Benchmarking with heavily-assymetric threads? (full message)

[Lev Serebryakov]

Hello Jmh-dev,

 Sorry for premature send

 I need to benchmark some sort of single producer single consumer ring
buffer (queue), which is blocking on both sides. There is one more trick:
read and write batches are not equal. Here could be situation, when writer
writes 1000 element per invocation (and blocks till there is space for 1000
elements) and reader reads 16 elements per invocation (and blocks till there
are at least 16 elements).

 And here is a problem: looks like JMH try to equalize number of calls to
different benchmarks in one group and when it is impossible it start to run
iterations up to timeout, and like this.

 For example, when imbalance is 1:10, one iteration takes when it is
configured to be 10 seconds. When imbalance is 1:1000, I got *interrupted*
iterations. Results looks meaningless, too.

-- 
Best regards,
 Lev                            mailto:lev at serebryakov.spb.ru