rmuir opened a new pull request, #12667: URL: https://github.com/apache/lucene/pull/12667
Following up to @dweiss work, this gives us the same benchmarks as https://github.com/rmuir/vectorbench, just without the code duplication and maintenance hassle. Each method is simply invoked with and without `--add-modules=jdk.incubator.vector`, so we can easily compare the performance of both the scalar and vector implementations. Same results we were getting before, with less code. I also like that it tests the "real" codepath through lucene, we just call VectorUtil methods with different JVM args. ``` Benchmark (size) Mode Cnt Score Error Units VectorUtilBenchmark.binaryCosineScalar 1024 thrpt 5 0.763 ± 0.012 ops/us VectorUtilBenchmark.binaryCosineVector 1024 thrpt 5 3.487 ± 0.080 ops/us VectorUtilBenchmark.binaryDotProductScalar 1024 thrpt 5 1.404 ± 0.032 ops/us VectorUtilBenchmark.binaryDotProductVector 1024 thrpt 5 7.120 ± 0.133 ops/us VectorUtilBenchmark.binarySquareScalar 1024 thrpt 5 1.109 ± 0.898 ops/us VectorUtilBenchmark.binarySquareVector 1024 thrpt 5 5.938 ± 1.490 ops/us VectorUtilBenchmark.floatCosineScalar 1024 thrpt 5 0.612 ± 0.071 ops/us VectorUtilBenchmark.floatCosineVector 1024 thrpt 5 5.953 ± 0.118 ops/us VectorUtilBenchmark.floatDotProductScalar 1024 thrpt 5 1.926 ± 0.060 ops/us VectorUtilBenchmark.floatDotProductVector 1024 thrpt 5 11.865 ± 1.057 ops/us VectorUtilBenchmark.floatSquareScalar 1024 thrpt 5 1.386 ± 0.065 ops/us VectorUtilBenchmark.floatSquareVector 1024 thrpt 5 9.689 ± 0.102 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
