line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
=head1 NAME |
2
|
|
|
|
|
|
|
|
3
|
|
|
|
|
|
|
Data::Entropy::Algorithms - basic entropy-using algorithms |
4
|
|
|
|
|
|
|
|
5
|
|
|
|
|
|
|
=head1 SYNOPSIS |
6
|
|
|
|
|
|
|
|
7
|
|
|
|
|
|
|
use Data::Entropy::Algorithms |
8
|
|
|
|
|
|
|
qw(rand_bits rand_int rand_prob); |
9
|
|
|
|
|
|
|
|
10
|
|
|
|
|
|
|
$str = rand_bits(17); |
11
|
|
|
|
|
|
|
$i = rand_int(12345); |
12
|
|
|
|
|
|
|
$i = rand_int(Math::BigInt->new("1000000000000")); |
13
|
|
|
|
|
|
|
$j = rand_prob(1, 2, 3); |
14
|
|
|
|
|
|
|
$j = rand_prob([ 1, 2, 3 ]); |
15
|
|
|
|
|
|
|
|
16
|
|
|
|
|
|
|
use Data::Entropy::Algorithms qw(rand_fix rand rand_flt); |
17
|
|
|
|
|
|
|
|
18
|
|
|
|
|
|
|
$x = rand_fix(48); |
19
|
|
|
|
|
|
|
$x = rand(7); |
20
|
|
|
|
|
|
|
$x = rand_flt(0.0, 7.0); |
21
|
|
|
|
|
|
|
|
22
|
|
|
|
|
|
|
use Data::Entropy::Algorithms |
23
|
|
|
|
|
|
|
qw(pick pick_r choose choose_r shuffle shuffle_r); |
24
|
|
|
|
|
|
|
|
25
|
|
|
|
|
|
|
$item = pick($item0, $item1, $item2); |
26
|
|
|
|
|
|
|
$item = pick_r(\@items); |
27
|
|
|
|
|
|
|
@chosen = choose(3, $item0, $item1, $item2, $item3, $item4); |
28
|
|
|
|
|
|
|
$chosen = choose_r(3, \@items); |
29
|
|
|
|
|
|
|
@shuffled = shuffle($item0, $item1, $item2, $item3, $item4); |
30
|
|
|
|
|
|
|
$shuffled = shuffle_r(\@items); |
31
|
|
|
|
|
|
|
|
32
|
|
|
|
|
|
|
=head1 DESCRIPTION |
33
|
|
|
|
|
|
|
|
34
|
|
|
|
|
|
|
This module contains a collection of fundamental algorithms that |
35
|
|
|
|
|
|
|
use entropy. They all use the entropy source mechanism described in |
36
|
|
|
|
|
|
|
L. |
37
|
|
|
|
|
|
|
|
38
|
|
|
|
|
|
|
=cut |
39
|
|
|
|
|
|
|
|
40
|
|
|
|
|
|
|
package Data::Entropy::Algorithms; |
41
|
|
|
|
|
|
|
|
42
|
9
|
|
|
9
|
|
78458
|
{ use 5.006; } |
|
9
|
|
|
|
|
36
|
|
|
9
|
|
|
|
|
424
|
|
43
|
9
|
|
|
9
|
|
55
|
use warnings; |
|
9
|
|
|
|
|
18
|
|
|
9
|
|
|
|
|
302
|
|
44
|
9
|
|
|
9
|
|
47
|
use strict; |
|
9
|
|
|
|
|
16
|
|
|
9
|
|
|
|
|
326
|
|
45
|
|
|
|
|
|
|
|
46
|
9
|
|
|
9
|
|
46
|
use Carp qw(croak); |
|
9
|
|
|
|
|
17
|
|
|
9
|
|
|
|
|
557
|
|
47
|
9
|
|
|
9
|
|
628
|
use Data::Entropy qw(entropy_source); |
|
9
|
|
|
|
|
14
|
|
|
9
|
|
|
|
|
542
|
|
48
|
9
|
|
|
|
|
1399
|
use Data::Float 0.008 qw( |
49
|
|
|
|
|
|
|
have_subnormal min_normal_exp significand_bits |
50
|
|
|
|
|
|
|
float_is_finite float_parts float_sign mult_pow2 copysign |
51
|
9
|
|
|
9
|
|
10670
|
); |
|
9
|
|
|
|
|
89795
|
|
52
|
9
|
|
|
9
|
|
83
|
use Params::Classify 0.000 qw(is_ref); |
|
9
|
|
|
|
|
154
|
|
|
9
|
|
|
|
|
585
|
|
53
|
|
|
|
|
|
|
|
54
|
|
|
|
|
|
|
our $VERSION = "0.007"; |
55
|
|
|
|
|
|
|
|
56
|
9
|
|
|
9
|
|
52
|
use parent "Exporter"; |
|
9
|
|
|
|
|
22
|
|
|
9
|
|
|
|
|
78
|
|
57
|
|
|
|
|
|
|
our @EXPORT_OK = qw( |
58
|
|
|
|
|
|
|
rand_bits rand_int rand_prob |
59
|
|
|
|
|
|
|
rand_fix rand rand_flt |
60
|
|
|
|
|
|
|
pick_r pick choose_r choose shuffle_r shuffle |
61
|
|
|
|
|
|
|
); |
62
