llama.cpp
ggml : optimize cuda ssm_scan using warp-level reduction
#18505

Merged

ggml : optimize cuda ssm_scan using warp-level reduction #18505

am17an merged 3 commits into ggml-org:master from Aadeshveer:ggml-cuda-ssm-scan-opt

ggml : optimize cuda ssm_scan using warp-level reduction

64b08fed

github-actions added Nvidia GPU

github-actions added ggml

gabe-l-hart approved these changes on 2026-01-05

Aadeshveer force pushed 279 days ago

Aadeshveer requested a review from

gabe-l-hart 279 days ago

am17an approved these changes on 2026-01-06

ggml : apply code review suggestions (style, const, constexpr)

67f2d003

Aadeshveer force pushed to 67f2d003 278 days ago

Aadeshveer requested a review from

am17an 278 days ago

ggml : add TODO regarding stride consistency

4c8271c5

am17an merged 24af22fc into master 278 days ago

Reviewers

am17an

gabe-l-hart

Assignees

No one assigned

Labels

Nvidia GPU ggml

Milestone

No milestone