rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1558372829
Here's a summary of where the perf sits for these various functions on my machines. It only takes 5 minutes to run a pass just for vector size of 1024 dimensions only to get an idea: ``` $ git clone https://github.com/rmuir/vectorbench.git $ cd vectorbench $ mvn verify $ java -jar target/vectorbench.jar -p size=1024 ``` intel skylake (256-bit vectors): ``` Benchmark (size) Mode Cnt Score Error Units BinaryCosineBenchmark.cosineDistanceNew 1024 thrpt 5 3.626 ± 0.045 ops/us BinaryCosineBenchmark.cosineDistanceOld 1024 thrpt 5 0.790 ± 0.086 ops/us BinaryDotProductBenchmark.dotProductNew 1024 thrpt 5 7.122 ± 0.329 ops/us BinaryDotProductBenchmark.dotProductOld 1024 thrpt 5 1.835 ± 0.039 ops/us BinarySquareBenchmark.squareDistanceNew 1024 thrpt 5 6.392 ± 0.057 ops/us BinarySquareBenchmark.squareDistanceOld 1024 thrpt 5 1.545 ± 0.247 ops/us FloatCosineBenchmark.cosineNew 1024 thrpt 5 6.074 ± 0.089 ops/us FloatCosineBenchmark.cosineOld 1024 thrpt 5 0.631 ± 0.006 ops/us FloatDotProductBenchmark.dotProductNew 1024 thrpt 5 12.108 ± 0.152 ops/us FloatDotProductBenchmark.dotProductOld 1024 thrpt 5 2.014 ± 0.020 ops/us FloatSquareBenchmark.squareNew 1024 thrpt 5 9.504 ± 0.171 ops/us FloatSquareBenchmark.squareOld 1024 thrpt 5 1.412 ± 0.028 ops/us ``` mac m1 arm (128-bit vectors): ``` Benchmark (size) Mode Cnt Score Error Units BinaryCosineBenchmark.cosineDistanceNew 1024 thrpt 5 2.259 ± 0.011 ops/us BinaryCosineBenchmark.cosineDistanceOld 1024 thrpt 5 1.046 ± 0.002 ops/us BinaryDotProductBenchmark.dotProductNew 1024 thrpt 5 6.142 ± 0.002 ops/us BinaryDotProductBenchmark.dotProductOld 1024 thrpt 5 3.107 ± 0.002 ops/us BinarySquareBenchmark.squareDistanceNew 1024 thrpt 5 6.142 ± 0.002 ops/us BinarySquareBenchmark.squareDistanceOld 1024 thrpt 5 3.100 ± 0.016 ops/us FloatCosineBenchmark.cosineNew 1024 thrpt 5 7.840 ± 0.006 ops/us FloatCosineBenchmark.cosineOld 1024 thrpt 5 1.076 ± 0.001 ops/us FloatDotProductBenchmark.dotProductNew 1024 thrpt 5 12.467 ± 0.005 ops/us FloatDotProductBenchmark.dotProductOld 1024 thrpt 5 3.823 ± 0.001 ops/us FloatSquareBenchmark.squareNew 1024 thrpt 5 14.329 ± 0.061 ops/us FloatSquareBenchmark.squareOld 1024 thrpt 5 3.185 ± 0.002 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
