]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
overview / pkg / ggml / sources / whisper.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 1caed1d)
hexagon: eliminate scalar VTCM loads via HVX splat helpers (llama/22993)
authorTrivikram Reddy <redacted>
Wed, 13 May 2026 00:28:02 +0000 (19:28 -0500)
committerGeorgi Gerganov <redacted>
Thu, 14 May 2026 18:26:48 +0000 (21:26 +0300)
commitbcaf4498269fea44b47621a4f0e8dff562c93cd4
tree36e65b052b30dfe9a9d22045432f292f069777a8tree
parent1caed1d2bae10279bc56261c2e03c3a86a27f4bbcommit | diff
hexagon: eliminate scalar VTCM loads via HVX splat helpers (llama/22993)

* hexagon: add hvx_vec_repl helpers and use those for splat-from-vtcm usecase

* hmx-mm: optimize per-group scale handling

* hmx-fa: optimize slope load from vtcm

* hmx-fa: use aligned access where possible in hmx-utils

* hexagon: add hvx_vec_repl_2x_f16 helper and consolidate repl helpers

---------

Co-authored-by: Max Krasnyansky <redacted>
ggml/src/ggml-hexagon/htp/hmx-flash-attn-ops.c diff | blob | history
ggml/src/ggml-hexagon/htp/hmx-matmul-ops.c diff | blob | history
ggml/src/ggml-hexagon/htp/hmx-utils.h diff | blob | history
ggml/src/ggml-hexagon/htp/hvx-repl.h [new file with mode: 0644] blob
ggml/src/ggml-hexagon/htp/hvx-utils.h diff | blob | history
Packaging of ggerganov/whisper.cpp