Did you use default ('client') harmony mode? It's about 10-20% slower then
-Xem:server one and tuned for fast application startup. I hope that someday
we will merge all our configs and have only 1 mode to work to avoid these
confusions.On 1/25/07, Leo Li <[EMAIL PROTECTED]> wrote:
Sorry, missing one thing. The numbers in the list are mseconds to run tests.:) On 1/25/07, Leo Li <[EMAIL PROTECTED]> wrote: > > Hi, all: > I have just ran the benchmark tests provided by the dacapo. Although > I just got the result from my machine with default heap size , it > really suprises me: Harmony at least has similar performance output as RI. > > Environment: windows XP sp2 > Result: > > > > > RI > > Harmony > > *antrl* > > 12235 > > 13000 > > *bloat* > > 21890 > > Failure: exception thrown out > > *chart* > > 20672 > > 17672 > > *eclipse* > > 98829 > > 74610 > > *fop* > > 4422 > > 4578 > > *hsqldb* > > Failure: Out of Memory > > 9250 > > *jython* > > 17969 > > Failure: maybe Harmony has different layout from RI > > *lusearch* > > 17140 > > 13657 > > *luindex* > > 28687 > > 24797 > > *pmd* > > 15129 > > 12312 > > *xalan* > > 28563 > > Failure: with xalan version conflicts > > (The failures on Harmony might be due to configuration.) > > Further, I still have some problem: > 1. Are there performace test suites focused on io,net and nio, where I > suspect the space to improve exists. > 2. Is there performance tests that can provide profiling messages to help > to diagnose hotspot? > > Thanks, > Leo > > > > On 1/25/07, Leo Li <[EMAIL PROTECTED]> wrote: > > > > On 1/25/07, Geir Magnusson Jr. <[EMAIL PROTECTED]> wrote: > > > > > > > > On Jan 24, 2007, at 11:00 PM, Robin Garner wrote: > > > > > > [SNIP] > > > > > > > http://cs.anu.edu.au/people/Robin.Garner/dacapo/regression/ - My > > > > focus is on DaCapo correctness, the performance test isn't > > > > particularly rigorous. > > > > > > > > IMO for harmony tracking SPECjvm98, SPECjbb2005 and dacapo would be > > > > a good start. > > > > > > > > One thing you definitely should do in a dedicated Harmony > > > > performance test is test across a range of heap sizes. In a small > > > > heap GC improvements and allocation efficiency are more important - > > > > in large heaps locality effects and code quality dominate. > > > > > > > > If possible, testing on a range of machines would be good too. > > > > > > I'm hoping that by just making this "plug-n-play" in our build-test > > > CI and reporting system, we'll be running in many places in no time... > > > > > > > > +1 > > > > Add performance testing to auto build system will be a great help for us > > to monitor the effect of the modification in source code. > > > > geir > > > > > > > > > > > > -- > > Leo Li > > China Software Development Lab, IBM > > > > > > -- > Leo Li > China Software Development Lab, IBM > -- Leo Li China Software Development Lab, IBM
-- Mikhail Fursov
