madshi bug tracker

View Issue Details Jump to Notes ] Issue History ] Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0000284madVRbugpublic2015-04-26 01:302018-01-21 21:17
ReporterGCRaistlin 
Assigned Tomadshi 
PrioritynormalSeverityblockReproducibilityalways
StatusclosedResolutionunable to reproduce 
PlatformC2D E6420OSWinXPOS VersionSP3
Summary0000284: Going full screen and then back on the 2nd monitor causes 100% CPU usage
DescriptionSee below.
Steps To Reproduce1. Download and open the test DVD: http://rusfolder.com/43539310 [^]
2. Press Pause, drag the MPC-HC window to the monitor 0000002.
3. Go full screen, then go back.
MPC-HC starts to eat all CPU resources. In fact, you'll hardly be able to close MPC-HC not through the console (I used Far for this).
Additional InformationUnder "2nd monitor" I mean literally second, not only "not primary". The issue can't be reproduced on the monitor 0000001 even if the primary monitor is the monitor 0000002.
TagsNo tags attached.
madVR Version0.87.17 and later
Media Player (with version info)MPC-HC 1.7.8.162
Splitter (with version info)LAV (internal)
Decoder (with version info)LAV (internal)
DecodingCUDA
Deinterlacingauto mode
DXVA2 Scaling Activeno
Aero / Desktop CompositionOff
Problem occurs with modewindowed mode
GPU ManufacturerNVidia
GPU ModelGF 8800 GTX
GPU Driver Version340.52
Attached Fileszip file icon madVR - freeze report (1).txt.zip [^] (7,419 bytes) 2015-05-02 15:47
zip file icon madVR - freeze report (2).txt.zip [^] (8,822 bytes) 2015-05-02 16:18
zip file icon madVR - freeze report (3).txt.zip [^] (7,465 bytes) 2015-05-02 18:35
zip file icon madVR - freeze report (4).txt.zip [^] (7,312 bytes) 2015-05-02 19:22

- Relationships
related to 0000285closedmadshi Changing refresh rate causes 100% CPU usage 

-  Notes
(0000979)
madshi (administrator)
2015-05-02 11:40

Can't reproduce that problem here with v0.87.21. Can you double check with v0.87.21? If the problem still occurs for you, please in the situation where MPC-HC seems to be frozen, please Ctrl+Alt+Shift+Break/Pause. With a bit of luck you may find a freeze report file on your desktop after a couple of seconds. If so, please zip it and upload it here (the bug tracker only accepts zip attachments).
(0000981)
GCRaistlin (reporter)
2015-05-02 15:47

I've discovered the issue with v0.87.21 exactly, then I've just decided to determine in which version it was introduced. So I've reproduced it many times.
(0000983)
madshi (administrator)
2015-05-02 15:56

Hmmmm... Unfortunately you didn't have the MPC-HC debug symbols (*.PDB) installed when you made that freeze report. Sorry, should have mentioned that. Can you please get the MPC-HC and LAV debug symbols? They need to match the exact MPC-HC/LAV version you have installed, otherwise the results will be useless. Maybe just update to the latest LAV/MPC-HC builds, to be safe, and get the latest matching PDB symbols, as well. Then please create a new freeze report. The freeze report will then contain information about what MPC-HC and LAV were doing at the time of the freeze, too.

First impression from your freeze report is that the MPC-HC main thread isn't responding to messages at the moment of the freeze. But due to the lacking MPC-HC PDB symbols I can't see where the MPC-HC main thread got stuck.
(0000986)
GCRaistlin (reporter)
2015-05-02 16:19

Done
(0000987)
madshi (administrator)
2015-05-02 17:16

Ok, both freeze reports look pretty much identical, as I suspected. So I guess we can forget about 285 for now. If we find a fix for 284, it will likely also fix 285.

So, I've noticed 2 things in the freeze report:

1) In both cases it seems a key was pressed right before the freeze occurred. Is that true? Can you try to reproduce the problem by using a different means of input, e.g. mouse? Does the same problem still occur?

2) The problem seems to have to do with DXVA processing (decoding, scaling and/or deinterlacing). What happens if you disable all that stuff, does that make the problem go away? If so, can you please try to identify which of the three (decoding, deint, or scaling) is causing the issue?
(0000988)
GCRaistlin (reporter)
2015-05-02 18:35

> In both cases it seems a key was pressed right before the freeze occurred. Is that true? Can you try to reproduce the problem by using a different means of input, e.g. mouse?
Yes, I've went fullscreen and back by Alt-Enter.

>The problem seems to have to do with DXVA processing
In fact, I've reset both MPC-HC and madVR settings to defaults. I don't see any DXVA options enabled.
(0000989)
madshi (administrator)
2015-05-02 18:44

Can you try double clicking instead of Alt-Enter?

Please check Ctrl+J, probably deinterlacing is on? Try turning it off. Does that change anything?
(0000990)
GCRaistlin (reporter)
2015-05-02 19:22

>Can you try double clicking instead of Alt-Enter?
See madVR - freeze report (3).txt.zip

>Please check Ctrl+J, probably deinterlacing is on? Try turning it off. Does that change anything?
Yes, deinterlacing was on (auto). With deinterlacing disabled, the issue isn't reproducible. In this case, freezing occurs when switching deinterlacing by hot key. See madVR - freeze report (4).txt.zip
(0000991)
madshi (administrator)
2015-05-02 19:33

Hmmmm... Normally those freeze reports are really useful, but in your case, although I can see what all the threads are doing, I don't fully see the reason for the freeze. It's not my impression that madVR seems to be at fault. It seems to me that the MPC-HC thread which created the madVR instance fails to react to messages for some reason.

