🚀 go-pugleaf

bbum@object.com (Bill Bumgarner) writes:
>I would like to see any machine that doesn't dog out when running a
>couple of compiles.

We have quite a few that can do it. None have the brand name NeXT
though. Some are considerably more expensive, some aren't.

>As far as text scrolling is concerned;  scrolling text within a set of 32-bit 
>rectangles of screen space (which is what scrolling text really is-- usually  
>you are only scrolling a single rectangle, though often the windowmanager is  
>forced to scroll a piece of text through multiple, odd-shaped, discontiguous  
>hunks of screen space) is a nasty performance hit as well... If the entire  
>contents of whatever is being scrolled is cached on the ND, it shouldn't be a  
>big deal.  But if the image has to be moved across the backplane (either as a  
>straightforward bitmap or as a set of graphics primitives), then you get nailed  
>on the lack of throughput through the backplane... which isn't terribly high.
>

Yes you're quite correct about the memory bandwidth problems of
scrolling but other systems don't seem to suffer as much as the NeXT.
Is the DPS interpreter? The blitting code on the ND's i860? Or what?
Too many software layers?

I once did a window system port (X11) for a custom 24 bit (really 32)
display we built. The display was connected to custom computer we
built and hosted on a SparcStation IPC (memory i/f across the SBus to
another custom bus) and although its scrolling performance was nothing
to write home about (I did very little optimisation as it was only a
short lived prototype) it didn't suffer as much as the NeXT.

Don't misunderstand me .... I love these NeXT machines (I consider
them to be the most useful machine that comes straight out of the box,
or they were when you got lots of s/w bundled with them) - they're a
nice mix of the Mac, a UNIX workstation and the beginnings of a decent
distributed computing environment.
-- 
Andy Newman (andy@research.canon.oz.au)