The main thing that changed since the last measurement was making the
core classes into sealed classes, and then wrapping them with a
dynamic shell class
LzNode, LzView, etc.
On Tue, Feb 12, 2008 at 5:02 PM, P T Withington <[EMAIL PROTECTED]> wrote:
> Phew!
>
>
>
> On 2008-02-12, at 16:40 EST, Henry Minsky wrote:
>
> > Oops, right, the swf9 should be the faster one.
> >
> > On Tue, Feb 12, 2008 at 4:16 PM, Donald Anderson
> > <[EMAIL PROTECTED]> wrote:
> >> Henry,
> >>
> >> I hope you have the headings reversed - looks like SWF9 takes
> >> longer...
> >>
> >>
> >>
> >> On Feb 12, 2008, at 4:10 PM, Henry Minsky wrote:
> >>
> >>> So, working around that compiler bug
> >>> <canvas width="1000" height="600">
> >>> <handler name="oninit">
> >>> var i;
> >>> i = 1000;
> >>> var starttime;
> >>> starttime = getTimer();
> >>> while (i > 0) {
> >>> i--;
> >>> new lz.view(canvas, {x: i, y: i});
> >>> }
> >>>
> >>> new lz.text(canvas,{x:0, y:20, text: 'elapsed time:
> >>> '+getTimer()});
> >>> var ctime;
> >>> ctime = getTimer() - starttime;
> >>> new lz.text(canvas,{x:0, y:30, text: ('initialize new views
> >>> time: '+ctime )} );
> >>> </handler>
> >>> </canvas>
> >>>
> >>>
> >>>
> >>> SWF8
> >>> total elapsed time 137 msec
> >>> creating views: 97 msec
> >>>
> >>> SWF9
> >>> total elapsed time 450 msec
> >>> creating views: 270 msec
> >>>
> >>> So it looks like a speedup of about three from swf8 to swf9 for view
> >>> creation.
> >>>
> >>>
> >>>
> >>>
> >>> --
> >>> Henry Minsky
> >>> Software Architect
> >>> [EMAIL PROTECTED]
> >>
> >>
> >> --
> >>
> >> Don Anderson
> >> Java/C/C++, Berkeley DB, systems consultant
> >>
> >> voice: 617-547-7881
> >> email: [EMAIL PROTECTED]
> >> www: http://www.ddanderson.com
> >>
> >>
> >>
> >>
> >>
> >
> >
> >
> > --
> > Henry Minsky
> > Software Architect
> > [EMAIL PROTECTED]
>
>
--
Henry Minsky
Software Architect
[EMAIL PROTECTED]