Hmmmm... That makes me wonder: In the MPC-HC settings under "Playback", do you have the option "use work thread to contruct the filter graph" enabled? If not, try that. Maybe restoring MPC-HC to the default settings might also be worth a try.
(0000993)
GCRaistlin (reporter)
2015-05-03 20:46

> It seems to me that the MPC-HC thread which created the madVR instance fails to react to messages for some reason.
Note that madVR 0.87.14 works just fine under the same conditions.

> In the MPC-HC settings under "Playback", do you have the option "use work thread to contruct the filter graph" enabled?
Yes.

>Maybe restoring MPC-HC to the default settings might also be worth a try.
As I wrote above, I use MPC-HC default settings when reproducing the issue.
(0000995)
madshi (administrator)
2015-05-03 21:04

Can you please try build v0.87.15 and v0.87.16, so I know which build exactly introduced the problem for you?
(0000996)
GCRaistlin (reporter)
2015-05-03 23:54

The issue is reproducible with 0.87.15 and later. To be precise, for the first time with 0.87.16 I couldn't reproduce it by going full screen and back on the second monitor. Then I dropped MPC-HC window back to the first monitor and launched Display applet to check bug 285 by changing refresh rate. Here (before changing refresh rate) I got the freeze.
The second and all the following tries of 0.87.15 gave the freeze "in a general way" - by going full screen and back on the second monitor.
(0001019)
GCRaistlin (reporter)
2015-05-11 16:10

The issue still exists in 0.88.2.
(0001021)
GCRaistlin (reporter)
2015-05-11 23:08

The issue still exists in 0.88.4.
(0002096)
madshi (administrator)
2018-01-14 19:59

Same comment as in 0000285:

2.5 years later. Do you still have this issue? I was never able to reproduce it, so I wonder... FWIW, the freeze report seems to suggest that DXVA itself gets stuck for some reason, but it's not clear to me what this has to do with the refresh rate change. Also, there have been so many changes in all the software, and the driver and OS, so it might be worth to double check if the problem even still exists, anymore?
(0002165)
madshi (administrator)
2018-01-21 21:17

Closing this due to lack of feedback/reply. Please feel free to reopen at any time.

- Issue History
Date Modified Username Field Change
2015-04-26 01:30 GCRaistlin New Issue
2015-05-02 11:40 madshi Note Added: 0000979
2015-05-02 11:40 madshi Assigned To => madshi
2015-05-02 11:40 madshi Status new => feedback
2015-05-02 15:47 GCRaistlin Note Added: 0000981
2015-05-02 15:47 GCRaistlin Status feedback => assigned
2015-05-02 15:47 GCRaistlin File Added: madVR - freeze report (1).txt.zip
2015-05-02 15:56 madshi Note Added: 0000983
2015-05-02 15:56 madshi Status assigned => feedback
2015-05-02 16:18 GCRaistlin File Added: madVR - freeze report (2).txt.zip
2015-05-02 16:19 GCRaistlin Note Added: 0000986
2015-05-02 16:19 GCRaistlin Status feedback => assigned
2015-05-02 17:16 madshi Note Added: 0000987
2015-05-02 17:16 madshi Status assigned => feedback
2015-05-02 18:35 GCRaistlin Note Added: 0000988
2015-05-02 18:35 GCRaistlin Status feedback => assigned
2015-05-02 18:35 GCRaistlin File Added: madVR - freeze report (3).txt.zip
2015-05-02 18:44 madshi Note Added: 0000989
2015-05-02 18:44 madshi Status assigned => feedback
2015-05-02 19:22 GCRaistlin Note Added: 0000990
2015-05-02 19:22 GCRaistlin Status feedback => assigned
2015-05-02 19:22 GCRaistlin File Added: madVR - freeze report (4).txt.zip
2015-05-02 19:33 madshi Note Added: 0000991
2015-05-02 19:33 madshi Status assigned => feedback
2015-05-03 20:46 GCRaistlin Note Added: 0000993
2015-05-03 20:46 GCRaistlin Status feedback => assigned
2015-05-03 21:04 madshi Note Added: 0000995
2015-05-03 21:04 madshi Status assigned => feedback
2015-05-03 23:54 GCRaistlin Note Added: 0000996
2015-05-03 23:54 GCRaistlin Status feedback => assigned
2015-05-11 16:10 GCRaistlin Note Added: 0001019
2015-05-11 23:08 GCRaistlin Note Added: 0001021
2018-01-14 19:58 madshi Relationship added related to 0000285
2018-01-14 19:59 madshi Note Added: 0002096
2018-01-14 19:59 madshi Status assigned => feedback
2018-01-21 21:17 madshi Note Added: 0002165
2018-01-21 21:17 madshi Status feedback => closed
2018-01-21 21:17 madshi Resolution open => unable to reproduce


Copyright © 2000 - 2012 MantisBT Group
Powered by Mantis Bugtracker