Hi all, my question is about the enormous varianz in data throughput and the resulting duration of TPTP when profiling JUnit Tests. In more detail:
I use TPTP to create dynamic traces of a JUnit TestRun (probably somebody remember my questions regarding traceModel some months ago). Now i created some performance statistics to convince my professor. The figures are like: how long does it take to trace a complete TestSuite, how long does it take to trace exactly one class ... and so on. My Obervation is: The data througput underliles an enormous varianz. To rule out hardware reasons i repeated the same tests several times with (nearly) the same finding. For example i profiled org.apache.commons.beantutils and org.apache.commons.codec (current trunk). The figures are: Complete Suite (19 TestClasses): Processed: 7976174 bytes, bytes/second: 725106, events processed: 68126, events/second: 6193, elapsed time: 11 seconds - WrapDynaBeanTestCase Processed: 243065 bytes, bytes/second: 48613, events processed: 2092, events/second: 418, elapsed time: 5 seconds - LongConverterTestCase Processed: 28236 bytes, bytes/second: 14118, events processed: 252, events/second: 126, elapsed time: 2 seconds - DoubleMetaphoneTest Processed: 49416440 bytes, bytes/second: 1764872, events processed: 420642, events/second: 15022, elapsed time: 28 seconds - QuotedPrintableCodec Processed: 71720 bytes, bytes/second: 35860, events processed: 626, events/second: 313, elapsed time: 2 seconds As you see the results vary from 126 evens/seconds to 15022 events/seconds ... wow?! Furthermore it is interersting that profiling the whole suite which contains 19 Testclasses is profiled in 11 seconds while profiling one single class WrapDynaBeanTestCase lasts 5 seconds. Is that only due to tptp overhead (starting and maintaing data collectors, etc.)? My environment is: - Windows XP Prof - Sun Jdk 1.5.0_12-b04 - Eclipse 3.3.1.1 M20071023-1652 - TPTP-4.4.0.3-200709261752 Any help and ideas are highly appreciated, Cheers, Thomas E.-E. P.S: do you know which class produces the log output (which i used above) and which elapsed time is measured there? I tried to measure the time myself and had no chance to get similar values. Is it from creating the Launch-Object till Termination of the process? _______________________________________________ tptp-tracing-profiling-tools-dev mailing list tptp-tracing-profiling-tools-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/tptp-tracing-profiling-tools-dev