llama.cpp
[SYCL] Use native subgroup size for K-quant DMMV kernels on Intel
#21700

Merged

[SYCL] Use native subgroup size for K-quant DMMV kernels on Intel #21700

ggerganov merged 1 commit into ggml-org:master from PMZFX:opt/kquant-dmmv-subgroup16

PMZFX requested a review 76 days ago

github-actions added ggml

github-actions added SYCL

PMZFX changed the title ~~[SYCL] Use subgroup size 16 for K-quant DMMV kernels on Intel (2.3x–2.7x pp on Arc B70)~~ [SYCL] Use native subgroup size for K-quant DMMV kernels on Intel 76 days ago

arthw closed this 75 days ago

arthw reopened this 75 days ago

arthw approved these changes on 2026-04-11

PMZFX force pushed from 839c7e22 to 6f28d8cc 69 days ago

SYCL: use native subgroup size for K-quant DMMV

4213f77b

PMZFX force pushed from 6f28d8cc to 4213f77b 11 days ago

arthw approved these changes on 2026-06-15

arthw added merge ready

ggerganov merged e3bb1add into master 10 days ago

Reviewers

arthw

Assignees

No one assigned

Labels

ggml merge ready SYCL

Milestone

No milestone

llama.cpp [SYCL] Use native subgroup size for K-quant DMMV kernels on Intel #21700 Merged

[SYCL] Use native subgroup size for K-quant DMMV kernels on Intel #21700

llama.cpp
[SYCL] Use native subgroup size for K-quant DMMV kernels on Intel
#21700

Merged