npu: a 17..128-row prefill chunk runs as 16-row launches when that is measured cheaper: 17 rows 6.95 s -> 1.94 s (fixes #1959) - #1963

npu: a 17..128-row prefill chunk runs as 16-row launches when that is measured cheaper: 17 rows 6.95 s -> 1.94 s (fixes #1959)#1963

bong-water-water-bong merged 8 commits into

main1bit-MONSTER/engine:mainfrom

cl-npuperf/prefill-narrow-split1bit-MONSTER/engine:cl-npuperf/prefill-narrow-splitCopy head branch name to clipboard