gunnarmorling / concatbenchmark

Benchmark to analyze the impact of biased locking deprecation and potential removal

String Concat Benchmark

Benchmark to analyze the impact of biased locking deprecation and potential removal.

Build

mvn clean verify

Results

java -jar target/benchmarks.jar -f 1 -wi 5 -i 5

# JMH version: 1.25.2
# VM version: JDK 15, OpenJDK 64-Bit Server VM, 15+36-1562

Benchmark                                 Mode  Cnt          Score          Error  Units
StringConcatBenchmark.testStringBuffer   thrpt    5   39366534,323 ±  2374661,524  ops/s
StringConcatBenchmark.testStringBuilder  thrpt    5  166810460,263 ± 11673046,169  ops/s

java -XX:+UseBiasedLocking -XX:BiasedLockingStartupDelay=0 -jar target/benchmarks.jar -f 1 -wi 5 -i 5

# JMH version: 1.25.2
# VM version: JDK 15, OpenJDK 64-Bit Server VM, 15+36-1562
# VM options: -XX:+UseBiasedLocking -XX:BiasedLockingStartupDelay=0

Benchmark                                 Mode  Cnt          Score         Error  Units
StringConcatBenchmark.testStringBuffer   thrpt    5   86009387,696 ± 6154578,359  ops/s
StringConcatBenchmark.testStringBuilder  thrpt    5  159590813,813 ± 4191626,720  ops/s

Open Questions

Why is StringBuffer impacted that much by biased locking deactivation, shouldn't it be subject to lock elision?

License

This code base is available ander the Apache License, version 2.