Re: [Jmol-users] memory allocation

[Timothy D. <mol...@ma...>]

On Dec 13, 2007, at 12:30 PM, Craig T Martin wrote:

> While we're talking memory allocation, and with respect to:
>
> "And unfortunately the memory of applets whose window/tab was  
> closed it not always release "
>
> For the stuff that I'm doing, I've known for some time that after  
> loading (and closing) various windows, Jmol eventually stops  
> working (Jmol, not the browser, hangs on a subsequent load) and I  
> have to restart the browser. This on a Mac running either Safari or  
> FireFox (latest everything, but it's been happening for more than a  
> year). If I look at the memory allocated to Safari, for example, I  
> can see a step increase on every new Jmol applet. So it is clearly  
> not releasing memory.
>
I have investigated this a bit, since I do a lot of development under  
OSX/Safari, and it is not clear to me what is going on.

Safari declares a virtual memory size according to the JVM max  
allocation. in my case, allocating 1024 mb to the JVM, Safari gets  
1.6g of vmem when it first loads an applet (even an empty applet).   
of course, this is unrelated to physical memory - Safari generally  
starts with approx 225mb of physmem on my machine.

loading 4 different 1mb (uncompressed) pdb files into 4 different  
applets on a single page increases Safari's physmem usage by 12 mb -  
that's 3x the actual combined pdb file sizes. I've been able to  
repeat this multiple times with different files.

the 'extra' physmem remains allocated to Safari as long as the window  
with the applets is open (obvious).  at least half of it, sometimes  
all of it, remains even after the window is closed.  closing the  
window and clearing the browser cache sometimes causes Safari to drop  
back down to its initial 225 mb of physmem; sometimes it does not.   
(during all of this time, the vmem does not change from 1.6 gb.)

I found this on the Web, and it looks like it could be helpful, so  
I'll pass it along to those wiser than myself:

<http://wrapper.tanukisoftware.org/doc/english/properties.html>

(fair warning: while the link was working 5 minutes ago, it is  
currently offline <shrug>)


>

not sure if I helped or not. :-)

tim
-- 
Timothy Driscoll                                em: mo...@vb...
Virginia Bioinformatics Institute               ph: 540-231-3007
Bioinformatics I: M-1                           im: molvisions
Washington St., Blacksburg, VA 24061

04-16-07.  We will not forget you.