Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!mailrus!cornell!klassen
From: klassen@gvax.cs.cornell.edu (R. Victor Klassen)
Newsgroups: comp.graphics
Subject: Re: How does pr_vector() work ?
Message-ID: <38280@cornell.UUCP>
Date: 7 Mar 90 15:34:48 GMT
References: <1990Feb25.222157.26389@allgfx.agi.oz>
Sender: nobody@cornell.UUCP
Reply-To: klassen@gvax.cs.cornell.edu (R. Victor Klassen)
Organization: Cornell Univ. CS Dept, Ithaca NY
Lines: 32

In article <1990Feb25.222157.26389@allgfx.agi.oz> shn@allgfx.agi.oz (Si-Hanh NGUYEN) writes:
>
>Does anybody know the detailed algorithm used in pr_vector() of Sun Pixrect ?
>The manual says that "pr_vector draws balanced vectors. (The technique used
>is to balance the Breshenham error term)".
>
There are two papers of which I am aware which address what I beleive
to be the balancing to which the Sun manual refers.

%A J. Boothroyd
%A P.A. Hamilton
%T Exactly reversible plotter paths
%J Australian Computer Journal
%V 2
%N 1
%P 20-21
%D 1970
%K reversible incremental interpolation

%A D.E. Field
%T Incremental linear interpolation
%J ACM Transactions on Graphics
%V 4
%N 1
%P 1-11
%D January 1985
%K DDA Bresenham scan conversion

The issue is one of rounding down or toward zero.  I don't have anything more
at my fingertips, but I think you want to round down when you do the division
in the initialization, which isn't what most processors do if you're in the
wrong octant.