[ 
https://issues.apache.org/jira/browse/PIVOT-122?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12711114#action_12711114
 ] 

Noel Grandin commented on PIVOT-122:
------------------------------------

It may well be that you can find some stuff that can be speeded up, but outside 
of obvious problem, you've only got a couple of options

(a) something like Proguard, which strips away unnecessary stuff and optimises 
what is left

(b) triggering the class-loading as early as possible - I do this in one of my 
apps :
public static void main(String [] args) {
  // start loading these classes early because they take a long time
  new Thread() {
    public void run() {
        new ClassThatNeedsALongTimeToLoad();
    }
  }.start();

  .... everything else...

}

> Optimize WTK class loading
> --------------------------
>
>                 Key: PIVOT-122
>                 URL: https://issues.apache.org/jira/browse/PIVOT-122
>             Project: Pivot
>          Issue Type: Improvement
>          Components: wtk
>            Reporter: Todd Volkert
>            Assignee: Greg Brown
>             Fix For: 1.3
>
>
> Class initialization for pivot.wtk.<Class> often takes upwards of 300-500ms 
> (sometimes more).  This is because initializing one class in that package 
> often triggers initialization of many other classes, which is to be expected. 
> Nonetheless, 500ms seems extreme for class loading.
> This problem can be seen by creating a simple skeleton app that simply loads 
> a WTKX hierarchy and adds it to the display.  Using WTKX binding, you'll see 
> that the first call to bind() will take a long time (the aforementioned 
> 500ms), then an immediately subsequent call to bind() will take drastically 
> less, even though it's still instantiating the objects anew.  The difference 
> is that the first call to bind has to run WTK class initialization.
> This may be something we just live with, but it should at least be 
> investigated to see if we can't find a smoking gun.  The premise is that 
> compiled byte code running on an nGHz processor shouldn't take half a second 
> to run class initialization.  Informal tests show this time to be the last 
> vestige of slow load times for Pivot apps.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to