home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!dtix!darwin.sura.net!ukma!cs.widener.edu!dsinc!satalink!bert.tyler
- From: bert.tyler@satalink.com (Bert Tyler)
- Newsgroups: sci.fractals
- Subject: Data-dependent bug in Fra
- Message-ID: <1222.1004.uupcb@satalink.com>
- Date: 20 Nov 92 17:55:00 GMT
- Reply-To: bert.tyler@satalink.com (Bert Tyler)
- Organization: Datamax/Satalink Connection * Ivyland, PA (215) 443-9434
- Lines: 28
-
- Jon Noring / Barry Smith reported problems while running Fractint and
- the image generated by the command-line options:
-
- type=complexnewton corners=-6/5/5/-4 params=12/2/12/2 maxiter=2500
- periodicity=0
-
- First the good news: I just generated that fractal with no problems
- using the current internal developer's test version of Fractint on
- my 486/50DX2. Took 21 minutes, 39 seconds for a 1024x768x256 image.
-
- The big difference (both in terms of speed and in terms of
- eventual success) was the fact that my PC had a math coprocessor
- onboard and the other PCs didn't. The newton algorithms involve
- power functions, and the parameters used by this particular
- complexnewton involve a *lot* of powers, and this really stresses
- our internal math routines on FPU-less PCs. In this case, it stressed
- them until they broke.
-
- My guess as to why one fellow got a visible error message while the other
- got a trashed screen is that the second fellow's system threw his
- PC partially (but not completely) out of its graphics video state
- when it attempted to display that internal error message, and
- Fractint's graphics<->text video logic couldn't handle what was left.
-
- Bert Tyler (bert.tyler@satalink.com)
- ---
- . DeLuxe./386 1.25 #343sa . Void where prohibited. Batteries not included.
-
-