Hi HN, the last half year I spent some of my spare time on learning Vulkan (mainly the compute part) and applied it directly in developing a gaussian splatting renderer. I am not sure where I will go with this, but I thought some of you might find it interesting and/or have suggestions on how to proceed.

Ultimately, I want to Klartraum to be an extensive neural network and rendering inference engine (no training, backprop, autograd) running on embedded devices and VR headsets with high performance. My major obstacle right now is that writing GLSL compute kernels is tedious compared to CUDA ...

What do you think should be added so the library would be of use for others?