opencl: quant lm_head / decode GEMV and medium-batch GEMM optimizations (speculative decoding/MTP) #26477
wanghqc
force pushed
from
fa2eacc9
to
eb0607ee
107 days ago
wanghqc
force pushed
from
eb0607ee
to
127316f3
102 days ago
wanghqc
force pushed
from
127316f3
96 days ago
wanghqc
force pushed
to
8b0fb4b0
88 days ago
lhez
force pushed
from
8b0fb4b0
77 days ago
opencl: quant lm_head / decode GEMV and medium-batch GEMM optimizations
af7dffa8
opencl: guard q4_K/q6_K tiled_ns convert-kernel registration for non-…
b517cd33
opencl: gate q4_K MUL_MAT+GLU fusion dispatch to Adreno
aaf2e90e
opencl: require the noshuffle weight layout in the q4_K GLU fusion gate
11384baf
opencl: do not take the vectorized f16 mrow GEMV path on an unaligned…
c6d0997b
opencl: pass the new get_scale_min_k4 stride argument at the row-majo…
f6dcd645
opencl: enable the q4_K split-K decode GEMV only where it is measured…
c7ca886d
opencl: record the X1-85 split-K datapoint (neutral, exclusion confir…
d8d712b3
opencl: restrict the tiled lm_head/embed GEMV default to X2E/A8X
3e83f77b
opencl: fix q4_K variant kernels to read the transposed scales layout
b7386ebf
opencl: keep the flat-GEMV large-m escape opt-in
6dbdc9b7
opencl: guard the o4 GEMV store against the rounded-up dispatch tail
10289e3b
opencl: restore the tiled q4_K/q6_K layout on tensor read-back
ad3cef32
opencl: split-K for the q8_0 decode GEMV at small M
9ceaa97a
opencl: keep the q6_K noshuffle correctness escape ahead of the opt-i…
7833efff
lhez
force pushed
to
7833efff
77 days ago
lhez
marked this pull request as ready for review 77 days ago
lhez
requested a review
77 days ago
lhez
approved these changes
on 2026-09-03
lhez
merged
95ef7fc1
into master 76 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub