carterkozak commented on pull request #573: URL: https://github.com/apache/logging-log4j2/pull/573#issuecomment-924387654
> imagine the results may depend on the distinct number and shape of formatter patterns used in an application, as that may push C2 to optimize differently for monomorphic, bimorphic, or megamorphic calls and some of the potential pitfalls of class hierarchy optimizations @schlosna In this case the additional subclasses shouldn't impact overall performance because the PatternLayout operates with an array of LogEventPatternConverter instances based on the input pattern. There are dozens of types that may be used, we're exchanging some for others to reduce the variables introduced into the system. Running with java 11: This PR: ``` $ java -jar log4j-perf/target/benchmarks.jar ".*PatternLayoutBenchmark.serializableMCDEx" -f 1 -wi 4 -i 4 -w 4s -r 4s -t 1 -bm avgt ... Benchmark Mode Cnt Score Error Units PatternLayoutBenchmark.serializableMCDEx avgt 4 84.154 ± 10.530 ns/op ``` (second run after running both branches a first time, to verify the results are reproducible) ``` Benchmark Mode Cnt Score Error Units PatternLayoutBenchmark.serializableMCDEx avgt 4 87.192 ± 16.008 ns/op ``` Compared to the 2.x branch: ``` java -jar log4j-perf/target/benchmarks.jar ".*PatternLayoutBenchmark.serializableMCDEx" -f 1 -wi 4 -i 4 -w 4s -r 4s -t 1 -bm avgt ... Benchmark Mode Cnt Score Error Units PatternLayoutBenchmark.serializableMCDEx avgt 4 152.972 ± 2.614 ns/op ``` Second run of release-2.x after the second run of the PR branch: ``` Benchmark Mode Cnt Score Error Units PatternLayoutBenchmark.serializableMCDEx avgt 4 154.212 ± 8.509 ns/op ``` -- 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: notifications-unsubscr...@logging.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
