Jeff Bolz 6efcd65945 vulkan: optimize flash attention split_k_reduce (#14554) ...

* vulkan: allow FA split_k with smaller KV values

* vulkan: spread split_k_reduce work across more threads

k_num can get rather large. Use the whole workgroup to reduce the M/L values.

Launch a thread for each element in the HSV dimension of the output. Helps a
lot for large HSV (like deepseek).

2025-07-08 20:11:42 +02:00

..

cmake

ggml-cpu : rework weak alias on apple targets (#14146)

2025-06-16 13:54:15 +08:00

include

CUDA: add bilinear interpolation for upscale (#14563)

2025-07-08 10:11:18 +08:00

src

vulkan: optimize flash attention split_k_reduce (#14554)

2025-07-08 20:11:42 +02:00

.gitignore

vulkan : cmake integration (#8119)

2024-07-13 18:12:39 +02:00

CMakeLists.txt

ggml : remove kompute backend (#14501)

2025-07-03 07:48:32 +03:00