Loading...

XML

Word

Printable

Sprint:
2021 Week 16-18 (from Apr 19)
Steps to Reproduce:

Hide

The source code for the JMH benchmark used to obtain these results can be found here:
https://github.com/triceo/optaplanner-score-director-benchmark/tree/752

Attached are async profiler flamegraphs for all the runs.

Show
The source code for the JMH benchmark used to obtain these results can be found here: https://github.com/triceo/optaplanner-score-director-benchmark/tree/752 Attached are async profiler flamegraphs for all the runs.
Release Note Text:
Undefined
Docs QE Status:
NEW
QE Status:
NEW

There appears to be a performance regression that showed up between 7.52.0.Beta1 and Final, on the OptaPlanner Cloud Balancing example:

	Beta1	Final	Diff
CS-D	197763	181622	- 8 %
DRL	206996	205583	- 1 %*

At the same time, the Machine Reassignment experienced a significant improvement:

	Beta1	Final	Diff
CS-D	12005	14176	+ 18 %
DRL	11655	11956	+ 3 %

Numbers in operations per second, on otherwise quiet hardware.
Asterisk ( * ) indicates results within the margin of error of the benchmark.

Some possible clues:

is cloned by

RHDM-1710 Object hashing in fact handle leads to performance regression in OptaPlanner

relates to

PLANNER-2399 Evaluate improved CS-D accumulators