[Laszlo-dev] I am getting 60% cpu usage in swf9/firefox/osx for the test case in LPP-8072 in 4.2

[P T Withington]

It seems I can duplicate your result in clean trunk, too.  I  
apparently had trimmed down the size of the XML file for testing.  Now  
that I put it back to the original size, I am seeing the 60% when I  
put the mouse over the player.

Next, I'm going to try turning off all mouseEnabled by hand and see if  
I still get that...

On 2009-04-28, at 07:13EDT, P T Withington wrote:

> I'm going to see if I can replicate your results.
>
> You _do_ have to wait a while for the instantiator to finish, before  
> you make your cpu observations.
>
> One possibility is that 4.2 is missing other optimizations that are  
> swamping your changes.
>
> On 2009-04-27, at 19:47EDT, Henry Minsky wrote:
>
>> So maybe this is a bad test case; I looked at swfdump of that  
>> games.swf
>> resource, and  it's a fairly complex vector graphic. So much of the  
>> CPU
>> usage we see will probably befrom the graphics intensive redraw  
>> every frame.
>> Perhaps a better test would be views with just simple backgrounds  
>> or bitmaps
>> or something. Or
>> turn on bitmap caching on the resource.
>>
>>
>>
>> On Mon, Apr 27, 2009 at 7:40 PM, Henry Minsky <hminsky at laszlosystems.com 
>> >wrote:
>>
>>> Well, actually the CPU usage behavior I see seems to be as follows
>>> + If I select the test app, and leave the mouse on top of a soccer  
>>> ball
>>> image, the CPU stays pegged
>>> at 60% forever.
>>>
>>> + If I command-tab to another OSX  process, the CPU usage goes  
>>> down to 15%
>>> or so
>>>
>>> + If I then click back anywhere in the browser window to give it  
>>> focus,
>>> the CPU still stays at 15%, but if I click the mouse
>>> on the app again, it's like it activates the Flash player and CPU  
>>> goes up
>>> to 60% again.
>>>
>>> + If I leave the mouse over the blank <window> or the debugger  
>>> window,
>>> rather than the tiled image array,
>>> CPU usage is not as high, more like
>>> 20% or so.
>>>
>>> As an experiment I modified the swf9 kernel to force  
>>> mouseEnabled=false on
>>> every sprite, but don't see any measureable difference in CPU  
>>> usage in that
>>> case :-(
>>>
>>>
>>>
>>>
>>>
>>> On Mon, Apr 27, 2009 at 7:07 PM, Henry Minsky <hminsky at laszlosystems.com 
>>> >wrote:
>>>
>>>> Never mind, I think I was not waiting long enough to let the  
>>>> thing settle
>>>> down.
>>>>
>>>>
>>>> On Mon, Apr 27, 2009 at 7:01 PM, Henry Minsky <hminsky at laszlosystems.com 
>>>> >wrote:
>>>>
>>>>>
>>>>> So even with a clean 4.2 branch, if I run the test case in  
>>>>> LPP-8072, with
>>>>> the wrapping layout with 1100 images and text views, I always  
>>>>> get 60%
>>>>> cpu usage. It never drops below that when the app has focus. I  
>>>>> have the
>>>>> thing in non-debug mode, with no
>>>>> dev-console. I even modified the swf9 kernel so that  
>>>>> mouseEnabled was
>>>>> always false on every sprite, but it doesn't seem to make
>>>>> any difference, the CPU usage stays pegged at 60%. Strangely, if  
>>>>> I am
>>>>> running with the debugger, and I move the mouse over
>>>>> the debugger window, CPU usage goes down to 22%.If I select the
>>>>> debugger's input field for focus, the CPU goes down to 20%.
>>>>>
>>>>> But leaving the mouse over the tiled graphics shoots up to 60%.
>>>>>
>>>>> I thought your testing showed a lot lower CPU usage..
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Henry Minsky
>>>>> Software Architect
>>>>> hminsky at laszlosystems.com
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Henry Minsky
>>>> Software Architect
>>>> hminsky at laszlosystems.com
>>>>
>>>>
>>>>
>>>
>>>
>>> --
>>> Henry Minsky
>>> Software Architect
>>> hminsky at laszlosystems.com
>>>
>>>
>>>
>>
>>
>> -- 
>> Henry Minsky
>> Software Architect
>> hminsky at laszlosystems.com
>