[WiP] madVR support in MP1 (3 Viewers)

[kenwonders]

It's ok Seb I saw you'd pushed 85. I have replicated the black screen red text error with dshowhelper pdb. Millions of nvd3dum worker threads, the only thing I coming back from 'dshowhelper' call stack search is this:

Not Flagged 3256 0 Worker Thread msvcr120.dll!_threadstartex dshowhelper.dll!CAMEvent::Wait Below Normal
ntdll.dll!_NtWaitForSingleObject@12()
KernelBase.dll!_WaitForSingleObjectEx@12()
KernelBase.dll!_WaitForSingleObject@8()
dshowhelper.dll!CAMEvent::Wait(unsigned long dwTimeout) Line 128
dshowhelper.dll!LogThread(void * param) Line 259
msvcr120.dll!_callthreadstartex() Line 376
msvcr120.dll!_threadstartex(void * ptd) Line 354
kernel32.dll!@BaseThreadInitThunk@12()
ntdll.dll!__RtlUserThreadStart()
ntdll.dll!__RtlUserThreadStart@8()

Will try the zap hang one again in a bit.

EDIT: Every time it hang (both ways) memory usage of MP is up over 600mb when usually for me it's under 200mb

 

[azzuro]

Hey seb. I have moved to madvr on my htpc.
The contrast seems better than evr.
For now video and tv work great.
Is combined with rtsp rework.

 

[kenwonders]

Here's a zap hang stack:

Not Flagged 3892 0 Worker Thread madVR.ax thread dshowhelper.dll!CCritSec::Lock Highest
ntdll.dll!_NtWaitForAlertByThreadId@8()
ntdll.dll!@RtlpWaitOnAddressWithTimeout@16()
ntdll.dll!RtlpWaitOnAddress()
ntdll.dll!RtlpWaitOnCriticalSection()
ntdll.dll!RtlpEnterCriticalSectionContended()
ntdll.dll!_RtlEnterCriticalSection@4()
dshowhelper.dll!CCritSec::Lock() Line 50
dshowhelper.dll!CAutoLock::CAutoLock(CCritSec * plock) Line 98
dshowhelper.dll!MPMadPresenter::RenderOsd(const char * name, __int64 frameStart, tagRECT * fullOutputRect, tagRECT * activeVideoRect) Line 271
madVR.ax!4a4876bf()
[Frames below may be incorrect and/or missing, no symbols loaded for madVR.ax]
madVR.ax!4a49b3f5()

Hoping this is the same cause as the madVR error, as the black screen is now far more likely than this issue.

 

[kenwonders]

After playing Elysium multiple times, shortly into playback MP crashed by itself. I attached, paused and got this, hope it's of use. Let me know if there's anything else you need.

Not Flagged 3256 0 Worker Thread madVR.ax thread nvd3dum.dll!6d8f35a0 Highest
ntdll.dll!_NtWaitForSingleObject@12()
KernelBase.dll!_WaitForSingleObjectEx@12()
KernelBase.dll!_WaitForSingleObject@8()
nvd3dum.dll!6d8f35a0()
[Frames below may be incorrect and/or missing, no symbols loaded for nvd3dum.dll]
nvd3dum.dll!6d8f424a()
nvd3dum.dll!6d906b30()
d3d9.dll!_Present()
d3d9.dll!_DdBltLH@4()
d3d9.dll!CSwapChain:resentMain(struct tagRECT const *,struct tagRECT const *,struct HWND__ *,struct _RGNDATA const *,unsigned long)
d3d9.dll!CBaseDevice:resentMain(struct tagRECT const *,struct tagRECT const *,struct HWND__ *,struct _RGNDATA const *,unsigned long)
d3d9.dll!CBaseDevice:resentEx(struct tagRECT const *,struct tagRECT const *,struct HWND__ *,struct _RGNDATA const *,unsigned long)
dshowhelper.dll!MPMadPresenter::RenderOsd(const char * name, __int64 frameStart, tagRECT * fullOutputRect, tagRECT * activeVideoRect) Line 299
madVR.ax!4a4876bf()
madVR.ax!4a49aa06()

 

[kilik360]

Good news this evening, I installed v85 and test a 24p movie and... same things drop frames a lot. Then I opened MP config to disable RefreshRateChanger and use the option with madvr instead.

I open MP start a 24p movie and... Tada ! Fluid playback has coming back from the dead !!! Thanks to kenwonders for the hint. I can now test future release again.

But this still unanswer the question... what is causing my trouble with RefreshRateChanger ? It does the exact same thing I had with MP2 and morpheus_xx was able to nailed it.

Here's my logs guys, keep up the great work.

 

[Sebastiii]

Good news this evening, I installed v85 and test a 24p movie and... same things drop frames a lot. Then I opened MP config to disable RefreshRateChanger and use the option with madvr instead.

I open MP start a 24p movie and... Tada ! Fluid playback has coming back from the dead !!! Thanks to kenwonders for the hint. I can now test future release again.

