Articles describes how to make pure C# rendering video on VMR9 with custom allocator presenter over. VMR9 Allocator Presenter in C# with. My effect sample. So, we have a bug saying there is a leak in the VMR9Allocator sample (https. The VMRAllocator2 sample does not seem to suffer from the same problem.

So, we have a bug saying there is a leak in the VMR9Allocator sample (While I don't know much about either VMR or the VMR Allocator, I do know some things about leaks. So, I have spent some time mucking around here, and I believe I have made things better. First, a few notes: • I am running on vista sp1 64bit. • I am compiling all the apps to run x86 (not x64). • I am using the foo.avi test file to perform my tests • I am watching committed bytes, not working set size.

Scribd Username And Password Crack Torent. After each Play, I click the Close menu option and wait a few moments for the memory to settle. Now, I was able to see a good sized leak when I started running the code.

Sms Client Library.dll there. I did a couple of things that made minor improvements: • Add this code to the Allocator.Dispose method. Marshal.Release(unmanagedDevice); This apparently kept the 'Device' from getting released. However, if the previous call to GetObjectByValue was doing an unexpected AddRef, then we should have to add the same code down in AdviseNotify. But if you DO add the same code down in AdviseNotify, the program crashes (so don't do it). I spent some time trying to figure out where the extra AddRef was coming from, but I was not successful.

Maybe people who understand how VMR works can provide that last needed bit. In the meantime, this seems to have greatly reduced the amount of memory being used. Hate to bring up this old issue, hopefully I can shed some light on it though. Some time ago I found a similar issue with code similar to the VMRAllocator sample, and in tracing the problem I found it came down to the D3D Device not being released correctly - I figured there was still a ref to it somewhere, the memory was never being released and was still showing up in a profiler after a forced gc collect too. I could replicate this with the VMRAllocator sample, by opening & closing several files in the same instance of the app (as described in bug# 1754448). Using VidMemWatch I could see allocated video memory steadily increase, and a profiler would show the d3d device was not being released.

The VMRAllocator2 sample does not seem to suffer from the same problem, and one of the main differences is that VMRAllocator uses AllocateSurfaceHelper but VMRAllocator2 does not. At the time I still needed a render loop similar to VMRAllocator, so I ended up combining portions of the two demo apps so that AllocateSurfaceHelper was no longer used.