Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!dogie.macc.wisc.edu!uwvax!astroatc!nicmad!madnix!perry
From: perry@madnix.UUCP (Perry Kivolowitz)
Newsgroups: comp.sys.amiga
Subject: Re: The Art Department and Memory
Message-ID: <1640@madnix.UUCP>
Date: 25 Nov 90 17:21:12 GMT
References: <5652@crash.cts.com>
Reply-To: perry@madnix.UUCP (Perry Kivolowitz)
Organization: ASDG Incorporated
Lines: 50

In article <5652@crash.cts.com> seanc@pro-party.cts.com (Sean Cunningham) writes:
>In-Reply-To: message from perry@madnix.UUCP
>
> 
>You brought up something I was interested in.  Working with a larger file on
>disk.
> 
>Even if you can't now, do you think TAD will ever let you work with images
>that would be just to large for RAM.  Like some scanned in photos for a
>magazine spread or something, where they're sometimes 200MB or more?  

Actually, we have been working on a uniform and transparent virtual memory
model (program specific, not system wide) for more than a year. This work
will see its first commercial release in our Dupont 4Cast driver module 
for ADPro. 

A print made with the Dupont 4Cast may consume as much as 80MBytes of data
so VM was essential. Because we had been working on this aspect for over a
year, implementing the 4Cast driver was not as laborious as people on other
platforms have found it. 

With the passage of time, and more work, some future software will employ
still greater levels of VM integration. So far we have achieved our goals
of designing an extremely flexible VM architecture which impacts performance
as little as possible. Some performance hit is unavoidable though, with the
conflict between time and space being what it is and all that.

But, you know, being able to manipulate massive images is a mixed blessing.
Sure, it is nice to be able to do it. But consider that each of the possibly
hundreds of millions of pixels needs at least a little attention from the 
CPU during processing. Doing anything hundreds of millions of times takes
time. Doing something hundreds of millions of times from disk, takes a little
more time. So while VM gives you the ability to work with massive images,
you may find practical reasons why you wouldn't want to.

Currently, TAD and ADPro give you the fastest image processing available
on the Amiga. Some of our speed advantage is attributable to being 
RAM-based. But, most of our speed advantage comes from coding. For 
example, its our coding that allows ADPro's PostScript saver to convert 
an image to PostScript as much as 100 times faster than ProPage.

Summary: VM is great, its coming, but RAM prices are dropping more 
quickly than faster CPU boards (which make using massive images practical)
are becoming available.

PK
-- 
Perry Kivolowitz, ASDG Inc. ``We look for things. Things that make us go.''
	UUCP:  {harvard|rutgers|ucbvax}!uwvax!astroatc!nicmad!madnix!perry
	CIS:   76004,1765 PLINK: pk-asdg