| Re: [rmf-dev] performance testing |
|
I've pushed two new change request: ReqIF10Util: added tests for setting values, factored out common code. ReqIF10UtilTest: added benchmarking using JUnitBenchmarks. The first one is just about additional tests to ReqIF10UtilTest. This should be trivial to approve and merge. The second one depends on the first and adds benchmarking using JUnitBenchmarks. If you find this useful, you may want to merge it, but I we may want to hold it. This is because JUnitBenchmarks is still in development, and also because I had to patch it to make it work with OSGi, as I explained in the gerrit message: ReqIF10Util has recently been changed to use EMF reflective api. To get an idea of how this affects performance, the ReqIF10UtilTest is now benchmarked using the JUnitBenchmarks utility library residing at: https://github.com/carrotsearch/junit-benchmarks However due to class loading issues when running within OSGi, a fix is required to JUnitBenchmarks so it can properly display timing results. The fix can be found at: https://github.com/fmjrey/junit-benchmarks/tree/osgi The runtime jar included in this plugin contains the fix. While it makes basic usage possible, BenchmarkOptions annotations are still not working, therefore the following JVM arguments may be useful for those wanting to specify global values: -Djub.rounds.benchmark=500 -Djub.rounds.warmup=5 In the end, using such setup, I was able to see no major impact on performance with the recent changes I've made to ReqIF10Util. Now off to presentation stuff... Cheers, François |