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
=head1 FUNCTIONS |
64
|
|
|
|
|
|
|
|
65
|
|
|
|
|
|
|
All of these functions use entropy. The entropy source is not an |
66
|
|
|
|
|
|
|
explicit input in any case. All functions use the current entropy source |
67
|
|
|
|
|
|
|
maintained by the C module. To select an entropy source |
68
|
|
|
|
|
|
|
use the C function in that module, or alternatively |
69
|
|
|
|
|
|
|
do nothing to use the default source. |
70
|
|
|
|
|
|
|
|
71
|
|
|
|
|
|
|
=head2 Fundamental entropy extraction |
72
|
|
|
|
|
|
|
|
73
|
|
|
|
|
|
|
=over |
74
|
|
|
|
|
|
|
|
75
|
|
|
|
|
|
|
=item rand_bits(NBITS) |
76
|
|
|
|
|
|
|
|
77
|
|
|
|
|
|
|
Returns NBITS bits of entropy, as a string of octets. If NBITS is |
78
|
|
|
|
|
|
|
not a multiple of eight then the last octet in the string has its most |
79
|
|
|
|
|
|
|
significant bits set to zero. |
80
|
|
|
|
|
|
|
|
81
|
|
|
|
|
|
|
=cut |
82
|
|
|
|
|
|
|
|
83
|
|
|
|
|
|
|
sub rand_bits($) { |
84
|
42
|
|
|
42
|
1
|
14170
|
my($nbits) = @_; |
85
|
42
|
100
|
|
|
|
296
|
croak "need a non-negative number of bits to dispense" |
86
|
|
|
|
|
|
|
unless $nbits >= 0; |
87
|
41
|
|
|
|
|
110
|
return entropy_source->get_bits($nbits); |
88
|
|
|
|
|
|
|
} |
89
|
|
|
|
|
|
|
|
90
|
|
|
|
|
|
|
=item rand_int(LIMIT) |
91
|
|
|
|
|
|
|
|
92
|
|
|
|
|
|
|
LIMIT must be a positive integer. Returns a uniformly-distributed random |
93
|
|
|
|
|
|
|
integer in the range [0, LIMIT). LIMIT may be either a native integer, |
94
|
|
|
|
|
|
|
a C object, or an integer-valued C object; |
95
|
|
|
|
|
|
|
the returned number is of the same type. |
96
|
|
|
|
|
|
|
|
97
|
|
|
|
|
|
|
=cut |
98
|
|
|
|
|
|
|
|
99
|
|
|
|
|
|
|
sub rand_int($) { |
100
|
20640
|
|
|
20640
|
1
|
91631
|
my($limit) = @_; |
101
|
20640
|
100
|
|
|
|
50836
|
croak "need a positive upper limit for random variable" |
102
|
|
|
|
|
|
|
unless $limit > 0; |
103
|
20639
|
|
|
|
|
58780
|
return entropy_source->get_int($limit); |
104
|
|
|
|
|
|
|
} |
105
|
|
|
|
|
|
|
|
106
|
|
|
|
|
|
|
=item rand_prob(PROB ...) |
107
|
|
|
|
|
|
|
|
108
|
|
|
|
|
|
|
=item rand_prob(PROBS) |
109
|
|
|
|
|
|
|
|
110
|
|
|
|
|
|
|
Returns a random integer selected with non-uniform probability. The |
111
|
|
|
|
|
|
|
relative probabilities are supplied as a list of non-negative integers |
112
|
|
|
|
|
|
|
(multiple PROB arguments) or a reference to an array of integers (the |
113
|
|
|
|
|
|
|
PROBS argument). The relative probabilities may be native integers, |
114
|
|
|
|
|
|
|
C objects, or integer-valued C objects; |
115
|
|
|
|
|
|
|
they must all be of the same type. At least one probability value must |
116
|
|
|
|
|
|
|
be positive. |
117
|
|
|
|
|
|
|
|
118
|
|
|
|
|
|
|
The first relative probability value (the first PROB or the first element |
119
|
|
|
|
|
|
|
of PROBS) is the relative probability of returning 0. The absolute |
120
|
|
|
|
|
|
|
probability of returning 0 is this value divided by the total of all |
121
