rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1557631563
Seems to take quite a hit on my 256. And I suspect if you tried to make a "512 version" of the existing code it might be much better too? ByteVector.SPECIES_128 -> ShortVector.SPECIES_256 -> IntVector.SPECIES_512. No need to splitting into "parts". I'm also concerned it will error out if the user e.g. has only 64-bit vectors as the only possible size (e.g. avx disabled or vectorization not supported for the architecture). But maybe we can play with some of the idea more. ``` Benchmark (size) Mode Cnt Score Error Units BinaryDotProductBenchmark.dotProductNew 1024 thrpt 5 7.174 ± 0.602 ops/us BinaryDotProductBenchmark.dotProductNewNew 1024 thrpt 5 5.559 ± 0.109 ops/us BinaryDotProductBenchmark.dotProductOld 1024 thrpt 5 1.868 ± 0.019 ops/us ``` -- 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
