lrahmann

> > The primary issue was DescriptorsHeap::allocate failing for CPU only descriptor heaps > > What was the reason behind the failure? I honestly have no idea. Probably just the...

@DarioSamo it bugs me aswell, but the exception does come from one of the nvidia_driver dlls. That is probably the reason why its not documented anywhere. If you check out...

> > When debugging it locally the nvidia driver is spewing E_OUT_OF_MEMORY exceptions whenever you try to allocate a new DescriptorHeap (non shader visible) after hitting this limit. I can...

@DarioSamo Alternative to that would be to go down the SDL route and simply only hand out single ones. I can also look into implementing a variable size allocation manager...

@DarioSamo pushed a version which uses RBTrees to create a structure similar to the one described in your referenced post. It does allocate multiple buffers within a unified global address...

> Where does it crash under this latest commit? It crashes under the master commit as well so I don't think its related to my PR. I just retested this...

@DarioSamo see my comment, I can allocate new blocks. Its just done in chunks of 1024. I tested this PR our game and against the minimal example from the issue...

No issue, I dug a bit deeper and for reference here are the constants for DiligentEngine https://github.com/DiligentGraphics/DiligentCore/blob/e8f871737f33619de5ceec7474e6906131472c5b/Graphics/GraphicsEngine/interface/GraphicsTypes.h#L3763 They split this number between the different heap types. Also they do create...

Okay my bad, the crash on master was related to a version mismatch in the agility sdk. Now both editors run fine.

will fix the issue with the CI asap. there are a few open questions from my side: - Is crashing inside the functions a good way to fail on error...