]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
overview / pkg / ggml / sources / whisper.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 21c65a7)
SYCL: improve MoE prefill throughput (llama/23142)
authorAlexey Kopytko <redacted>
Fri, 22 May 2026 12:50:17 +0000 (21:50 +0900)
committerGeorgi Gerganov <redacted>
Mon, 25 May 2026 09:26:07 +0000 (12:26 +0300)
commitb0c9f9005926301e8a8306bd300e86dbd09db41d
tree1a60de9a31c196ed216284d0a63a59ea8c17fe53tree
parent21c65a78a3472275db7ccdc40078fa4a3858724acommit | diff
SYCL: improve MoE prefill throughput (llama/23142)

- change `k_copy_src1_to_contiguous` so that uses a precomputed contiguous mapping where all rows "owned" by an expert are in one slice with a know starts and ends
- switch the `O(n_as * n_routed_rows)` contraption to a counting sort-based procedure with `O(n_as + n_routed_rows)` complexity
ggml/src/ggml-sycl/ggml-sycl.cpp diff | blob | history
Packaging of ggerganov/whisper.cpp