Files
whisper.cpp/ggml
Jeff Bolz 191def71ce vulkan: mul_mat_id coopmat2 optimizations (llama/15546)
* vulkan: mul_mat_id coopmat2 optimizations

Add a path for when the tile fits in BN/2, similar to what we have for mul_mat.

Only call fetch_scales/store_scales once per QUANT_K block, and once at the
beginning in case start_k is not aligned.

* Also add a path for BN/4 - worth a couple more percent
2025-09-20 13:42:45 +03:00
..
2025-09-20 13:42:39 +03:00