RE: [Gcl-devel] Mandelbrot for the terminal behaves differentlywithGCL and ABCL

[Mike Thomas]

Hi again.

| I've was on leave yesterday and am catching up on other stuff so
| can't look
| further into this at the moment.  I have CLISP, LW Personal and
| Corman here
| so hopefully will have some basis on which to work when I get back to your
| email.

I ran Edi's first mandel function on CLISP, LWP, Corman and GCL compiled and
interpreted, all on Windows.

CLISP and Corman agreed with each other.

The other three disagreed with all others - LWP quite radically so, and
including the fact that GCL compiled gave slightly different results to GCL
interpreted.

Without going to the trouble of printing out all of the numbers, I would put
all of these results down to roundoff and/or precision differences which, in
iterative numerical algorithms, can alter the iterations required to meet a
given termination condition.

Cheers

Mike Thomas

(defun mandel ()
    (loop for y from -1 to 1.1 by 0.1 do
      (loop for x from -2 to 1 by 0.04 do
        (let* ((c 126)
               (z (complex x y))
               (a z))
          (loop while (and (< (abs
                               (setq z (+ (* z z) a)))
                              2)
                           (> (decf c) 32)))
          (princ (code-char c))))
      (princ #\Newline)))