Mercurial > hg > Members > kono > os9 > sbc09

comparison basic/README.txt @ 57:2088fd998865

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
sbc09 directry clean up
author Shinji KONO <kono@ie.u-ryukyu.ac.jp>
date Mon, 23 Jul 2018 16:07:12 +0900
parents
children
comparison
equal deleted inserted replaced
56:4fa2bdb0c457 57:2088fd998865
1 BASIC AND FLOATING POINT ROUTINES FOR THE 6809
2 ==============================================
3
4 sbc09 stands for Lennart Benschop 6809 Single Board Computer.
5 It contains a assembler and simulator for the Motorola M6809 processor.
6
7 copyleft (c) 1994-2014 by the sbc09 team, see AUTHORS for more details.
8 license: GNU General Public License version 2, see LICENSE for more details.
9
10
11
12 FLOATING POINT ROUTINES FOR THE 6809
13 ------------------------------------
14
15 They are intended to be used with the sbc09 system. These routines
16 should be fairly portable to any 6809-based system.
17
18 As it is an unfinished program (intended to become a full-featured
19 BASIC interpreter one day), I never released it before and I almost
20 forgot about it. Fortunately it was still on a backup CD-R that I
21 made in 2001.
22
23
24 FILES
25 - - -
26
27 makeflot.c Conversion tool to convert floatnum.src to floatnum.inc
28
29 floatnum.inc Floating point constants to be included in main program.
30 floatnum.src Same constants, but not converted to binary.
31
32 fbasic.asm RPN calculator with floating point (just to test the FP routines.
33 This was intended to be part of a larger Basic interpreter,
34 but this was never finished).
35
36 basic.asm Tiny Basic
37 basic.txt Tiny Basic instructions
38
39 It was originally planned to turn this into a full-fledged BASIC
40 interpreter (maybe somewhat like BBC Basic), but this never
41 happened. It is now a rudimentary RPN calculator, just to test the
42 floating point routines. Each number or command needs to be on a separate
43 line.
44
45
46
47
48 MAKE THE PROGRAMS
49 - - - - - - - - -
50
51 Simple:
52
53 make
54
55
56 Or in single steps:
57
58 compile the helper tool ...
59
60 ./makeflot <floatnum.src >floatnum.inc
61
62
63 assemble the FP calculator ...
64
65 ./a09 fbasic.asm
66
67
68 assemble Tiny Basic (integer only) ...
69
70 ./a09 basic.asm
71
72
73
74
75 RUN THE PROGRAMS
76 - - - - - - - -
77
78
79 Start the board simulator
80
81 ../v09
82
83 You should see the prompt "Welcome to BUGGY version 1.0"
84
85 Type the command
86
87 xl400
88
89 Press the escape character Control-]
90 (e.g. on Linux for a german style keyboard Control+AltGr+9)
91
92 Then you see the v09> prompt.
93
94 Type the command
95
96 ufbasic
97
98 Now the file "fbasic" will be uploaded to the board.
99
100 Type the command
101
102 g400
103
104 Now you can type floating point numbers and commands (RPN style), each
105 on a different line, like this
106
107 2
108 3
109 *
110 6.00000000E+00
111
112 1
113 0
114 /
115
116 The last calculation breaks back to the monitor.
117
118 The following commands are available (see the source):
119 + - * / (the normal arithmetic operators).
120 = compare top two numbers on stack (and leave them), show < = or >
121 i round to integer (round to -Inf, like BASIC INT() function).
122 q square root
123 s sin
124 c cos
125 t tan
126 a atan
127 l ln
128 e exp
129 d duplicate number on stack
130 x exchange top numbers on stack.
131 r remove top of stack.
132
133
134
135 IMPLEMENTATION NOTES
136 - - - - - - - - - -
137
138 This is a 40-bit float, like many microcomputers of the 80s had,
139 including the Commodore 64, the ZX-Spectrum, the BBC and others. It
140 has an 8-bit exponent and a 32-bit mantissa (with hidden leading bit).
141 The basic operations (including square root) should be as accurate as
142 can be expected.
143
144 It does not do IEEE-754 features, such as Infinity, NaN, +/-zero and
145 subnormal numbers, but appears to work quite reasonably.
146
147 Trig functions deviate a few places in the 9th decimal. In particular
148 sin(pi/2) shows as 9.99999998E-01 instead of 1.00000000E+00. I
149 consider this acceptable and consistent with what could be expected.
150
151 The Log function deviates a few places in the 8th decimal. LN(5) appears to
152 be about worst-case. I find this a bit disappointing.
153
154 2
155 l
156 5
157 l
158 +
159 e
160
161 should show exactly 10, but it shows 9.99999970E+00 instead. This is
162 not caused by the exp function, but by the log of 5 (as I checked with
163 Python).
164
165