VRAM optimization + matrix multiplication discussion#1935

JohannesGaessler wants to merge 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:cuda-pool-testJohannesGaessler/llama.cpp:cuda-pool-testCopy head branch name to clipboard