/videotools/mkvtoolnix/)
**A screenshot can be made by pressing the "print screen" button on the keyboard and opening paint, then paste (CTRL+V)

Please upload screenshots to http://imageshack.us/
Please upload samples to http://sharebee.com and post all the mirrors

**If the bug causes mpc-hc to crash, attach a minidump or upload it it to http://sharebee.com and post all the mirrors. You can make a mini dump by starting mplayerc.exe in a command line(with administrator rights) with the option "mplayerc.exe /minidump" and upload the ".dmp" file generated after crash.

Step Four: Write a detailed explanation of how to reproduce the problem

[Steveo08]

i have a question to dxva and evr custom and using colorspace on geforce 8600 gt hdmi.

in the nvidia control panel under digital color output for my plasma tv i am using rgb.

but now i have seen that dxva and evr custom used nv12 instead of rgb.

so should i now chage the color output in the nvidia control panel to YCbCr444?

[mark0077]

i have a question to dxva and evr custom and using colorspace on geforce 8600 gt hdmi.

in the nvidia control panel under digital color output for my plasma tv i am using rgb.

but now i have seen that dxva and evr custom used nv12 instead of rgb.

so should i now chage the color output in the nvidia control panel to YCbCr444?

Correct me if I am wrong, but the signal will be converted to RGB in any case. If you output YCbCr444 from your gfx card, your TV will do the conversion. If you output in RGB from your gfx card, your gfx card will convert the nv12 to RGB.

My preference is still to output RGB from ffdshow, eliminating the need for the gfx card or tv to do any conversions. Experience has thought me not to trust either gfx card or tv for good conversions in all scenarios.

[Aleksoid1978]

@Aleksoid1978, can you post the working sample's you made?

Here:

http://aleksoid.tosei.ru/Samples/test_2.mkv
http://aleksoid.tosei.ru/Samples/test_1.mkv

[ADude]

Well if you rather I don't post at all until I have time to sit down and write documentation I can do that.

We are not looking - in this thread - for carefully written documentation - just an explanation of how to use the new settings for vsync.

Some guy started to ask for a formal manual, but that was not the origin of the request.

[marc99]

Here:

http://aleksoid.tosei.ru/Samples/test_2.mkv
http://aleksoid.tosei.ru/Samples/test_1.mkv

Samples has bframes=0.

Limitation with Vista (64bit, NVidia 8600GT) is:

Reference Frames - Max bFrames (bPyramid Disabled)
16 - 1
15 - 3
14 - 7
13 - 16
Reference Frames - Max bFrames (bPyramid Enabled)
16 - 4
15 - 7
14 - 16

[Aleksoid1978]

Samples has bframes=0.

Limitation with Vista (64bit, NVidia 8600GT) is:

Reference Frames - Max bFrames (bPyramid Disabled)
16 - 1
15 - 3
14 - 7
13 - 16
Reference Frames - Max bFrames (bPyramid Enabled)
16 - 4
15 - 7
14 - 16

How check bFrames count in video ???

[Steveo08]

Correct me if I am wrong, but the signal will be converted to RGB in any case. If you output YCbCr444 from your gfx card, your TV will do the conversion. If you output in RGB from your gfx card, your gfx card will convert the nv12 to RGB.

My preference is still to output RGB from ffdshow, eliminating the need for the gfx card or tv to do any conversions. Experience has thought me not to trust either gfx card or tv for good conversions in all scenarios.



but what is, it i don't use ffdshow, how can i prefer rgb with evr?

[tetsuo55]

Samples has bframes=0.

Limitation with Vista (64bit, NVidia 8600GT) is:

Reference Frames - Max bFrames (bPyramid Disabled)
16 - 1
15 - 3
14 - 7
13 - 16
Reference Frames - Max bFrames (bPyramid Enabled)
16 - 4
15 - 7
14 - 16

Hi marc99, everyone technical person i ask always says there is no limit for bframes?

How did you get these values?

From this list it appears like every 4 b frames count as 1 ref frame(while the specs say any value between 1 and 16 should only be 1 ref frame)

[Aleksoid1978]

I make some test clips, 1920, ref_num=16 with bFrames=9 :

1920x1080 - freez after 3-5 second playback, but 1920x800 - play fine.

here sample 1920x800, that fime play with DXVA - http://aleksoid.tosei.ru/Samples/test_4.mkv

The conclusion - this is not the case, well, or not only that