]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
CUDA: fix FA occupancy, optimize tile kernel (llama/15982)
authorJohannes Gäßler <redacted>
Wed, 17 Sep 2025 13:32:42 +0000 (15:32 +0200)
committerGeorgi Gerganov <redacted>
Sat, 20 Sep 2025 10:33:50 +0000 (13:33 +0300)
commit5437f5bceea5a252db3bb883f73f1ed95f4068ee
tree28cf30febaf6fd0ce614be35121231eb000e03ff
parentc3d23819c44cd5d394e38153be9cff854e1b8487
CUDA: fix FA occupancy, optimize tile kernel (llama/15982)
src/ggml-cuda/common.cuh
src/ggml-cuda/fattn-common.cuh
src/ggml-cuda/fattn-tile.cu
src/ggml-cuda/vendors/hip.h