|
|
|
|
|
|
|
the relative probability values. Similarly the second value controls |
122
|
|
|
|
|
|
|
the probability of returning 1, and so on. |
123
|
|
|
|
|
|
|
|
124
|
|
|
|
|
|
|
=cut |
125
|
|
|
|
|
|
|
|
126
|
|
|
|
|
|
|
sub rand_prob(@) { |
127
|
154
|
100
|
100
|
154
|
1
|
95373
|
my $probs = @_ == 1 && is_ref($_[0], "ARRAY") ? $_[0] : \@_; |
128
|
154
|
|
|
|
|
211
|
my $total = 0; |
129
|
154
|
|
|
|
|
401
|
for(my $i = @$probs; $i--; ) { |
130
|
754
|
|
|
|
|
897
|
my $prob = $probs->[$i]; |
131
|
754
|
100
|
|
|
|
1675
|
croak "probabilities must be non-negative" unless $prob >= 0; |
132
|
753
|
|
|
|
|
1504
|
$total += $prob; |
133
|
|
|
|
|
|
|
} |
134
|
153
|
100
|
|
|
|
432
|
croak "can't have nothing possible" if $total == 0; |
135
|
152
|
|
|
|
|
411
|
for(my $i = @$probs; $i--; ) { |
136
|
370
|
|
|
|
|
446
|
my $prob = $probs->[$i]; |
137
|
370
|
|
|
|
|
398
|
$total -= $prob; |
138
|
370
|
100
|
|
|
|
1476
|
return $i if entropy_source->get_prob($total, $prob); |
139
|
|
|
|
|
|
|
} |
140
|
|
|
|
|
|
|
} |
141
|
|
|
|
|
|
|
|
142
|
|
|
|
|
|
|
=back |
143
|
|
|
|
|
|
|
|
144
|
|
|
|
|
|
|
=head2 Numbers |
145
|
|
|
|
|
|
|
|
146
|
|
|
|
|
|
|
=over |
147
|
|
|
|
|
|
|
|
148
|
|
|
|
|
|
|
=item rand_fix(NBITS) |
149
|
|
|
|
|
|
|
|
150
|
|
|
|
|
|
|
Returns a uniformly-distributed random NBITS-bit fixed-point fraction in |
151
|
|
|
|
|
|
|
the range [0, 1). That is, the result is a randomly-chosen multiple of |
152
|
|
|
|
|
|
|
2^-NBITS, the multiplier being a random integer in the range [0, 2^NBITS). |
153
|
|
|
|
|
|
|
The value is returned in the form of a native floating point number, so |
154
|
|
|
|
|
|
|
NBITS can be at most one greater than the number of bits of significand |
155
|
|
|
|
|
|
|
in the floating point format. |
156
|
|
|
|
|
|
|
|
157
|
|
|
|
|
|
|
With NBITS = 48 the range of output values is the same as that of the |
158
|
|
|
|
|
|
|
Unix C function. |
159
|
|
|
|
|
|
|
|
160
|
|
|
|
|
|
|
=cut |
161
|
|
|
|
|
|
|
|
162
|
|
|
|
|
|
|
sub rand_fix($) { |
163
|
516
|
|
|
516
|
1
|
101117
|
my($nbits) = @_; |
164
|
516
|
100
|
|
|
|
1475
|
croak "need a non-negative number of bits to dispense" |
165
|
|
|
|
|
|
|
unless $nbits >= 0; |
166
|
515
|
50
|
|
|
|
1056
|
croak "can't generate more than ".(significand_bits+1). |
167
|
|
|
|
|
|
|
" bits of fixed-point fraction" |
168
|
|
|
|
|
|
|
if $nbits > significand_bits+1; |
169
|
515
|
|
|
|
|
595
|
my $frac = 0.0; |
170
|
515
|
|
|
|
|
1179
|
for(my $pos = 24; $pos <= $nbits; $pos += 24) { |
171
|
516
|
|
|
|
|
7514
|
$frac += mult_pow2(rand_int(1 << 24), -$pos); |
172
|
|
|
|
|
|
|
} |
173
|
515
|
|
|
|
|
8963
|
$frac += mult_pow2(rand_int(1 << ($nbits % 24)), -$nbits); |
174
|
515
|
|
|
|
|
11299
|
return $frac; |
175
|
|
|
|
|
|
|
} |
176
|
|
|
|
|
|
|
|
177
|
|
|
|
|
|
|
=item rand([LIMIT]) |
178
|
|
|
|
|
|
|
|
179
|
|
|
|
|
|
|
Generates a random fixed-point fraction by C and then multiplies |
180
|
|
|
|
|
|
|
