]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
hexagon: further optimization and tuning of matmul and dot kernels (llama/19407)
authorMax Krasnyansky <redacted>
Thu, 12 Feb 2026 07:04:27 +0000 (23:04 -0800)
committerGeorgi Gerganov <redacted>
Sun, 15 Feb 2026 19:44:37 +0000 (21:44 +0200)
commit304205679c650c3e5977e0b01b7b9bd022336767
tree6bf27c4a9b9de39f45a24ce8d00a0967c0113485
parent0326fd37dd4170f0264a909c1ab58ec230dee02b
hexagon: further optimization and tuning of matmul and dot kernels (llama/19407)

* ggml-hexagon: implement 2x2 matmul kernel

* hexmm: implement vec_dot_rx2x2 for Q8_0 and MXFP4

* hexagon: fix editor config failures

* hexagon: refactor matmul ops to use context struct and remove wrappers

Also implement vec_dot_f16 2x2

* hexagon: refactor dyn quantizers to use mmctx

* hexagon: remove mm fastdiv from op_ctx

* hexagon: refactor matmul entry point to reduce code duplication

---------

Co-authored-by: Trivikram Reddy <redacted>
ggml/src/ggml-hexagon/htp/htp-ops.h
ggml/src/ggml-hexagon/htp/matmul-ops.c