hexagon: optimize HMX matmul operations #21071
optimize hmx_mat_mul functions by calculating row and column tiles up…
de56c35d
refactor core_dot_chunk_fp16 to use size_t for tile counts and improv…
b2b21a36
wip
5e18f4ed
set scale outside of loop
a2628321
wip
ee95d92a
refactor core_mma_chunk_fp16 and mat_mul_qk_0_d16a32 to use size_t fo…
33d94316
wip
3a97015f
wip
6e291d8e
refactor transfer_output_chunk_fp16_to_fp32 to use size_t for dimensions
f43d68ca
refactor core_dot_chunk_fp16 to use size_t for tile row stride calcul…
42bd08ca
wip
ee951463
refactor hmx_mat_mul functions to use hvx_vec_splat_f16 for column sc…
91d88a35
chraac
requested a review
16 days ago
chraac
commented
on 2026-03-27
refactor hmx_mat_mul_permuted_w16a32_batched to streamline scale sett…
55d72581
refactor core_dot_chunk_fp16 to improve tile stride calculations for …
362c62ce
Merge branch 'master' into dev-hmx-opt
e31e30a6
refactor hmx_mat_mul functions to use Q6_V_vsplat_R for column scales…
7c1a5a39
Merge branch 'master' into dev-hmx-opt
3cd80415
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub