Re: [Jmol-developers] Jmol and Memory

[Nicolas V. <nve...@cl...>]

Miguel wrote:
>> browser               opened/Mb    +page/Mb    +1p84.pdb/Mb
>> MSIE6.0...              19            43            62
>> Firefox1.5.0.3          18            42            65-->47
>> Opera8.51               14            37            57
>>
>> So I would have to say that Opera is doing something right here, and
>> Firefox has some way of reducing its memory footprint that the other
>> browsers don't have.
>>     
>
> I am quite surprised to hear this. Applications don't usually return
> memory to the OS once they have allocated it.
>   
It's not the first time I see this with browsers.
I saw it when I also had memory usage problems for a web page with 
problems between JavaScript and ActiveX referencing each other.

> The Java plugin is definitely not returning memory to the OS.
>
>   
>> [Later: Opera and MSIE both slowly dropped down
>> about 10 Mb as well.]
>>     
>
> Seems very strange to me.
>
>   
>> Ah, now the behavior at http://fusion.stolaf.edu/chemistry/jmol/xtalx
>> (jmol 10.x) is quite different. In that case, loading up maleic.cif,
>> which has 18 crystals, and displaying the {1 1 1} unit cell produces an
>> interesting effect. Firefox is cycling on about a 20-second cycle
>> between 58 and 88 Mb memory usage, picking up about 2 Mb per second
>> until it gets to 88, then dropping again suddenly to 58 Mb. MSIE is just
>> sitting at 63 Mb; Opera is just sitting at 56 Mb. I wonder what in the
>> world is going on with Firefox? Note that this is whith repeating
>> setTimeout loop of 10 ms. When I change that to 100 ms, the cycling is
>> still there, just quite sow and not hardly to the same extent.
>>
>> Wow, here's something interesting: I do a getProperty auxiliaryInfo,
>> which gives me the information shown below for each of those 18 models,
>> and Firefox just shot up to 107 Mb and continues to climb to 115 Mb --
>> cycling now down to 50 Mb and climbing rapidly again.  What could that
>> be about?
>>     
>
> As I said in one of my messages yesterday, strange things must be
> happening when object references get passed from Java to JavaScript
> through the LiveConnect interface.
>   

Yes, probably.
For example, if references to JavaScript objects are passed to other 
technologies, then JavaScript garbage collector is totally lost.

>> Miguel, I'd be interested in knowing, if you can't even get past loading
>> the applet, what your maximum memory allocation is set at,
>>     
>
> I have never set it.
>
> Default on Linux for Java Plugin seems to be 64 Mb.
>   
The memory usage of the browser is probably not in the Java Plugin.
With JavaScript, last year I add a problem with the browser taking > 1 
Gb memory after letting it display a page for a day.

>   
>> or at least
>> what your process report says for the amount of memory used by Mozilla
>> right after the crash.
>>     
>
> It does not crash ... it just hangs. No report is generated.
>
>
> To help you out I just added some java memory usage statistics to the
> popup menu.
>
> Bring up the popup menu and bring up the About submenu.
>
> total = amount of memory requested from the operating system
> free = amount of free memory (always less than total)
> maximum = maximum total that Java will request.
>
> Note that a garbage collection is forced before the menu is displayed so
> that the 'free' number will be (basically) correct.
>   

Nice.

You have also removed the links to "Mouse Manual" and "I18n".

Nico