llama.cpp
ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa)
#30572

Merged

ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa) #30572

JohannesGaessler merged 1 commit into ggml-org:master from thelittlefireman:fix-fattn-tile-oob-check

ggml-cuda: fattn do not check oob if k_VKQ_max = (k_VKQ_max - nbatch_fa)

fb84a21b

thelittlefireman requested a review 15 days ago

thelittlefireman changed the title ~~ggml-cuda: fattn do not check oob if k_VKQ_max = (k_VKQ_max - nbatch_fa)~~ ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa) 15 days ago

github-actions added ggml

github-actions added CUDA

am17an approved these changes on 2026-10-16

JohannesGaessler approved these changes on 2026-10-18

JohannesGaessler merged 5aeac37a into master 12 days ago

Reviewers

JohannesGaessler

am17an

Assignees

No one assigned

Labels

ggml CUDA

Milestone

No milestone

llama.cpp ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa) #30572 Merged

ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa) #30572

llama.cpp
ggml-cuda: fattn do not check oob if k_VKQ_0 = (k_VKQ_max - nbatch_fa)
#30572

Merged