JMH v0.9.2

[Aleksey Shipilev]

Hi there,

JMH 0.9.2 is released and available in Maven Central (thanks again
Evgeny!). This is the stabilization release towards JMH 1.0, which
should happen later in July. Please test migrate to, and test 0.9.2!

Highlights of JMH 0.9.2:

 * "batch size" is now available for all benchmark modes; it makes sense
to use only in those modes which measure individual timings (e.g.
SingleShot and SampleTime), but other modes are also compensating for
the batch size to yield comparable results.

 * runners print stdout/stderr now, if no profilers are claiming it; it
was changed before in the course of profiler re-work, and *Java*
System.out/err was propagated without problems. Unfortunately, if you
were using native printers, e.g. with VM options like -XX:+PrintGC, the
output was swallowed. It isn't now.

 * human-readable output for multiple forked runs is improved: now we
print the aggregated result over all the forks, not the individual
results. This makes the on-going results exactly the same as printed in
the final report.

 * "perf"/"perfasm" auto-detection fixed: it does not print cryptic
messages when perf is not available on system now.

 * "perfasm" profiler usability is improved with grouping profiles by
method name, and by library name.

 * "perfasm" now uses hotspot.log as the intermediate file, which means
you can get the perf-annotated hotspot.log with
-Djmh.perfasm.saveLog=true now.

 * "stack" profiler now groups methods by thread state (patch by Sergey
Kuksenko).

 * ASM generators are updated to ASM 5.0.3 now, which allows us to parse
the Java 8 bytecode; it does not matter much, because reflection-based
generators seem to be working just fine.

Enjoy!

Thanks,
-Aleksey.