This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
I have measured startup time by our performance tests from qa module (startup time and memory consumption) on Sol 8, UltraSparc60, RAM=1GB, JDK1.4.2_02 Results for [nb winsys](031104) first start 50612 49634 second start 42339 38230 third start 40514 37501 results for [nb dev](200311041900) first start 32307 32443 second start 26320 26820 third start 27591 25523 As you can see build from winsys branch is about 50% longer than the build from trunk (built on sources from same date). By the way it seems like memory consumtion lowers ~122MB(trunk) vs. ~96MB(winsys) .
Created attachment 12074 [details] ide.log file with performance logs from WINSYS build
Created attachment 12075 [details] ide.log file with performance logs from TRUNK build
Please provide the method you used for the measuring. Using the same method for trunk and branch doesn't mean you are measuring comparable things.
It seems like Peter is right, but "feel" of the startup is the same, so reassigne back.
OK, first I solve the thing, currently we load all TopComponent on startup, even when not opened... then we will see whether it was the main problem.
Marian, there must be something really screwed with your enviroment, as you got different results even in _completly_ unrelated places that virtually can't be influenced by the WS changes, like splash screen show time and even top-logging initialization. Moreover the changes shouldn't that much influence (drastically reduce) used memory, so I suspect you had problems with free memory (or the lack thereof). Now, I've tried it on my machine (OK, different architecture, notebook P4@2GHz) and I got different results, with well-limited scope of changes in the log. I performed log.startup as well as manual wall-time measurement. The start log measurements are similar to yours - 6.6s vs 9.7s, but the difference in real time between starting the application and seeing the main window is way different: 10.8s vs 12.3 (14%) Explanation: I realized (when printing the log to the console) that there is very big delay between last log entry (Preparation finished) and the time when I really see the main window. The old WS used a lot of deferred performers and similar beasts that were actually exected as hooks asynchronously th the main (measuring) thread, even long after the main thread finished. This is why I resorted to using stopwatch... Now the real differences in logged parts: *) load lookup cache finished, took 614ms vs 1001ms - this seems to me more like some fix-only-in-trunk thing *) *debugger* (i.e. webdebug, dcore, djpda) ModuleInstalls - this may be caused by some ws changes, 950 vs. 1200ms total *) Main window opened dT=1234 vs 3526ms - different strategy, don't necessarily mean any change for user.
Fixed in branch [winsys_29836] Now there should be loaded only opened TopComponents during startup. core/windows/../Central.java 1.1.2.115 ModeImpl.java 1.1.2.51 PersistenceHanlder.java 1.1.2.51 TopComponentGroupImpl.java 1.1.2.10 WindowManagerImpl.java 1.1.2.112 model/DefaultModeModel.java 1.1.2.25 DefaultModel.java 1.1.2.54 DefaultTopComponentGroupModel.java 1.1.2.8 ModeModel.java 1.1.2.21 Model.java 1.1.2.45 TopComponentGroupModel.java 1.1.2.7 TopComponentSubModel.java 1.1.2.11
verified in [nb_dev](200311261900) Tonda is working on the patch for NB3.5 now, but he had did some measurement and it seems like we are faster, good job guys ;)