it by LIMIT, returning the result. LIMIT defaults to 1, and if it |
181
|
|
|
|
|
|
|
is 0 then that is also treated as 1. The length of the fixed-point |
182
|
|
|
|
|
|
|
fraction is 48 bits, unless that can't be represented in the native |
183
|
|
|
|
|
|
|
floating point type, in which case the longest possible fraction will |
184
|
|
|
|
|
|
|
be generated instead. |
185
|
|
|
|
|
|
|
|
186
|
|
|
|
|
|
|
This is a drop-in replacement for C: it produces exactly the |
187
|
|
|
|
|
|
|
same range of output values, but using the current entropy source instead |
188
|
|
|
|
|
|
|
of a sucky PRNG with linear relationships between successive outputs. |
189
|
|
|
|
|
|
|
(C does the type of calculation described, but using the |
190
|
|
|
|
|
|
|
PRNG C to generate the fixed-point fraction.) The details of |
191
|
|
|
|
|
|
|
behaviour may change in the future if the behaviour of C |
192
|
|
|
|
|
|
|
changes, to maintain the match. |
193
|
|
|
|
|
|
|
|
194
|
|
|
|
|
|
|
Where the source of a module can't be readily modified, it can be made |
195
|
|
|
|
|
|
|
to use this C by an incantation such as |
196
|
|
|
|
|
|
|
|
197
|
|
|
|
|
|
|
*Foreign::Module::rand = \&Data::Entropy::Algorithms::rand; |
198
|
|
|
|
|
|
|
|
199
|
|
|
|
|
|
|
This must be done before the module is loaded, most likely in a C |
200
|
|
|
|
|
|
|
block. It is also possible to override C for all modules, |
201
|
|
|
|
|
|
|
by performing this similarly early: |
202
|
|
|
|
|
|
|
|
203
|
|
|
|
|
|
|
*CORE::GLOBAL::rand = \&Data::Entropy::Algorithms::rand; |
204
|
|
|
|
|
|
|
|
205
|
|
|
|
|
|
|
This function should not be used in any new code, because the kind |
206
|
|
|
|
|
|
|
of output supplied by C is hardly ever the right thing to use. |
207
|
|
|
|
|
|
|
The C idiom to generate a random integer has non-uniform |
208
|
|
|
|
|
|
|
probabilities of generating each possible value, except when C<$n> is a |
209
|
|
|
|
|
|
|
power of two. For floating point numbers, C can't generate most |
210
|
|
|
|
|
|
|
representable numbers in its output range, and the output is biased |
211
|
|
|
|
|
|
|
towards zero. In new code use C to generate integers and |
212
|
|
|
|
|
|
|
C to generate floating point numbers. |
213
|
|
|
|
|
|
|
|
214
|
|
|
|
|
|
|
=cut |
215
|
|
|
|
|
|
|
|
216
|
9
|
|
|
9
|
|
4802
|
use constant RAND_NBITS => 48 > significand_bits+1 ? significand_bits+1 : 48; |
|
9
|
|
|
|
|
20
|
|
|
9
|
|
|
|
|
16572
|
|
217
|
|
|
|
|
|
|
|
218
|
|
|
|
|
|
|
sub rand(;$) { |
219
|
258
|
|
|
258
|
1
|
137514
|
my($limit) = @_; |
220
|
258
|
100
|
100
|
|
|
559
|
return rand_fix(RAND_NBITS) * |
221
|
|
|
|
|
|
|
(!defined($limit) || $limit == 0.0 ? 1.0 : $limit); |
222
|
|
|
|
|
|
|
} |
223
|
|
|
|
|
|
|
|
224
|
|
|
|
|
|
|
=item rand_flt(MIN, MAX) |
225
|
|
|
|
|
|
|
|
226
|
|
|
|
|
|
|
Selects a uniformly-distributed real number (with infinite precision) |
227
|
|
|
|
|
|
|
in the range [MIN, MAX] and then rounds this number to the nearest |
228
|
|
|
|
|
|
|
