Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site ecsvax.UUCP
Path: utzoo!watmath!clyde!bonnie!akgua!mcnc!ecsvax!bet
From: bet@ecsvax.UUCP (Bennett E. Todd III)
Newsgroups: net.graphics
Subject: Re: Mandelbrot programs posted to net.sources
Message-ID: <426@ecsvax.UUCP>
Date: Mon, 16-Sep-85 17:18:29 EDT
Article-I.D.: ecsvax.426
Posted: Mon Sep 16 17:18:29 1985
Date-Received: Wed, 18-Sep-85 05:42:00 EDT
References: <417@ecsvax.UUCP> <2197@ukma.UUCP>
Reply-To: duccpc!bet@ecsvax.UUCP (Bennett E. Todd III)
Distribution: net
Organization: Duke University Computation Center
Lines: 56

In article <2197@ukma.UUCP> sean@ukma.UUCP (Sean Casey) writes:
>Could you post some good sample values for input?

Sorry 'bout that -- an obvious omission. Here are the headers from the
files I have generated so far:

Centered at -0.240000+0.000000i, scale 1.757800+1.528500i, 320x200
Centered at -0.720000+0.250000i, scale 0.019531+0.016984i, 320x200
Centered at -0.915000+0.270000i, scale 0.019531+0.016984i, 320x200
Centered at -1.410000+0.000000i, scale 0.019531+0.016984i, 320x200
Centered at -0.160000+1.030000i, scale 0.019531+0.016984i, 320x200
Centered at -0.550000+0.640000i, scale 0.019531+0.016984i, 320x200
Centered at -0.760000+0.110000i, scale 0.019531+0.016984i, 320x200
Centered at -0.240000+0.000000i, scale 1.757800+1.171866i, 1008x336
Centered at -0.760000+0.110000i, scale 0.025476+0.016984i, 1008x336
Centered at -0.720000+0.250000i, scale 0.025476+0.016984i, 1008x336
Centered at -0.915000+0.270000i, scale 0.025476+0.016984i, 1008x336
Centered at -1.410000+0.000000i, scale 0.025476+0.016984i, 1008x336
Centered at -0.160000+1.030000i, scale 0.025476+0.016984i, 1008x336
Centered at -0.550000+0.640000i, scale 0.025476+0.016984i, 1008x336

The 320x200 frames go nicely on the Color Graphics Adaptor; the 1008x336
come out about 4x6" on the Color Printer. The former take a few hours;
the latter take more (one that I timed took 14 hours). This is with
DeSmet C using the 8087 library; not optimal floating point performance
by any means. Also, it looks to me like the zoomed in pictures on the
Color Printer (90x zoom if I recall the numbers correctly) show pretty
bad graininess around the border of the set. Either the double-precision
8087 support is slightly bad, or 255 isn't enough iterations. I fear the
latter. Again, sorry about the omission. And to explain about the input
parameters again, the command

	compute name center=a+bi scale=c+di hpix=e vpix=f

creates a data file a little over hpix*vpix bytes long, for the picture
centered at a+bi, going from a-c to a+c in the real direction and b-d to
b+d in the imaginary, hpix pixels horizontally by vpix vertically. To
cut down on typing I added a "like=name" option, so after computing one
of the pictures you can get the scale, hpix, and vpix from it and just
specify a new center by

	compute newname center=a+bi like=oldname

As you can see, I should have put more effort into the algorithm and
less into the user interface:-).

Anybody have any concrete info on how many iterations are required to
properly resolve the set at a given zoom?

-Bennett
-- 

"Hypocrisy is the vaseline of social intercourse." (Who said that?)

Bennett Todd -- Duke Computation Center, Durham, NC 27706-7756; (919) 684-3695
UUCP: ...{decvax,seismo,philabs,ihnp4,akgua}!mcnc!ecsvax!duccpc!bet