dequantize + matrix multiplication CUDA kernels#2043

JohannesGaessler wants to merge 2 commits intoggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:cuda-matrix-matrix-4JohannesGaessler/llama.cpp:cuda-matrix-matrix-4Copy head branch name to clipboard