llama.cpp
ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table
#30737
Merged
Merged
ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table #30737
chaxu01
merged
5 commits into
ggml-org:master
from
boshjerns:submit/cpu-gelu-quick
CISC
commented
on 2026-10-18
ggml-cpu : compute f32 gelu_quick and geglu_quick without the fp16 table
6e179150
ggml-cpu : keep the GGML_GELU_QUICK_FP16 table path
b7685ead
ggml-cpu : add the VXE gelu_quick path
6d25cfcb
tests : add test-gelu-quick for the f32 CPU kernels
d1505fc1
taronaeo
approved these changes
on 2026-10-19
tests : drop inf/nan inputs from test-gelu-quick
dea39d84
chaxu01
approved these changes
on 2026-10-20
Login to write a write a comment.
Login via GitHub