representable floating point value, which it returns. (Actually it is |
229
|
|
|
|
|
|
|
only I the function worked this way: in fact it never generates |
230
|
|
|
|
|
|
|
the number with infinite precision. It selects between the representable |
231
|
|
|
|
|
|
|
floating point values with the probabilities implied by this process.) |
232
|
|
|
|
|
|
|
|
233
|
|
|
|
|
|
|
This can return absolutely any floating point value in the range [MIN, |
234
|
|
|
|
|
|
|
MAX]; both MIN and MAX themselves are possible return values. All bits |
235
|
|
|
|
|
|
|
of the floating point type are filled randomly, so the range of values |
236
|
|
|
|
|
|
|
that can be returned depends on the details of the floating point format. |
237
|
|
|
|
|
|
|
(See L for low-level floating point utilities.) |
238
|
|
|
|
|
|
|
|
239
|
|
|
|
|
|
|
The function Cs if MIN and MAX are not both finite. If MIN is |
240
|
|
|
|
|
|
|
greater than MAX then their roles are swapped: the order of the limit |
241
|
|
|
|
|
|
|
parameters actually doesn't matter. If the limits are identical then |
242
|
|
|
|
|
|
|
that value is always returned. As a special case, if the limits are |
243
|
|
|
|
|
|
|
positive zero and negative zero then a zero will be returned with a |
244
|
|
|
|
|
|
|
randomly-chosen sign. |
245
|
|
|
|
|
|
|
|
246
|
|
|
|
|
|
|
=cut |
247
|
|
|
|
|
|
|
|
248
|
|
|
|
|
|
|
sub rand_flt($$) { |
249
|
11000
|
|
|
11000
|
1
|
6371939
|
my($a, $b) = @_; |
250
|
11000
|
50
|
33
|
|
|
399273
|
croak "bounds for rand_flt() must be finite" |
251
|
|
|
|
|
|
|
unless float_is_finite($a) && float_is_finite($b); |
252
|
11000
|
100
|
|
|
|
535448
|
if($a == $b) { |
253
|
2000
|
100
|
66
|
|
|
27211
|
return $_[rand_int(2)] |
254
|
|
|
|
|
|
|
if $a == 0.0 && float_sign($a) ne float_sign($b); |
255
|
1000
|
|
|
|
|
2459
|
return $_[0]; |
256
|
|
|
|
|
|
|
} |
257
|
9000
|
100
|
|
|
|
28765
|
($a, $b) = ($b, $a) if abs($a) < abs($b); |
258
|
9000
|
50
|
|
|
|
244935
|
my($prm_sign, $prm_max_exp, $prm_max_sgnf) = |
259
|
|
|
|
|
|
|
$a == 0.0 ? ("+", min_normal_exp, 0.0) : float_parts($a); |
260
|
9000
|
100
|
|
|
|
826254
|
my($b_sign, $b_exp, $b_sgnf) = |
261
|
|
|
|
|
|
|
$b == 0.0 ? ("+", min_normal_exp, 0.0) : float_parts($b); |
262
|
9000
|
|
|
|
|
170637
|
my($min_exp, $min_sgnf); |
263
|
0
|
|
|
|
|
0
|
my($opp_max_exp, $opp_max_sgnf); |
264
|
9000
|
100
|
|
|
|
26450
|
if($b_sign eq $prm_sign) { |
265
|
6000
|
|
|
|
|
11696
|
($min_exp, $min_sgnf) = ($b_exp, $b_sgnf); |
266
|
6000
|
|
|
|
|
10297
|
($opp_max_exp, $opp_max_sgnf) = (min_normal_exp, 0.0); |
267
|
|
|
|
|
|
|
} else { |
268
|
3000
|
|
|
|
|
4671
|
($min_exp, $min_sgnf) = (min_normal_exp, 0.0); |
269
|
3000
|
|
|
|
|
4947
|
($opp_max_exp, $opp_max_sgnf) = ($b_exp, $b_sgnf); |
270
|
|
|
|
|
|
|
} |
271
|
9000
|
|
|
|
|
11680
|
TRY_AGAIN: |
272
|
|
|
|
|
|
|
my $exp = $prm_max_exp; |
273
|
9000
|
|
|
|
|
12118
|
my $bdone = significand_bits; |
274
|
9000
|
50
|
|
|
|
32301
|
$bdone = 28 if $bdone > 28; |
275
|
9000
|
|
