[WiP] madVR support in MP1 (6 Viewers)

[Sebastiii]

V36

I think (at least on my PC) succeed to force madVR to destroy his instance, so i would like to know if it's better for you guys

 

[Sebastiii]

So the next step test, would be to stop the TV after 6 or 8 zap (wait 10 or 20s) and try again and stop/wait etc.
It will confirm that it's really an issue with memory not freed.

When i register madVR[debug].ax, we can see that on zap the graph is not destroyed when zapping from tuner to tuner. When doing the stop, after few second the madVR instance is freed (because Garbage Collector do his job).

When talking with madshi, he tell me that madVR destroye his instance when all related COM object are release and the counter goes to 0. But with .NET, the release COM object are done async and memory is freed by garbage collector. So right now i don't know if we can do another way lol

Maybe @arion_p can have an idea ?

I'm sure i did some dirty hack a few years back with something like GC.Collect() Or failing that a huge MP pause that the user will think is a HD spinning up or something.

If I stop the TV every few zaps, it is not going bad.

Yep there is such code (i know that we shouldn't force it) "GC.Collect();" but even with it madVR was not destroyed quickly, but it seems that i have something to test with V36

 

[Sebastiii]

I will surely upload a new bin because on my dev PC the stop/destroy is not always trigger.

 

[daWooky]

nope
v36, no destroy here

 

[Sebastiii]

V37
I'm using madVR.ax the debug version, so i hope it will be ok (for the timing i have added).

 

[kenwonders]

I had just made some new logs... This also affects EVR on the v35 build. If you start Dave when Blackadder II is on (like right now), the 4:3 aspect seems to confuse things and you get a small window in the top left. If you 'change' to this channel from a working 16:9 channel, it shows full screen but over to the far left instead of centred, going to the guide and back corrects this.

Right that bug logged onto testing v37!

 

[daWooky]

v37 no change- screenshot 23
memory usage goes up after stop or card switch occured
similar conclusion as on v3x before no hardcore zapzerap on two cards, zap on only one is ok

the 4:3 aspect seems to confuse things

that funny bug i saw that as well a few versions ago

 

[Sebastiii]

v37 no change- screenshot 23
no hardcore zapzerap on two cards, zap on only one is ok

the 4:3 aspect seems to confuse things

that funny bug i saw that as well a few versions ago

When did you get the Direct3D Device failed already ?
Could you test with watchdog and upload log plz ?

 

[kenwonders]

ITS WORKING! for me anyway. I just did over 50 (yes really) changes without issue.

 

[Sebastiii]

ITS WORKING! for me anyway. I just did over 50 (yes really) changes without issue.

Ah that's a good news (even if to achieve this the code is not great)
So maybe for @daWooky , the timing need to be raise.