Repository navigation
The vector implementation for JavaFastPFOR is slower than the scalar one. #1
Description
Benchmark result
Benchmark Mode Cnt Score Error Units
IntCompressionBench.fastPFORCompression thrpt 12 210.045 ± 4.269 ops/s
IntCompressionBench.fastPFORDecompression thrpt 12 621.802 ± 6.700 ops/s
IntCompressionBench.vectorFastPFORCompression thrpt 12 87.594 ± 0.363 ops/s
IntCompressionBench.vectorFastPFORDecompression thrpt 12 165.013 ± 4.739 ops/s
How to reproduce
Follow https://github.com/mulugetam/VectorJavaFastPFOR/tree/main/jmh
Run
numactl -m 0 -C 2 java --add-modules jdk.incubator.vector -jar target/benchmarks.jar IntCompressionBenchEnvironment
AWS m5.xlarge, ubuntu-22.04
JDK 19.0.1
Gradle URL: https://services.gradle.org/distributions-snapshots/gradle-7.6-20220924231105+0000-bin.zip
Activity
lemire commentedon Nov 25, 2022
What is your hardware?
XSAM commentedon Nov 26, 2022
AWS EC2 m5.xlarge