rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1556224675
i made the benchmarks easier to run with something like this: ``` git clone https://github.com/rmuir/vectorbench cd vectorbench mvn verify java -jar target/vectorbench.jar ``` I can confirm everything works on aarch64 and i am experimenting with the unrolling. definitely if we remove the unrolling it gets way slower. so i tried to unroll again (4x instead of 2x), it is only a slight improvement in performance on my skylake: ``` Benchmark (size) Mode Cnt Score Error Units DotProductBenchmark.dotProductNew 1024 thrpt 5 9.997 ± 0.999 ops/us DotProductBenchmark.dotProductNewNew 1024 thrpt 5 11.285 ± 0.161 ops/us DotProductBenchmark.dotProductOld 1024 thrpt 5 2.024 ± 0.028 ops/us ``` But on the aarch64 mac additionall unrolling basically doubles throughput again (7.785 -> 14.912). I am experimenting more with this. -- 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
