Reply to thread

Message

<blockquote data-quote="Sebastiii" data-source="post: 1195535" data-attributes="member: 60104">Ok <img src="data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7" class="smilie smilie--sprite smilie--sprite1" alt=":)" title="Smile&nbsp; &nbsp; :)" loading="lazy" data-shortname=":)" /> I really need to reproduce it or reproduce it on your computer and test different try <img src="data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7" class="smilie smilie--sprite smilie--sprite1" alt=":)" title="Smile&nbsp; &nbsp; :)" loading="lazy" data-shortname=":)" />Yesterday, i have found the leak on my HTPC <img src="data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7" class="smilie smilie--sprite smilie--sprite1" alt=":)" title="Smile&nbsp; &nbsp; :)" loading="lazy" data-shortname=":)" /> It does happen like this for some reason when we disconnect the link between the video renderer to madVR, somehow when release madVR filter, it didn't fully release and didn't call the c++ destruction instance and then the memory is not freed even if the madVR icon is removed.So i added back on C# side the finalRelease comObject call instead of just release comObject and when doing that, the destructor is called on C++ and then GPU memory is freed.It will surely not change anything about the hang/freeze, from log it doesn't seems to be a deadlock (hard freeze and MP must be killed) but some kind of no refresh.It seems that i can reproduce something similar, while debugging on htpc, mp was running in fullscreen and on top i was having baretail + GPUz to follow live log and GPU memory, so focus MP but have those 2 tools with always on top settings, it was hard to trigger Fullscreen video to goes to little or going to home etc. but MP was reacting and the video was working and i was able to do pause etc.</blockquote>

[QUOTE="Sebastiii, post: 1195535, member: 60104"] Ok :) I really need to reproduce it or reproduce it on your computer and test different try :) Yesterday, i have found the leak on my HTPC :) It does happen like this for some reason when we disconnect the link between the video renderer to madVR, somehow when release madVR filter, it didn't fully release and didn't call the c++ destruction instance and then the memory is not freed even if the madVR icon is removed. So i added back on C# side the finalRelease comObject call instead of just release comObject and when doing that, the destructor is called on C++ and then GPU memory is freed. It will surely not change anything about the hang/freeze, from log it doesn't seems to be a deadlock (hard freeze and MP must be killed) but some kind of no refresh. It seems that i can reproduce something similar, while debugging on htpc, mp was running in fullscreen and on top i was having baretail + GPUz to follow live log and GPU memory, so focus MP but have those 2 tools with always on top settings, it was hard to trigger Fullscreen video to goes to little or going to home etc. but MP was reacting and the video was working and i was able to do pause etc. [/QUOTE]

Verification