kaivalnp commented on PR #15010: URL: https://github.com/apache/lucene/pull/15010#issuecomment-3134744902
Sharing some benchmarks.. `10_x` with `-reindex`: ``` recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn beamWidth quantized index(s) index_docs/s force_merge(s) num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType 0.963 1.889 1.888 0.999 100000 100 50 64 250 no 11.57 8641.55 24.98 1 77.49 292.969 292.969 HNSW ``` `10_x` without `-reindex`: ``` recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn beamWidth quantized index(s) index_docs/s force_merge(s) num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType 0.963 1.202 1.200 0.999 100000 100 50 64 250 no 0.00 Infinity 0.12 1 77.49 292.969 292.969 HNSW ``` This PR with `-reindex`: ``` recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn beamWidth quantized index(s) index_docs/s force_merge(s) num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType 0.962 1.193 1.192 0.999 100000 100 50 64 250 no 11.47 8716.12 24.22 1 77.47 292.969 292.969 HNSW ``` This PR without `-reindex`: ``` recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn beamWidth quantized index(s) index_docs/s force_merge(s) num_segments index_size(MB) vec_disk(MB) vec_RAM(MB) indexType 0.962 1.177 1.176 0.999 100000 100 50 64 250 no 0.00 Infinity 0.12 1 77.47 292.969 292.969 HNSW ``` **Note**: I ran these with `-XX:CompileCommand=inline,*PanamaVectorUtilSupport.*` to avoid the cold-start problem! I see \~37% improvement in the "search after indexing" case, and no difference in the "only search" case -- the trend matches the original PR -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org