[rescue] SGI tidbits

[Joshua D Boyd]

On Sun, Sep 29, 2002 at 10:59:26AM -0400, Dave McGuire wrote:

> >But seriously, why on earth would it suck down 80MB?
> 
>   Much of Mozilla's user interface is compiled-in, run-time-interpreted 
> Javascript.  That seems (to me) to account for much of its unbelievably 
> shitty performance on even the fastest of machines.  It was a really, 
> really stupid decision.

That could be the reason for the slow speed.  But it doesn't seem right
to me.  After, Emacs isn't that bad, even on the newer more graphical
heavy versions.

But, more usefully, just look at the speed difference of sawfish versus
enlightenment.  Enlightment is slow, and all C/C++, where as sawfish is
run-time-interpreted, and reasonbly fast, even on older slower PCs.

And as to the 80 megs issue, I could say why for sure.  They are
probably holding at least 2 copies of every image on the page, but that
doesn't account for that much.  They could render the entire page to a
bitmap and hold it in memory, along with all the elements, but that also
doesn't seem to account for the memory used (but it would account to the
speed of scrolling versus the crawl of a horizontal resize.

I'd imagine that loading the source code to memory takes a lot of space
due to all the other bloat.

Basically, Mozilla probably doesn't waste all that memory in one spot.
It is probably wasted in lots of spots. Some of them are probably a good
place to spend memory (I think rendering the entire page to a bitmap is
a good idea for redrawing speed purposes), and others are probably bad
(like failing to deallocate resources from the X server properly, which
would be indicated if Mozilla's memory useage grows over time.  My
mozilla, which has been running for less than a day on my P2-350 (I
usually always leave it up, but had shut it down yesterday while I
fiddled with my new moniter), is currently using only 14megs of ram,
which seems kinda reasonable, assuming that it prerenders each page, and
considering that I have 9 graphics heavy pages open.

-- 
Joshua D. Boyd