On Mon, 10 Aug 2009 12:21:11 -0300 Gustavo Sverzut Barbieri
<barbi...@profusion.mobi> said:

> On Mon, Aug 10, 2009 at 6:56 AM, Enlightenment
> SVN<no-re...@enlightenment.org> wrote:
> > Log:
> >        * edje: Reduce sizeof (Edje_Calc_Params).
> >
> >        Note: It doesn't really impact edje memory foot print yet. But in
> >        the plan to do a computation cache inside edje, this structure
> >        will be used a lot (I am planning to do this feature at some point,
> >        but no ETA yet, and be reassured it will be optionnal so we can
> >        choose between CPU load or memory load).
> >
> >        Note: As I was looking for similar area of improvements,
> >        Edje_Part_Description could really use an union to reduce it's size,
> >        but as we load this structure directly from an Eet file, we need
> >        union in Eet first. And this should be part of a comming Edje file
> >        format break.
> 
> Better than union is to have the single part in one structure and
> specific bits in their own structure. Depending on how we do the Eet +
> union support, we may think on how to do it to cover this case as
> well.
> 
> I'd say instead of allocate memory and fill it, one could give a
> "type" value to user callback and then it would receive the correct
> Eet_Data_Descriptor for that subtype. So union would return the data
> descriptor with the same struct size for all types, while dynamic
> would check (switch/case) which one to use, and return the fields
> properly.
> 
> This is likely to reduce memory consumption a lot because we often
> have LOTS of rectangle that have almost no field, while we have very
> few TEXT/TEXTBLOCK/GRADIENT that consume most memory.
> 
> BTW, something that could improve memory there is using mempool.

even mempool will fragment. it isnt necessarily a silver bullet. not allocating
is better than anything else :) i am wondering if simple N structs with casting
is best (and a common header in all for common members for all types).

-- 
------------- Codito, ergo sum - "I code, therefore I am" --------------
The Rasterman (Carsten Haitzler)    ras...@rasterman.com


------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with 
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
enlightenment-devel mailing list
enlightenment-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/enlightenment-devel

Reply via email to