• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<HTML>
2<HEAD>
3<TITLE>Constructive Real Calculator and Library Implementation Notes</title>
4</head>
5<BODY BGCOLOR="#FFFFFF">
6<H1>Constructive Real Calculator and Library Implementation Notes</h1>
7</body>
8The calculator is based on the constructive real library consisting
9mainly of <TT>com.sgi.math.CR</tt> and <TT>com.sgi.math.UnaryCRFunction</tt>.
10The former provides basic arithmetic operations on constructive reals.
11The latter provides some basic operations on unary functions over the
12constructive reals.
13<H2>General approach</h2>
14The library was not designed to use the absolute best known algorithms
15and to provide the best possible performance.  To do so would have
16significantly complicated the code, and lengthened start-up times for
17the calculator and similar applications.  Instead the goals were to:
18<OL>
19<LI> Rely on the standard library to the greatest possible extent.
20The implementation relies heavily on <TT>java.math.BigInteger</tt>,
21in spite of the fact that it may not provide asymptotically optimal
22performance for all operations.
23<LI> Use algorithms with asymptotically reasonable performance.
24<LI> Keep the code, and especially the library code, simple.
25This was done both to make it more easily understandable, and to
26keep down class loading time.
27<LI> Avoid heuristics.  The code accepts that there is no practical way
28to avoid diverging computations.  The user interface is designed to
29deal with that.  There is no attempt to try to prove that a computation
30will diverge ahead of time, not even in cases in which such a proof is
31trivial.
32</ol>
33A constructive real number <I>x</i> is represented abstractly as a function
34<I>f<SUB>x</sub></i>, such that  <I>f<SUB>x</sub>(n)</i> produces a scaled
35integer approximation to <I>x</i>, with an error of strictly less than
362<SUP><I>n</i></sup>.  More precisely:
37<P>
38|<I>f<SUB>x</sub></i>(<I>n</i>) * 2<SUP><I>n</i></sup> - x| < 2<SUP><I>n</i></sup>
39<P>
40Since Java does not support higher order functions, these functions
41are actually represented as objects with an <TT>approximate</tt>
42function.  In order to obtain reasonable performance, each object
43caches the best previous approximation computed so far.
44<P>
45This is very similar to earlier work by Boehm, Lee, Cartwright, Riggle, and
46O'Donnell.
47The implementation borrows many ideas from the
48<A HREF="http://reality.sgi.com/boehm/calc">calculator</a>
49developed earlier by Boehm and Lee.  The major differences are the
50user interface, the portability of the implementation, the emphasis
51on simplicity, and the reliance on a general implementation of inverse
52functions.
53<P>
54Several alternate and functionally equivalent representations of
55constructive real numbers are possible.
56Gosper and Vuillemin proposed representations based on continued
57fractions.
58A representation as functions
59producing variable precision intervals is probably more efficient
60for larger scale computation.
61We chose this representation because it can be implemented compactly
62layered on large integer arithmetic.
63<H2>Transcendental functions</h2>
64The exponential and natural logarithm functions are implemented as Taylor
65series expansions.  There is also a specialized function that computes
66the Taylor series expansion of atan(1/n), where n is a small integer.
67This allows moderately efficient computation of pi using
68<P>
69pi/4 = 4*atan(1/5) - atan(1/239)
70<P>
71All of the remaining trigonometric functions are implemented in terms
72of the cosine function, which again uses a Taylor series expansion.
73<P>
74The inverse trigonometric functions are implemented using a general
75inverse function operation in <TT>UnaryCRFunction</tt>.  This is
76more expensive than a direct implementation, since each time an approximation
77to the result is computed, several evaluations of the underlying
78trigonometric function are needed.  Nonetheless, it appears to be
79surprisingly practical, at least for something as undemanding as a desk
80calculator.
81<H2>Prior work</h2>
82There has been much prior research on the constructive/recursive/computable
83real numbers and constructive analysis.  Relatively little of this
84has been concerned with issues related to practical implementations.
85<P>
86Several implementation efforts examined representations based on
87either infinite, lazily-evaluated decimal expansions (Schwartz),
88or continued fractions (Gosper, Vuillemin).  So far, these appear
89less practical than what is implemented here.
90<P>
91Probably the most practical approach to constructive real arithmetic
92is one based on interval arithmetic.  A variant that is close to,
93but not quite, constructive real arithmetic is described in
94<P>
95Oliver Aberth, <I>Precise Numerical Analysis</i>, Wm. C. Brown Publishers,
96Dubuque, Iowa, 1988.
97<P>
98The issues related to using this in a higher performance implementation
99of constructive real arithmetic are explored in
100<P>
101Lee and Boehm, "Optimizing Programs over the Constructive Reals",
102<I>ACM SIGPLAN 90 Conference on Programming Language Design and
103Implementation, SIGPLAN Notices 25</i>, 6, pp. 102-111.
104<P>
105The particular implementation strategy used n this calculator was previously
106described in
107<P>
108Boehm, Cartwright, Riggle, and O'Donnell, "Exact Real Arithmetic:
109A Case Study in Higher Order Programming", <I>Proceedings of the
1101986 ACM Lisp and Functional Programming Conference</i>, pp. 162-173, 1986.
111</body>
112</html>
113