]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: use registers instead of smem in topk-moe (#16647)
authorAman Gupta <redacted>
Sat, 18 Oct 2025 09:52:53 +0000 (17:52 +0800)
committerGitHub <redacted>
Sat, 18 Oct 2025 09:52:53 +0000 (11:52 +0200)
commit38355c6c8e43204e11a22daa7483082c0ff01e71
tree36bb5b69df3a88d2ae32588dd5269379ee8d807f
parent81387858f1fbcc1acedbd308486e1016618ca8f8
CUDA: use registers instead of smem in topk-moe (#16647)

Uses the technique used in the vulkan PR #16641. Neat trick!
ggml/src/ggml-cuda/topk-moe.cu