llama.cpp
opencl: fix several bugs where the backend aborts
#27630

Merged

opencl: fix several bugs where the backend aborts #27630

lhez merged 9 commits into ggml-org:master from qualcomm:hq/opencl-device-robustness-r0822

github-actions added ggml

github-actions added OpenCL

wanghqc marked this pull request as ready for review 63 days ago

wanghqc requested a review 63 days ago

max-krasnyansky approved these changes on 2026-09-09

opencl: do not abort on OpenCL 2.0 devices under a 3.0 platform

29e9848f

opencl: never enqueue a zero-size NDRange

72a5f45b

opencl: decline CONV_2D for non-contiguous operands

802bed68

opencl: q6_K set_tensor — retry + ALLOC_HOST_PTR fallback on alloc fa…

1635ef2b

opencl: generalize alloc retry to all set_tensor / get_tensor / buffe…

57df6570

opencl: actionable error on Adreno X1 per-context memory wall

9152ed46

opencl: clarify comments, refactor

0df91b84

opencl: revise error msg

ce7e22ea

lhez force pushed to ce7e22ea 59 days ago

opencl: format

9af32029

lhez force pushed to 9af32029 59 days ago

lhez requested a review from

max-krasnyansky 57 days ago

max-krasnyansky approved these changes on 2026-09-11

lhez approved these changes on 2026-09-11

lhez merged 8a56aedd into master 56 days ago

Reviewers

lhez

max-krasnyansky

Assignees

No one assigned

Labels

ggml OpenCL

Milestone

No milestone