Re: [CZT-Devel] memory optimisation - comments please...

[Leo F. <leo...@ne...>]

Hi Tim,

The Netbeans profiler is actually quite powerful and easy to use. There is even a OO-query (SQL like) language to analyse the heap dump (!!!)

Thanks for the suggestion - you were right: I managed to reduce the ArrayList/Object[] memory consumption from 24% to 16%
(e.g., in real terms from 105MB to 80MB - some good results as well).

And the changes weren't that significant: mostly at TermImpl and GNast templates for corejava + a few spots here and there. 

I am quite happy with the results, and the usability within Eclipse is now much much better / smooth for some of the larger specs I am handling
(e.g., bare in mind I also had in memory the proof state and VCG data /info )!!! 

Thanks for suggesting this :-) 

And others, please let me know your experiences - if it improve / deteriorate  etc....

Best,
Leo

On 25 Jan 2012, at 09:00, Tim Miller wrote:

> Hi Leo,
> 
> Thanks for looking into all of this. It is some useful analysis.
> 
> I have tried to use the getAnns() method without creating new references 
> to it. I violate this only to improve readability. I don't think it is 
> causing to many problems though because the annotation lists persist in 
> memory as long as the AST exists, so references to these lists elsewhere 
> won't have much of an effect.
> 
> One possibility to help with reducing the annotation list memory 
> consumption is to modify the gnast code such that the annotation list 
> for each term is initialised to null, and then only created the first 
> time the getAnns() method is called. I'd wager that for a lot of terms, 
> it is never called.
> 
> And regarding your previous email, I definitely think a 2-3% increase in 
> CPU use is a good trade off for halving memory usage. In the parser and 
> typechecker, we can always wait a bit longer for the result, but once 
> we're out of memory, there's nowhere to go!
> 
> Cheers,
> Tim
> 
> On 25/01/12 17:47, Leo Freitas wrote:
>> Hi guys,
>> 
>> I think I found another source of some of the retained instances that seem to have weird behaviour.
>> 
>> Term has a List<Object>  getAnns() method, which exposes information like LocAnn and TypeAnn and others.
>> It's called in 1112 places across 17 CZT sub-projects: that exposes the ArrayList / Object[] for retention from those
>> that use the result or keep it in maps or the like.
>> 
>> Having said that, I couldn't find a place of misuse in the scattered references that I searched for (e.g., this might
>> be just another red herring or a key point for modification/memory-performance-improvement).....
>> 
>> thoughts / comments / suggestions ?
>> 
>> Leo
>> ------------------------------------------------------------------------------
>> Keep Your Developer Skills Current with LearnDevNow!
>> The most comprehensive online learning library for Microsoft developers
>> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
>> Metro Style Apps, more. Free future releases when you subscribe now!
>> http://p.sf.net/sfu/learndevnow-d2d
>> _______________________________________________
>> CZT-Devel mailing list
>> CZT...@li...
>> https://lists.sourceforge.net/lists/listinfo/czt-devel
>> 
> 
> 
> ------------------------------------------------------------------------------
> Keep Your Developer Skills Current with LearnDevNow!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnow-d2d
> _______________________________________________
> CZT-Devel mailing list
> CZT...@li...
> https://lists.sourceforge.net/lists/listinfo/czt-devel