But this still unanswer the question... what is causing my trouble with RefreshRateChanger ? It does the exact same thing I had with MP2 and morpheus_xx was able to nailed it.

Here's my logs guys, keep up the great work.

I think refreshrate from MP didn't refresh the correct screen when using madVR no clue why but i think it's a root of cause.

 

[Sebastiii]

@kenwonders
https://forum.team-mediaportal.com/threads/madvr-support-in-mp1.133972/page-98#post-1185575
and
https://forum.team-mediaportal.com/threads/madvr-support-in-mp1.133972/page-98#post-1185583

It what we want now need to handle this too bad that i can't reproduce it

 

[Sebastiii]

Here's a zap hang stack:

Not Flagged 3892 0 Worker Thread madVR.ax thread dshowhelper.dll!CCritSec::Lock Highest
ntdll.dll!_NtWaitForAlertByThreadId@8()
ntdll.dll!@RtlpWaitOnAddressWithTimeout@16()
ntdll.dll!RtlpWaitOnAddress()
ntdll.dll!RtlpWaitOnCriticalSection()
ntdll.dll!RtlpEnterCriticalSectionContended()
ntdll.dll!_RtlEnterCriticalSection@4()
dshowhelper.dll!CCritSec::Lock() Line 50
dshowhelper.dll!CAutoLock::CAutoLock(CCritSec * plock) Line 98
dshowhelper.dll!MPMadPresenter::RenderOsd(const char * name, __int64 frameStart, tagRECT * fullOutputRect, tagRECT * activeVideoRect) Line 271
madVR.ax!4a4876bf()
[Frames below may be incorrect and/or missing, no symbols loaded for madVR.ax]
madVR.ax!4a49b3f5()

Hoping this is the same cause as the madVR error, as the black screen is now far more likely than this issue.

Hi

It would be needed to get the other call stack as well (the thread that calls the Shutdown())
So what we need to know is that if Shutdown() has been run prior to that crash

So pushed new code from Tourettes (Thanks) and attached release dll/pdb for reproduce the crash with hoping more information

 

[joecrow]

V85 is looking good for me, no crashes or hangs so far, will contiue testing.
I did notice on zapping, that it has gone back to displaying the channel I am changing from, briefly, during the switchover to the channel I am changing too. This behaviour however does not appear (subjectively) to take as long as that similarly reported on an earlier version. (i.e. not so clunky).

 

[alexdepalma]

Here's a zap hang stack:

Not Flagged 3892 0 Worker Thread madVR.ax thread dshowhelper.dll!CCritSec::Lock Highest
ntdll.dll!_NtWaitForAlertByThreadId@8()
ntdll.dll!@RtlpWaitOnAddressWithTimeout@16()
ntdll.dll!RtlpWaitOnAddress()
ntdll.dll!RtlpWaitOnCriticalSection()
ntdll.dll!RtlpEnterCriticalSectionContended()
ntdll.dll!_RtlEnterCriticalSection@4()
dshowhelper.dll!CCritSec::Lock() Line 50
dshowhelper.dll!CAutoLock::CAutoLock(CCritSec * plock) Line 98
dshowhelper.dll!MPMadPresenter::RenderOsd(const char * name, __int64 frameStart, tagRECT * fullOutputRect, tagRECT * activeVideoRect) Line 271
madVR.ax!4a4876bf()
[Frames below may be incorrect and/or missing, no symbols loaded for madVR.ax]
madVR.ax!4a49b3f5()

Hoping this is the same cause as the madVR error, as the black screen is now far more likely than this issue.

Hi


It would be needed to get the other call stack as well (the thread that calls the Shutdown())
So what we need to know is that if Shutdown() has been run prior to that crash

So pushed new code from Tourettes (Thanks) and attached release dll/pdb for reproduce the crash with hoping more information

Hello.
Testing V86Bis
I've found several thing before it freezes on channel change.

First, some MadVr Procesing errors messages at the first seconds of channel change (on new tuned channel). See picture.
I saw this message very occasionally before. Some times accompanied by errors in de-interlacing (interlace image not deinterlaced).

I've also seen on zap from channel A to channel B, OSD shows 3 different channel names on the screen before tuning channel B. no crash, no freeze, simple OSD showing several channels before A to B change. (A and B are in sequence in my channel list). OSD shown before the zap, consist on channels which aren't in between A and B. not related, may be in buffer memory?

And finally the freeze. This time, no timeshift running after. I changed channel once but happened like if few remote key presses where press in sequence, it jumped to several channels ahead. on returning to the pretended channel (several back), MP get freezes.

here logs
freeze around 11:55h

Edit: an additional thing to mention is that OSD zap (the image shown by MP regarding channel logo, channel, name, program name and progress bar) should be visible during few seconds on top of the new tuned channel. Here, it disappears before the actual channel is shown.
Also on this version, several times I get a bank image on new channel (audio is properly received). Very occasionally (rare) on V75 + 2gb core