[DRAFT] Speedup dequantize kernels#1221

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

dfyz:memory-bound-quantizedfyz/llama.cpp:memory-bound-quantizeCopy head branch name to clipboard