llama.cpp
sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs
#12858
Merged
Merged
sycl : implementation of reordered Q4_0 MMVQ for Intel GPUs #12858
Alcpz
merged
17 commits into
ggml-org:master
from
Alcpz:Alcpz/mmvq_q4_0_reorder
sycl : Implemented reorder Q4_0 mmvq
187451b6
sycl : Fixed mmvq being called when reorder is disabled
9c8d809f
NeoZhangJianyu
commented
on 2025-04-10
sycl : Improved comments in the quants header
52b16227
Rbiessy
commented
on 2025-04-11
ShanoToni
commented
on 2025-04-11
AD2605
commented
on 2025-04-11
Use static_assert
e8555ab4
safe_div -> ceil_div
b60d6376
Clarify qi comment
fc768f31
change the reorder tensor from init to execute OP
c7500c93
dbg
1e0c4cfe
Merge remote-tracking branch 'neozhang/jianyuzh_fix_reorder' into Alc…
de608194
Undo changes to test-backend-ops
dc19cd5f
Refactor changes on top of q4_0 reorder fix
351ef2b9
Merge remote-tracking branch 'upstream/master' into Alcpz/mmvq_q4_0_r…
34f7bede
Missing Reverts
d61dda32
Alcpz
commented
on 2025-04-28
Rbiessy
approved these changes
on 2025-04-30
Refactored opt_for_reorder logic to simplify code path
48480c89
AD2605
approved these changes
on 2025-05-06
ShanoToni
approved these changes
on 2025-05-06
qnixsynapse
approved these changes
on 2025-04-15
Merge branch 'master' into Alcpz/mmvq_q4_0_reorder
6afb367e
Explicit inlining and unroll
6fe27eb4
Renamed mul_mat_algo enum for consistency
e809b073
Alcpz
commented
on 2025-05-09
Reviewers
Assignees
No one assigned
Labels
ggml
SYCL
Milestone
No milestone
Login to write a write a comment.
Login via GitHub