[squeak-dev] Squeak 3.9 alpha image wanted

> Hello,
>
>
> can anyone tell me were to find the earliest available 3.9 alpha image?
>
> I want to identify the cause of the slowdown of browsers when you have
> many of them opened.
>
>   http://article.gmane.org/gmane.comp.lang.smalltalk.squeak.general/108359
>
> I am using the benchmark from Andreas. So far I see the problem in the
> current squeak-dev 3.9.1, Squeak3.9b-7029.image. Browser speed is OK
> in Squeak3.8-6665.image (Squeak3.8-6665-basic.zip).
>
> Cheers
>
> Matthias
>
> PS: Benchmark results on my machine:
> "Run this several times, do not close browser windows."
> win := Browser openBrowser topView.
> win extent: 600@350.
> br := win model.
> br setClassOrganizer.
> br systemCategoryListIndex: 2.
> World doOneCycle.
>
> Transcript cr; show: [1 to: 10 do:[:i|
>        br classListIndex: i.
>        World doOneCycle.
>  ]] timeToRun.
>
>
>
> sq3.9.1-7075dev08.04.1.image
> 1880
> 1591
> 1420
> 1860
> 2550
> 3341
> 3262
> 3821
> 4396
> 4802
>
> Squeak3.9-final-7067.image  07-Nov-2006 11:07
> 1001
> 1410
> 1842
> 2353
> 2794
> 3273
> 3809
> 4440
>
> Squeak3.9b-7029.image  04-May-2006 20:54
> 1161
> 1793
> 2421
> 3063
> 3565
> 4004
> 4556
> 4987
>
>
> Squeak3.8-6665.image Squeak3.8-6665-basic.zip   15-Jun-2005
> 942
> 918
> 930
> 945
> 975
> 985
> 1040
> 1076
>
>

>
> Matthias Berth wrote:
>> Hello,
>>
>>
>> can anyone tell me were to find the earliest available 3.9 alpha image?
>>
>> I want to identify the cause of the slowdown of browsers when you have
>> many of them opened.
>>
>>  
>> http://article.gmane.org/gmane.comp.lang.smalltalk.squeak.general/108359
>>
>> I am using the benchmark from Andreas. So far I see the problem in the
>> current squeak-dev 3.9.1, Squeak3.9b-7029.image. Browser speed is OK
>> in Squeak3.8-6665.image (Squeak3.8-6665-basic.zip).
>>
>> Cheers
>>
>> Matthias
>>

> nice wrote:
>
> > Even a simple
> >
> >    MessageTally spyOn: [Transcript cr; show:'hello'].
> >
> > exhibits a factor 2x slowdown between 3.8 and 3.10
> >
> > Nicolas
> >
> >
>
>  And the main difference lies in the drawing of the container, not the
> contents as you can test it:
>
>  Smalltalk garbageCollect.
>  theTranscriptWindow := SystemWindow allInstances detect: [:e | e label =
> 'Transcript']. "(I know, bad code, but cannot remember correct API)"
>
>  Time millisecondsToRun: [theTranscriptWindow drawDropShadowOn: World
> canvas].
>         => 4 36
>  Time millisecondsToRun: [World canvas drawMorph: theTranscriptWindow ].
>         => 4 21
>  Time millisecondsToRun: [theTranscriptWindow drawSubmorphsOn: World
> canvas].
>         => 18 11
>
>  Scores are given in 3.8 and 3.9/3.10 (where they are dependent on
> theTranscriptWindow bounds extent).
>  Notice the transparency and background gradient in 3.9/3.10...
>
>  Don't know how this connects to number of opened Browsers seriously
> impacting speed, but maybe that's a clue...
>
>
>
>  Nicolas
>
>
> >
> > Matthias Berth wrote:
> >
> > > Hello,
> > >
> > >
> > > can anyone tell me were to find the earliest available 3.9 alpha image?
> > >
> > > I want to identify the cause of the slowdown of browsers when you have
> > > many of them opened.
> > >
> > >
> http://article.gmane.org/gmane.comp.lang.smalltalk.squeak.general/108359
> > >
> > > I am using the benchmark from Andreas. So far I see the problem in the
> > > current squeak-dev 3.9.1, Squeak3.9b-7029.image. Browser speed is OK
> > > in Squeak3.8-6665.image (Squeak3.8-6665-basic.zip).
> > >
> > > Cheers
> > >
> > > Matthias
> > >
> > >
> >
>
>
>
>

>   http://article.gmane.org/gmane.comp.lang.smalltalk.squeak.general/108359
>
> I am using the benchmark from Andreas. So far I see the problem in the
> current squeak-dev 3.9.1, Squeak3.9b-7029.image. Browser speed is OK
> in Squeak3.8-6665.image (Squeak3.8-6665-basic.zip).
>
> Cheers
>
> Matthias
>
> PS: Benchmark results on my machine:
> "Run this several times, do not close browser windows."
> win := Browser openBrowser topView.
> win extent: 600@350.
> br := win model.
> br setClassOrganizer.
> br systemCategoryListIndex: 2.
> World doOneCycle.
>
> Transcript cr; show: [1 to: 10 do:[:i|
>        br classListIndex: i.
>        World doOneCycle.
>  ]] timeToRun.
>
>
>
> sq3.9.1-7075dev08.04.1.image
> 1880
> 1591
> 1420
> 1860
> 2550
> 3341
> 3262
> 3821
> 4396
> 4802
>
> Squeak3.9-final-7067.image  07-Nov-2006 11:07
> 1001
> 1410
> 1842
> 2353
> 2794
> 3273
> 3809
> 4440
>
> Squeak3.9b-7029.image  04-May-2006 20:54
> 1161
> 1793
> 2421
> 3063
> 3565
> 4004
> 4556
> 4987
>
>
> Squeak3.8-6665.image Squeak3.8-6665-basic.zip   15-Jun-2005
> 942
> 918
> 930
> 945
> 975
> 985
> 1040
> 1076
>
>
>  

> Matthias Berth wrote:
>> Hello,
>>
>>
>> can anyone tell me were to find the earliest available 3.9 alpha image?
>>
>> I want to identify the cause of the slowdown of browsers when you have
>> many of them opened.
>>  
>
> Much of the slowdown is the pane resize handles from the look
> enhancement. There is arguably a bug in Morph>>privateAddMorph: that
> sometime adds submorphs to the top left corner of the screen instead of
> to the
> top left corner of the owner morph. This causes massive screen redraws.
> Also the adding of the pane resizes causes
> whole window redraws more frequently than before they were added.
>
> Karl

> Matthias Berth wrote:
>> Hello,
>>
>>
>> can anyone tell me were to find the earliest available 3.9 alpha image?
>>
>> I want to identify the cause of the slowdown of browsers when you have
>> many of them opened.
>>  
>
> Much of the slowdown is the pane resize handles from the look
> enhancement. There is arguably a bug in Morph>>privateAddMorph: that
> sometime adds submorphs to the top left corner of the screen instead of
> to the
> top left corner of the owner morph. This causes massive screen redraws.
> Also the adding of the pane resizes causes
> whole window redraws more frequently than before they were added.
>
> Karl

Here is a workaround to I made a while back. I'm not sure how much