Jeff Bolz fbf720dc9f vulkan: Use cm2 decode_vector for mul_mat_id B matrix loads (llama/23991) ...

This allows vec4 loads of the B elements. Also increase BK to 64 when this is
enabled. Neither of these alone is consistently faster, but together these give
a nice speedup.

In ggml-vulkan.cpp, we need to make sure the B matrix alignment and stride are
multiples of 4.

2026-06-15 10:33:53 +03:00

..

cmake

ggml : Parallelize quant LUT init (llama/23595)

2026-05-25 12:26:07 +03:00

include

TP: quantized KV cache support (llama/23792)

2026-06-08 14:36:36 +03:00

src

vulkan: Use cm2 decode_vector for mul_mat_id B matrix loads (llama/23991)

2026-06-15 10:33:53 +03:00

.gitignore

…

CMakeLists.txt

ggml : bump version to 0.14.0 (ggml/1533)

2026-06-08 14:36:36 +03:00