llama.cpp
ggml : process data in smaller chunks in CUDA ggml_top_k() implementation to reduce temporary buffers memory usage
#24776
Merged
Merged
ggml : process data in smaller chunks in CUDA ggml_top_k() implementation to reduce temporary buffers memory usage #24776
fairydreaming
merged
6 commits into
ggml-org:master
from
fairydreaming:chunked-top-k
ggml : process data in smaller chunks in CUDA ggml_top_k() implementa…
c1a79a69
ggml : allocate tmp_dst only only once before the loop
ca88122e
chore : whitespaces
4d0e8387
ggml : use chunked processing in both CUDA CUB top-k and argsort impl…
8521ec42
chore : separate argsort_f32_i32_cuda_bitonic() call from return stat…
807a9f76
ORippler
approved these changes
on 2026-06-30
chore : replace ternary operators with min/max
f887bc24
JohannesGaessler
approved these changes
on 2026-07-09
Reviewers
Assignees
No one assigned
Labels
ggml
CUDA
Milestone
No milestone
Login to write a write a comment.
Login via GitHub