|
|
|
16197
|
my $prm_frng = $prm_max_sgnf * (1 << $bdone); |
276
|
9000
|
50
|
|
|
|
19402
|
if($prm_max_sgnf < 1.0) { |
277
|
|
|
|
|
|
|
# subnormal limit |
278
|
0
|
|
|
|
|
0
|
my $desired_rng = 1 << $bdone; |
279
|
0
|
|
|
|
|
0
|
while($bdone != significand_bits) { |
280
|
0
|
|
|
|
|
0
|
$bdone++; |
281
|
0
|
|
|
|
|
0
|
$prm_frng += $prm_frng; |
282
|
0
|
0
|
|
|
|
0
|
last if $prm_frng >= $desired_rng; |
283
|
|
|
|
|
|
|
} |
284
|
|
|
|
|
|
|
} |
285
|
9000
|
|
|
|
|
14875
|
my $prm_rng = int($prm_frng); |
286
|
9000
|
50
|
|
|
|
21631
|
$prm_rng++ if $prm_frng != $prm_rng; |
287
|
9000
|
|
|
|
|
16102
|
my $min_b = $bdone - ($exp - $min_exp); |
288
|
9000
|
100
|
|
|
|
30268
|
my $min_rng = $min_b >= 0 ? int(mult_pow2($min_sgnf, $min_b)) : 0; |
289
|
9000
|
|
|
|
|
159768
|
my $opp_b = $bdone - ($exp - $opp_max_exp); |
290
|
9000
|
100
|
|
|
|
22543
|
my $opp_frng = $opp_b >= 0 ? mult_pow2($opp_max_sgnf, $opp_b) : 0; |
291
|
9000
|
|
|
|
|
82356
|
my $opp_rng = int($opp_frng); |
292
|
9000
|
50
|
|
|
|
25920
|
$opp_rng++ if $opp_frng != $opp_rng; |
293
|
9000
|
|
|
|
|
23281
|
my $n = $min_rng + rand_int($prm_rng - $min_rng + $opp_rng); |
294
|
9000
|
|
|
|
|
19493
|
my($sg, $max_exp, $max_sgnf) = ($a, $prm_max_exp, $prm_max_sgnf); |
295
|
9000
|
100
|
|
|
|
20227
|
if($n >= $prm_rng) { |
296
|
1184
|
|
|
|
|
1412
|
$n -= $prm_rng; |
297
|
1184
|
|
|
|
|
2080
|
($sg, $max_exp, $max_sgnf) = ($b, $opp_max_exp, $opp_max_sgnf); |
298
|
|
|
|
|
|
|
} |
299
|
9000
|
|
33
|
|
|
27516
|
while($n == 0 && $exp - $bdone - 1 >= min_normal_exp) { |
300
|
0
|
|
|
|
|
0
|
$exp -= $bdone + 1; |
301
|
0
|
|
|
|
|
0
|
$n = rand_int(2 << $bdone); |
302
|
|
|
|
|
|
|
} |
303
|
9000
|
|
|
|
|
19696
|
for(my $bit = 16; $bit; $bit >>= 1) { |
304
|
45000
|
100
|
33
|
|
|
330314
|
if($bdone >= $bit && $exp - $bit >= min_normal_exp && |
|
|
|
66
|
|
|
|
|
305
|
|
|
|
|
|
|
$n < (2 << ($bdone - $bit))) { |
306
|
8440
|
|
|
|
|
9632
|
$bdone -= $bit; |
307
|
8440
|
|
|
|
|
18569
|
$exp -= $bit; |
308
|
|
|
|
|
|
|
} |
309
|
|
|
|
|
|
|
} |
310
|
9000
|
50
|
|
|
|
34533
|
goto TRY_AGAIN if $exp < $min_exp; |
311
|
9000
|
100
|
|
|
|
16698
|
my $top_sgnf = $exp == $max_exp ? $max_sgnf : 2.0; |
312
|
9000
|
100
|
|
|
|
25068
|
my $bot_sgnf = $exp == $min_exp ? $min_sgnf : 1.0; |
313
|
9000
|
|
|
|
|
27163
|
my $sgnf = mult_pow2($n, -$bdone); |
314
|
9000
|
|
|
|
|
291077
|
if(!have_subnormal && $exp == min_normal_exp && $sgnf < 1.0) { |
315
|
|
|
|
|
|
|
$top_sgnf = 1.0; |
316
|
|
|
|
|
|
|
$sgnf = 0.0; |
317
|
|
|
|
|
|
|
} else { |
318
|
9000
|
|
|
|
|
9864
|
$bot_sgnf = 1.0 if !have_subnormal && $exp == min_normal_exp && |
319
|
|
|
|
|
|
|
$bot_sgnf < 1.0; |
320
|
9000
|
|
|
|
|
23798
|
while($bdone != significand_bits) { |
321
|
9299
|
|
|
|
|
22181
|
my $bseg = significand_bits - $bdone; |
322
|
9299
|
100
|
|
|
|
19393
|
$bseg = 28 if $bseg > 28; |
323
|
9299
|
|
|
|
|
14317
|
$bdone += $bseg; |
324
|
9299
|
|
|
|
|
20848
|
