llama.cpp
opencl: tune the quant paths for Intel Xe-LP GPUs to improve its TG and PP performance
#26438

Merged

opencl: tune the quant paths for Intel Xe-LP GPUs to improve its TG and PP performance #26438

lhez merged 4 commits into ggml-org:master from qualcomm:hq/intel-quant-tuning-r0730

github-actions added ggml

github-actions added OpenCL

wanghqc force pushed 73 days ago

wanghqc force pushed to 5665bffc 59 days ago

opencl: Q4_K/Q5_K mul_mv N_DST 4->8 on Intel for 2x activation reuse

d87f431e

opencl: Q4_K mul_mm 8x8 tile fot Intel

df11be2b

opencl: Q5_K mul_mm 8x8 tile for Intel

2fee355a

opencl: Q4_K mul_mv N_DST 8->16 for Intel

b0e0fd19

lhez force pushed from 5665bffc to b0e0fd19 51 days ago

lhez approved these changes on 2026-08-30

lhez marked this pull request as ready for review 51 days ago

lhez requested a review 51 days ago

lhez requested a review from

max-krasnyansky 51 days ago

max-krasnyansky approved these changes on 2026-08-30

lhez merged 010be968 into master 51 days ago

Reviewers

max-krasnyansky

lhez

Assignees

No one assigned

Labels

ggml OpenCL

Milestone

No milestone