[unix-history] / usr / src / lib / libc / stdlib / random.3

.\" Copyright (c) 1983 Regents of the University of California.
.\" All rights reserved.  The Berkeley software License Agreement
.\" specifies the terms and conditions for redistribution.
.\"
.\"	@(#)random.3	5.1 (Berkeley) %G%
.\"
.TH RANDOM 3 "19 January 1983"
.UC 5
.SH NAME
random, srandom, initstate, setstate \- better random number generator; routines for changing generators
.SH SYNOPSIS
.nf
.B long  random()
.PP
.B srandom(seed)
.B int  seed;
.PP
.B char  *initstate(seed, state, n)
.B unsigned  seed;
.B char  *state;
.B int  n;
.PP
.B char  *setstate(state)
.B char  *state;
.fi
.SH DESCRIPTION
.PP
.I Random
uses a non-linear additive feedback random number generator employing a
default table of size 31 long integers to return successive pseudo-random
numbers in the range from 0 to 2\u\s731\s10\d\-1.  The period of this
random number generator is very large, approximately 16*(2\u\s731\s10\d\-1).
.PP
.I Random/srandom
have (almost) the same calling sequence and initialization properties as
.I rand/srand.
The difference is that
.IR rand (3)
produces a much less random sequence -- in fact, the low dozen bits
generated by rand go through a cyclic pattern.  All the bits generated by
.I random
are usable.  For example, \*(lqrandom()&01\*(rq will produce a random binary
value.
.PP
Unlike
.IR srand ,
.I srandom
does not return the old seed; the reason for this is that the amount of
state information used is much more than a single word.  (Two other
routines are provided to deal with restarting/changing random
number generators).  Like
.IR rand (3),
however,
.I random
will by default produce a sequence of numbers that can be duplicated
by calling
.I srandom
with 
.I 1
as the seed.
.PP
The
.I initstate
routine allows a state array, passed in as an argument, to be initialized
for future use.  The size of the state array (in bytes) is used by
.I initstate
to decide how sophisticated a random number generator it should use -- the
more state, the better the random numbers will be.
(Current "optimal" values for the amount of state information are
8, 32, 64, 128, and 256 bytes; other amounts will be rounded down to
the nearest known amount.  Using less than 8 bytes will cause an error).
The seed for the initialization (which specifies a starting point for
the random number sequence, and provides for restarting at the same
point) is also an argument.
.I Initstate
returns a pointer to the previous state information array.
.PP
Once a state has been initialized, the
.I setstate
routine provides for rapid switching between states.
.I Setstate returns a pointer to the previous state array; its
argument state array is used for further random number generation
until the next call to
.I initstate
or
.I setstate.
.PP
Once a state array has been initialized, it may be restarted at a
different point either by calling
.I initstate
(with the desired seed, the state array, and its size) or by calling
both
.I setstate
(with the state array) and
.I srandom
(with the desired seed).
The advantage of calling both
.I setstate
and
.I srandom
is that the size of the state array does not have to be remembered after
it is initialized.
.PP
With 256 bytes of state information, the period of the random number
generator is greater than 2\u\s769\s10\d, which should be sufficient for
most purposes.
.SH AUTHOR
Earl T. Cohen
.SH DIAGNOSTICS
.PP
If
.I initstate
is called with less than 8 bytes of state information, or if
.I setstate
detects that the state information has been garbled, error
messages are printed on the standard error output.
.SH "SEE ALSO"
rand(3)
.SH BUGS
About 2/3 the speed of
.IR rand (3C).
Commit	Line	Data
7db0d200 KM	1	.\" Copyright (c) 1983 Regents of the University of California.
	2	.\" All rights reserved. The Berkeley software License Agreement
	3	.\" specifies the terms and conditions for redistribution.
	4	.\"
	5	.\" @(#)random.3 5.1 (Berkeley) %G%
	6	.\"
	7	.TH RANDOM 3 "19 January 1983"
	8	.UC 5
	9	.SH NAME
	10	random, srandom, initstate, setstate \- better random number generator; routines for changing generators
	11	.SH SYNOPSIS
	12	.nf
	13	.B long random()
	14	.PP
	15	.B srandom(seed)
	16	.B int seed;
	17	.PP
	18	.B char *initstate(seed, state, n)
	19	.B unsigned seed;
	20	.B char *state;
	21	.B int n;
	22	.PP
	23	.B char *setstate(state)
	24	.B char *state;
	25	.fi
	26	.SH DESCRIPTION
	27	.PP
	28	.I Random
	29	uses a non-linear additive feedback random number generator employing a
	30	default table of size 31 long integers to return successive pseudo-random
	31	numbers in the range from 0 to 2\u\s731\s10\d\-1. The period of this
	32	random number generator is very large, approximately 16*(2\u\s731\s10\d\-1).
	33	.PP
	34	.I Random/srandom
	35	have (almost) the same calling sequence and initialization properties as
	36	.I rand/srand.
	37	The difference is that
	38	.IR rand (3)
	39	produces a much less random sequence -- in fact, the low dozen bits
	40	generated by rand go through a cyclic pattern. All the bits generated by
	41	.I random
	42	are usable. For example, \(lqrandom()&01\(rq will produce a random binary
	43	value.
	44	.PP
	45	Unlike
	46	.IR srand ,
	47	.I srandom
	48	does not return the old seed; the reason for this is that the amount of
	49	state information used is much more than a single word. (Two other
	50	routines are provided to deal with restarting/changing random
	51	number generators). Like
	52	.IR rand (3),
	53	however,
	54	.I random
	55	will by default produce a sequence of numbers that can be duplicated
	56	by calling
	57	.I srandom
	58	with
	59	.I 1
	60	as the seed.
	61	.PP
	62	The
	63	.I initstate
	64	routine allows a state array, passed in as an argument, to be initialized
65	for future use. The size of the state array (in bytes) is used by
66	.I initstate
67	to decide how sophisticated a random number generator it should use -- the
68	more state, the better the random numbers will be.
69	(Current "optimal" values for the amount of state information are
70	8, 32, 64, 128, and 256 bytes; other amounts will be rounded down to
71	the nearest known amount. Using less than 8 bytes will cause an error).
72	The seed for the initialization (which specifies a starting point for
73	the random number sequence, and provides for restarting at the same
74	point) is also an argument.
75	.I Initstate
76	returns a pointer to the previous state information array.
77	.PP
78	Once a state has been initialized, the
79	.I setstate
80	routine provides for rapid switching between states.
81	.I Setstate returns a pointer to the previous state array; its
82	argument state array is used for further random number generation
83	until the next call to
84	.I initstate
85	or
86	.I setstate.
87	.PP
88	Once a state array has been initialized, it may be restarted at a
89	different point either by calling
90	.I initstate
91	(with the desired seed, the state array, and its size) or by calling
92	both
93	.I setstate
94	(with the state array) and
95	.I srandom
96	(with the desired seed).
97	The advantage of calling both
98	.I setstate
99	and
100	.I srandom
101	is that the size of the state array does not have to be remembered after
102	it is initialized.
103	.PP
104	With 256 bytes of state information, the period of the random number
105	generator is greater than 2\u\s769\s10\d, which should be sufficient for
106	most purposes.
107	.SH AUTHOR
108	Earl T. Cohen
109	.SH DIAGNOSTICS
110	.PP
111	If
112	.I initstate
113	is called with less than 8 bytes of state information, or if
114	.I setstate
115	detects that the state information has been garbled, error
116	messages are printed on the standard error output.
117	.SH "SEE ALSO"
118	rand(3)
119	.SH BUGS
120	About 2/3 the speed of
121	.IR rand (3C).