$sgnf += mult_pow2(rand_int(1 << $bseg), -$bdone); |
325
|
|
|
|
|
|
|
} |
326
|
|
|
|
|
|
|
} |
327
|
9000
|
50
|
33
|
|
|
308397
|
goto TRY_AGAIN if $sgnf < $bot_sgnf || $sgnf >= $top_sgnf; |
328
|
9000
|
50
|
33
|
|
|
23123
|
$sgnf = $top_sgnf if $sgnf == $bot_sgnf && rand_int(2); |
329
|
9000
|
50
|
|
|
|
31905
|
return copysign($sgnf == 0.0 ? 0.0 : mult_pow2($sgnf, $exp), $sg); |
330
|
|
|
|
|
|
|
} |
331
|
|
|
|
|
|
|
|
332
|
|
|
|
|
|
|
=back |
333
|
|
|
|
|
|
|
|
334
|
|
|
|
|
|
|
=head2 Combinatorics |
335
|
|
|
|
|
|
|
|
336
|
|
|
|
|
|
|
=over |
337
|
|
|
|
|
|
|
|
338
|
|
|
|
|
|
|
=item pick(ITEM ...) |
339
|
|
|
|
|
|
|
|
340
|
|
|
|
|
|
|
Randomly selects and returns one of the ITEMs. Each ITEM has equal |
341
|
|
|
|
|
|
|
probability of being selected. |
342
|
|
|
|
|
|
|
|
343
|
|
|
|
|
|
|
=item pick_r(ITEMS) |
344
|
|
|
|
|
|
|
|
345
|
|
|
|
|
|
|
ITEMS must be a reference to an array. Randomly selects and returns |
346
|
|
|
|
|
|
|
one of the elements of the array. Each element has equal probability |
347
|
|
|
|
|
|
|
of being selected. |
348
|
|
|
|
|
|
|
|
349
|
|
|
|
|
|
|
This is the same operation as that performed by C, but using |
350
|
|
|
|
|
|
|
references to avoid expensive copying of arrays. |
351
|
|
|
|
|
|
|
|
352
|
|
|
|
|
|
|
=cut |
353
|
|
|
|
|
|
|
|
354
|
|
|
|
|
|
|
sub pick_r($) { |
355
|
155
|
|
|
155
|
1
|
44692
|
my($a) = @_; |
356
|
155
|
100
|
100
|
|
|
1427
|
croak "need a non-empty array to pick from" |
357
|
|
|
|
|
|
|
unless is_ref($a, "ARRAY") && @$a; |
358
|
152
|
|
|
|
|
356
|
return $a->[rand_int(@$a)]; |
359
|
|
|
|
|
|
|
} |
360
|
|
|
|
|
|
|
|
361
|
77
|
|
|
77
|
1
|
46523
|
sub pick(@) { pick_r(\@_) } |
362
|
|
|
|
|
|
|
|
363
|
|
|
|
|
|
|
=item choose(NCHOOSE, ITEM ...) |
364
|
|
|
|
|
|
|
|
365
|
|
|
|
|
|
|
Randomly selects NCHOOSE of the ITEMs. Each ITEM has equal probability |
366
|
|
|
|
|
|
|
of being selected. The chosen items are returned in a list in the same |
367
|
|
|
|
|
|
|
order in which they appeared in the argument list. |
368
|
|
|
|
|
|
|
|
369
|
|
|
|
|
|
|
=item choose_r(NCHOOSE, ITEMS) |
370
|
|
|
|
|
|
|
|
371
|
|
|
|
|
|
|
ITEMS must be a reference to an array. Randomly selects NCHOOSE of |
372
|
|
|
|
|
|
|
the elements in the array. Each element has equal probability of being |
373
|
|
|
|
|
|
|
selected. Returns a reference to an array containing the chosen items |
374
|
|
|
|
|
|
|
in the same order in which they appeared in the input array. |
375
|
|
|
|
|
|
|
|
376
|
|
|
|
|
|
|
This is the same operation as that performed by C, but using |
377
|
|
|
|
|
|
|
references to avoid expensive copying of arrays. |
378
|
|
|
|
|
|
|
|
379
|
|
|
|
|
|
|
=cut |
380
|
|
|
|
|
|
|
|
381
|
|
|
|
|
|
|
sub choose_r($$) { |
382
|
51
|
|
|
51
|
1
|
21766
|
my($nchoose, $a) = @_; |
383
|
51
|
100
|
|
|
|
583
|
croak "need a non-negative number of items to choose" |
384
|
|
|
|
|
|
|
unless $nchoose >= 0; |
385
|
49
|
100
|
100
|
|
|
955
|
croak "need a sufficiently large array to pick from" |
