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.
Summary: | Reduce the number of threads created by the RequestProcessor | ||
---|---|---|---|
Product: | platform | Reporter: | Petr Nejedly <pnejedly> |
Component: | -- Other -- | Assignee: | Petr Nejedly <pnejedly> |
Status: | VERIFIED FIXED | ||
Severity: | blocker | Keywords: | PERFORMANCE |
Priority: | P2 | ||
Version: | 3.x | ||
Hardware: | Sun | ||
OS: | Solaris | ||
Issue Type: | TASK | Exception Reporter: |
Description
Petr Nejedly
2002-01-30 15:16:44 UTC
It may also help people with solving some of the threading issues and help the Actions API with posting independent performers by implementing Annonymous RP (a generally paralelizing RP) It would also allow even greater adoption of private RPs in modules and reduce the usage of the shared RP.DEFAULT which slows down e.g. explorer folder expansions (used there at several places throughout the folder expansion process) Consider also: - do not call Thread.setPriority unless really necessary, it is a native call and may be expensive - do not create a stack trace for the "posting time" unless a system property is set explicitly asking for it; these stack traces are often unnecessary to diagnose a problem, and they slow down every RP request posting > do not call Thread.setPriority unless really necessary, it is a > native call and may be expensive Mmm, it takes about 3-4us on my machine, not that bad, but a simple wrapper caching the old priority and skipping setPriority call on 100:1 hit sped it up to 0.3-0.4us, so I'll do the old priority caching. > do not create a stack trace for the "posting time" unless a system > property is set explicitly asking for it; Till now the logic was opposite (keep traces by default), should I revert it now? BTW: it takes some tens of us for each request. Implemented in /cvs/openide/src/org/openide/util/RequestProcessor.java,v 1.55 Also implemented less confusing exception reporting. More than 3 weeks in the trunk builds without any significant problems. No more 22 RP in my VM ;-) Also exceptions reportring is more clear now. Verifying. |