rmuir commented on PR #12311: URL: https://github.com/apache/lucene/pull/12311#issuecomment-1560162059
if you want to see what i mean, try running `java -XX:UseAVX=0 -jar target/vectorbench.jar -p size=1024`: ``` Benchmark (size) Mode Cnt Score Error Units BinaryCosineBenchmark.cosineDistanceNew 1024 thrpt 5 0.011 ± 0.001 ops/us BinaryCosineBenchmark.cosineDistanceOld 1024 thrpt 5 0.799 ± 0.017 ops/us BinaryDotProductBenchmark.dotProductNew 1024 thrpt 5 0.028 ± 0.003 ops/us BinaryDotProductBenchmark.dotProductOld 1024 thrpt 5 1.798 ± 0.294 ops/us BinarySquareBenchmark.squareDistanceNew 1024 thrpt 5 0.027 ± 0.002 ops/us BinarySquareBenchmark.squareDistanceOld 1024 thrpt 5 1.540 ± 0.109 ops/us FloatCosineBenchmark.cosineNew 1024 thrpt 5 3.153 ± 0.401 ops/us FloatCosineBenchmark.cosineOld 1024 thrpt 5 0.664 ± 0.141 ops/us FloatDotProductBenchmark.dotProductNew 1024 thrpt 5 7.443 ± 0.315 ops/us FloatDotProductBenchmark.dotProductOld 1024 thrpt 5 1.689 ± 0.220 ops/us FloatSquareBenchmark.squareNew 1024 thrpt 5 5.439 ± 0.275 ops/us FloatSquareBenchmark.squareOld 1024 thrpt 5 1.155 ± 0.293 ops/us ``` How to detect such incredibly trappy performance and prevent it? `IntVector.SPECIES_PREFERRED` still says 128 in this case. -- 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
