Path: utzoo!attcan!uunet!ogicse!plains!bakke
From: bakke@plains.UUCP (Jeffrey P. Bakke)
Newsgroups: comp.graphics
Subject: B-Splines
Message-ID: <3834@plains.UUCP>
Date: 21 Mar 90 18:11:58 GMT
Reply-To: bakke@plains.UUCP (Jeffrey P. Bakke)
Organization: North Dakota State University, Fargo
Lines: 30

I have a project that I am currently working on that requires me to
create a polygon manipulation package for 2-D graphics.  Up to this
point everything has been great.  Its up and running but I'd like to
improve the code somewhat.  This is the problem, I have a function
called polygon_BSpline, that when sent the polygon structure, will
plot a BSpline curve using 4 control points.  My problem is, is that
this function, out of the whole 10,000 line program is the only function
that requires floating point.  

I was wondering of those of you in netland would have either a code segment,
generalized algorithm or good references for implementing a BSpline 
function via integers.  If I could get rid of the floating point, i can
dump the 20K FP Emulation library that Turbo C lags onto my program.

I attempted to convert the current Spline function that I had using long
integers and shifting the floating point into the integers.  Unfortunately
my implementation will overflow the long integers if I set a shift greater
than 8 bits (256 * the Floating point number).  And of course, if I do
use 8 bit shift into the integers, I lose so much precision that 
my spline becomes hash anyway...

Anyway, any pointers as to references, algorithms or code would be
greatly appreciated...


Jeff Bakke
bakke@plains.NoDak.edu


Oh, and i do read this group so feel free to post for others....