386
|
|
|
|
|
|
|
unless is_ref($a, "ARRAY") && @$a >= $nchoose; |
387
|
46
|
|
|
|
|
61
|
my $ntotal = @$a; |
388
|
46
|
|
|
|
|
75
|
my $nleave = $ntotal - $nchoose; |
389
|
46
|
|
|
|
|
57
|
my @chosen; |
390
|
46
|
|
|
|
|
113
|
for(my $i = 0; $i != $ntotal; $i++) { |
391
|
414
|
100
|
|
|
|
1002
|
if(entropy_source->get_prob($nleave, $nchoose)) { |
392
|
128
|
|
|
|
|
246
|
push @chosen, $a->[$i]; |
393
|
128
|
|
|
|
|
367
|
$nchoose--; |
394
|
|
|
|
|
|
|
} else { |
395
|
286
|
|
|
|
|
777
|
$nleave--; |
396
|
|
|
|
|
|
|
} |
397
|
|
|
|
|
|
|
} |
398
|
46
|
|
|
|
|
303
|
return \@chosen; |
399
|
|
|
|
|
|
|
} |
400
|
|
|
|
|
|
|
|
401
|
25
|
|
|
25
|
1
|
12996
|
sub choose(@) { @{choose_r(shift, \@_)} } |
|
25
|
|
|
|
|
82
|
|
402
|
|
|
|
|
|
|
|
403
|
|
|
|
|
|
|
=item shuffle(ITEM ...) |
404
|
|
|
|
|
|
|
|
405
|
|
|
|
|
|
|
Reorders the ITEMs randomly, and returns them in a list in random order. |
406
|
|
|
|
|
|
|
Each possible order has equal probability. |
407
|
|
|
|
|
|
|
|
408
|
|
|
|
|
|
|
=item shuffle_r(ITEMS) |
409
|
|
|
|
|
|
|
|
410
|
|
|
|
|
|
|
ITEMS must be a reference to an array. Reorders the elements of the |
411
|
|
|
|
|
|
|
array randomly. Each possible order has equal probability. Returns a |
412
|
|
|
|
|
|
|
reference to an array containing the elements in random order. |
413
|
|
|
|
|
|
|
|
414
|
|
|
|
|
|
|
This is the same operation as that performed by C, but using |
415
|
|
|
|
|
|
|
references to avoid expensive copying of arrays. |
416
|
|
|
|
|
|
|
|
417
|
|
|
|
|
|
|
=cut |
418
|
|
|
|
|
|
|
|
419
|
|
|
|
|
|
|
sub shuffle_r($) { |
420
|
19
|
|
|
19
|
1
|
5782
|
my($a) = @_; |
421
|
19
|
100
|
|
|
|
326
|
croak "need an array to shuffle" |
422
|
|
|
|
|
|
|
unless is_ref($a, "ARRAY"); |
423
|
18
|
|
|
|
|
64
|
$a = [ @$a ]; |
424
|
18
|
|
|
|
|
62
|
for(my $i = @$a; $i > 1; ) { |
425
|
126
|
|
|
|
|
322
|
my $j = rand_int($i--); |
426
|
126
|
|
|
|
|
193
|
@{$a}[$i, $j] = @{$a}[$j, $i]; |
|
126
|
|
|
|
|
14973
|
|
|
126
|
|
|
|
|
210
|
|
427
|
|
|
|
|
|
|
} |
428
|
18
|
|
|
|
|
146
|
return $a; |
429
|
|
|
|
|
|
|
} |
430
|
|
|
|
|
|
|
|
431
|
9
|
|
|
9
|
1
|
5023
|
sub shuffle(@) { @{shuffle_r(\@_)} } |
|
9
|
|
|
|
|
32
|
|
432
|
|
|
|
|
|
|
|
433
|
|
|
|
|
|
|
=back |
434
|
|
|
|
|
|
|
|
435
|
|
|
|
|
|
|
=head1 SEE ALSO |
436
|
|
|
|
|
|
|
|
437
|
|
|
|
|
|
|
L, |
438
|
|
|
|
|
|
|
L |
439
|
|
|
|
|
|
|
|
440
|
|
|
|
|
|
|
=head1 AUTHOR |
441
|
|
|
|
|
|
|
|
442
|
|
|
|
|
|
|
Andrew Main (Zefram) |
443
|
|
|
|
|
|
|
|
444
|
|
|
|
|
|
|
=head1 COPYRIGHT |
445
|
|
|
|
|
|
|
|
446
|
|
|
|
|
|
|
Copyright (C) 2006, 2007, 2009, 2011 |
447
|
|
|
|
|
|
|
Andrew Main (Zefram) |
448
|
|
|
|
|
|
|
|
449
|
|
|
|
|
|
|
=head1 LICENSE |
450
|
|
|
|
|
|
|
|
451
|
|
|
|
|
|
|
This module is free software; you can redistribute it and/or modify it |
452
|
|
|
|
|
|
|
under the same terms as Perl itself. |
453
|
|
|
|
|
|
|
|
454
|
|
|
|
|
|
|
=cut |
455
|
|
|
|
|
|
|
|
456
|
|
|
|
|
|
|
1; |