1tclrep/machineparameters(n) tclrep tclrep/machineparameters(n)
2______________________________________________________________________________
6
8 tclrep/machineparameters - Compute double precision machine parameters.
9
11 package require Tcl 8.4
12 package require snit
14 package require math::machineparameters 0.1
16 machineparameters create objectname ?options...?
18 objectname configure ?options...?
20 objectname cget opt
22 objectname destroy
24 objectname compute
26 objectname get key
28 objectname tostring
30 objectname print
32______________________________________________________________________________
34
36 The math::machineparameters package is the Tcl equivalent of the DLAMCH
37 LAPACK function. In floating point systems, a floating point number is
38 represented by
39 x = +/- d1 d2 ... dt basis^e
42 where digits satisfy
44 0 <= di <= basis - 1, i = 1, t
47 with the convention :
49 · t is the size of the mantissa
51 · basis is the basis (the "radix")
53 The compute method computes all machine parameters. Then, the get
55 method can be used to get each parameter. The print method prints a
56 report on standard output.
57
59 In the following example, one compute the parameters of a desktop under
60 Linux with the following Tcl 8.4.19 properties :
61 % parray tcl_platform
64 tcl_platform(byteOrder) = littleEndian
65 tcl_platform(machine) = i686
66 tcl_platform(os) = Linux
67 tcl_platform(osVersion) = 2.6.24-19-generic
68 tcl_platform(platform) = unix
69 tcl_platform(tip,268) = 1
70 tcl_platform(tip,280) = 1
71 tcl_platform(user) = <username>
72 tcl_platform(wordSize) = 4
73 The following example creates a machineparameters object, computes the
75 properties and displays it.
76 set pp [machineparameters create %AUTO%]
79 $pp compute
80 $pp print
81 $pp destroy
82 This prints out :
84 Machine parameters
87 Epsilon : 1.11022302463e-16
88 Beta : 2
89 Rounding : proper
90 Mantissa : 53
91 Maximum exponent : 1024
92 Minimum exponent : -1021
93 Overflow threshold : 8.98846567431e+307
94 Underflow threshold : 2.22507385851e-308
95 That compares well with the results produced by Lapack 3.1.1 :
97 Epsilon = 1.11022302462515654E-016
100 Safe minimum = 2.22507385850720138E-308
101 Base = 2.0000000000000000
102 Precision = 2.22044604925031308E-016
103 Number of digits in mantissa = 53.000000000000000
104 Rounding mode = 1.00000000000000000
105 Minimum exponent = -1021.0000000000000
106 Underflow threshold = 2.22507385850720138E-308
107 Largest exponent = 1024.0000000000000
108 Overflow threshold = 1.79769313486231571E+308
109 Reciprocal of safe minimum = 4.49423283715578977E+307
110 The following example creates a machineparameters object, computes the
112 properties and gets the epsilon for the machine.
113 set pp [machineparameters create %AUTO%]
116 $pp compute
117 set eps [$pp get -epsilon]
118 $pp destroy
119
122 · "Algorithms to Reveal Properties of Floating-Point Arithmetic",
123 Michael A. Malcolm, Stanford University, Communications of the
124 ACM, Volume 15 , Issue 11 (November 1972), Pages: 949 - 951
125 · "More on Algorithms that Reveal Properties of Floating, Point
127 Arithmetic Units", W. Morven Gentleman, University of Waterloo,
128 Scott B. Marovich, Purdue University, Communications of the ACM,
129 Volume 17 , Issue 5 (May 1974), Pages: 276 - 277
130
132 machineparameters create objectname ?options...?
133 The command creates a new machineparameters object and returns
134 the fully qualified name of the object command as its result.
135 -verbose verbose
137 Set this option to 1 to enable verbose logging. This
138 option is mainly for debug purposes. The default value
139 of verbose is 0.
140
142 objectname configure ?options...?
143 The command configure the options of the object objectname. The
144 options are the same as the static method create.
145 objectname cget opt
147 Returns the value of the option which name is opt. The options
148 are the same as the method create and configure.
149 objectname destroy
151 Destroys the object objectname.
152 objectname compute
154 Computes the machine parameters.
155 objectname get key
157 Returns the value corresponding with given key. The following
158 is the list of available keys.
159 · -epsilon : smallest value so that 1+epsilon>1 is false
161 · -rounding : The rounding mode used on the machine. The
163 rounding occurs when more than t digits would be required
164 to represent the number. Two modes can be determined
165 with the current system : "chop" means than only t digits
166 are kept, no matter the value of the number "proper"
167 means that another rounding mode is used, be it "round to
168 nearest", "round up", "round down".
169 · -basis : the basis of the floating-point representation.
171 The basis is usually 2, i.e. binary representation (for
172 example IEEE 754 machines), but some machines (like HP
173 calculators for example) uses 10, or 16, etc...
174 · -mantissa : the number of bits in the mantissa
176 · -exponentmax : the largest positive exponent before
178 overflow occurs
179 · -exponentmin : the largest negative exponent before
181 (gradual) underflow occurs
182 · -vmax : largest positive value before overflow occurs
184 · -vmin : largest negative value before (gradual) underflow
186 occurs
187 objectname tostring
189 Return a report for machine parameters.
190 objectname print
192 Print machine parameters on standard output.
193
195 This document, and the package it describes, will undoubtedly contain
196 bugs and other problems. Please report such in the category math of
197 the Tcllib Trackers [http://core.tcl.tk/tcllib/reportlist]. Please
198 also report any ideas for enhancements you may have for either package
199 and/or documentation.
200 When proposing code changes, please provide unified diffs, i.e the out‐
202 put of diff -u.
203 Note further that attachments are strongly preferred over inlined
205 patches. Attachments can be made by going to the Edit form of the
206 ticket immediately after its creation, and then using the left-most
207 button in the secondary navigation bar.
208
210 Copyright (c) 2008 Michael Baudin <michael.baudin@sourceforge.net>
211tcllib 1.0 tclrep/machineparameters(n)