mccullocht commented on PR #15742: URL: https://github.com/apache/lucene/pull/15742#issuecomment-3956642283
x86 benchmarks: ```Baseline: Benchmark (size) Mode Cnt Score Error Units VectorUtilBenchmark.binaryHalfByteDotProductBothPackedVector 1024 thrpt 15 23.675 ± 0.031 ops/us VectorUtilBenchmark.binaryHalfByteDotProductSinglePackedVector 1024 thrpt 15 11.855 ± 0.059 ops/us VectorUtilBenchmark.binaryHalfByteDotProductVector 1024 thrpt 15 68.691 ± 0.816 ops/us VectorUtilBenchmark.binaryHalfByteSquareBothPackedVector 1024 thrpt 15 30.269 ± 0.148 ops/us VectorUtilBenchmark.binaryHalfByteSquareSinglePackedVector 1024 thrpt 15 12.770 ± 0.139 ops/us VectorUtilBenchmark.binaryHalfByteSquareVector 1024 thrpt 15 72.246 ± 0.406 ops/us Experiment: Benchmark (size) Mode Cnt Score Error Units VectorUtilBenchmark.binaryHalfByteDotProductBothPackedVector 1024 thrpt 15 35.692 ± 0.152 ops/us VectorUtilBenchmark.binaryHalfByteDotProductSinglePackedVector 1024 thrpt 15 42.843 ± 0.523 ops/us VectorUtilBenchmark.binaryHalfByteDotProductVector 1024 thrpt 15 65.359 ± 0.618 ops/us VectorUtilBenchmark.binaryHalfByteSquareBothPackedVector 1024 thrpt 15 34.310 ± 0.105 ops/us VectorUtilBenchmark.binaryHalfByteSquareSinglePackedVector 1024 thrpt 15 42.061 ± 0.312 ops/us VectorUtilBenchmark.binaryHalfByteSquareVector 1024 thrpt 15 57.315 ± 0.760 ops/us ``` VectorUtilBenchmark.binaryHalfByteSquareVector is quite a bit slower, but I also don't know under what conditions we'd actually run this. I suspect we don't do a ton of distance comparisons using two unpacked int4 vectors. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
