ggml : process data in smaller chunks in CUDA ggml_top_k() implementation to reduce temporary buffers memory usage - #24776

Merged

fairydreaming merged 6 commits into

ggml-org:masterfrom

fairydreaming:chunked-top-k

Jul 9, 2026

Commits on Jun 18, 2026

ggml : process data in smaller chunks in CUDA ggml_top_k() implementation to reduce temporary buffers memory usage
sszymczy
committed
c1a79a6

Commits on Jun 19, 2026

ggml : allocate tmp_dst only only once before the loop
sszymczy
committed
ca88122
chore : whitespaces

fairydreaming
and
ggerganov
authored
4d0e838

Commits on Jun 25, 2026

ggml : use chunked processing in both CUDA CUB top-k and argsort implementations
sszymczy
committed
8521ec4

Commits on Jun 30, 2026