llama.cpp
hexagon: add support for Q1_0 and Q2_0
#30122
Merged

hexagon: add support for Q1_0 and Q2_0 #30122

tboinovski1
github-actions github-actions added ggml
github-actions github-actions added Hexagon
tboinovski1 tboinovski1 force pushed 16 days ago
tboinovski1 tboinovski1 force pushed 13 days ago
tboinovski1 tboinovski1 force pushed 12 days ago
tboinovski1 tboinovski1 force pushed 12 days ago
tboinovski1 hexagon: add support for q1_0 quant
a48b6761
tboinovski1 hexagon: enable Q1_0 weights in the 2D HMX MUL_MAT path
357cc5fa
max-krasnyansky hex-q1: cleanup gates, generalize tiling and remove in-loop divs
3c62d788
max-krasnyansky hex-tiling: document tiled formats in developer.md and update htp-ten…
7b00c315
tboinovski1 hex-q2: add support for Q2_0
07f3b45b
max-krasnyansky hex-q2: remove stale asserts and speacial cases
e614d278
max-krasnyansky hex-mm: fix MUL_MAT_NX assumption about square tiling, Q1/Q2 have non…
b93d1e63
max-krasnyansky hex-q1/2: minor optimizations (unrolls/un-unrolls/...)
c8c8a104
max-krasnyansky hex-q1: optimize Q1_0 layout to reduce unpack/vrmpy overhead
e9de33c8
max-krasnyansky hex-q1/q2_0: further optimize vec_dots
2fcd79f4
max-krasnyansky max-krasnyansky force pushed to 2fcd79f4 11 days ago
max-krasnyansky max-krasnyansky marked this pull request as ready for review 11 days ago
max-krasnyansky max-krasnyansky requested a review 11 days ago
github-actions github-actions added documentation
max-krasnyansky
max-krasnyansky max-krasnyansky changed the title hexagon: add support for Q1_0 hexagon: add support for Q1_0 and Q2_0 11 days ago
max-krasnyansky
max-krasnyansky approved these changes on 2026-10-17
max-krasnyansky max-krasnyansky added merge ready
lhez
lhez approved these changes on 2026-10-17
max-krasnyansky max-krasnyansky merged 797de63b into master 11 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone