cuda-kernels

LLM speculative inference server for heterogeneous hardware & consumer GPUs

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

Suckless no dependencies CUDA/C++ port of NVIDIA's DVLT, feed it images, get a 3D point cloud + camera poses. NO python. One fast 5MB binary.