On Wed 27 Jan 2010 at 01:26PM, [email protected] wrote:
> On Wed, Jan 27, 2010 at 03:20:05PM -0600, Shawn Walker wrote:
> > On 01/27/10 03:11 PM, [email protected] wrote:
> > >Folks,
> > >This is a change the I inherited from Dan. It implements the flyweight
> > >patter and __slots__ for fmris, versions, and dotsequences. Using this
> > >fix we get a 10-15% reduction in RSS during install and no performance
> > >regression.
> > >
> > > http://cr.opensolaris.org/~johansen/webrev-11029/
> >
> > This seems right to me. Have you poked around in the GUI to see if
> > it still behaves as expected? It should at this point (with b130+ I
> > think?).
>
> I have not. I assumed that since 9519 had been fixed, the GUI wasn't
> touching internals that it shouldn't be. This code passes the test
> suite and can install/uninstall.
>
> -j
J-- looking back at this webrev, I am reminded that one thing
which might be worth looking at is whether flyweight would help
if applied to the Version object itself-- perhaps minus the
timestamp.
That is to say, we probably have a lot of instances of Version
which are: 0.5.11,5.11-0.130:<something> and flyweighting up one level
(instead of or in addition too what you have now) might be at least
worth trying. Sorry I forgot to mention this yesterday, I had forgot
that I was planning to look into this.
-dp
--
Daniel Price, Solaris Kernel Engineering http://blogs.sun.com/dp
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
