1 |
.\" |
2 |
.\" Copyright 1998 by Luigi Rizzo, Dip. Ingegneria dell'Informazione, |
3 |
.\" Universitaet Berlin. See the source code for copyright details. |
4 |
.\" THERE IS ABSOLUTELY NO WARRANTY FOR THIS SOFTWARE. |
5 |
.\" |
6 |
.Dd July 15, 1998 |
7 |
.Dt FEC 3 |
8 |
.Os |
9 |
.Sh NAME |
10 |
.Nm fec_new, fec_encode, fec_encode, fec_free |
11 |
.Nd An erasure code in GF(2^m) |
12 |
.Sh SYNOPSIS |
13 |
.Fd #include <fec.h> |
14 |
.Ft void * |
15 |
.Fn fec_new "int k" "int n" |
16 |
.Ft void |
17 |
.Fn fec_encode "void *code" "void *data[]" "void *dst" "int i" "int sz" |
18 |
.Ft int |
19 |
.Fn fec_decode "void *code" "void *data[]" "int i[]" "int sz" |
20 |
.Ft void * |
21 |
.Fn fec_free "void *code" |
22 |
.Sh "DESCRIPTION" |
23 |
This library implements a simple (n,k) |
24 |
erasure code based on Vandermonde matrices. |
25 |
The encoder takes |
26 |
.Fa k |
27 |
packets of size |
28 |
.Fa sz |
29 |
each, and is able to produce up to |
30 |
.Fa n |
31 |
different encoded packets, numbered from 0 to n-1, |
32 |
such that any subset of |
33 |
.Fa k |
34 |
of them permits reconstruction of the original data. |
35 |
.Pp |
36 |
The data structures necessary for the encoding/decoding must |
37 |
first be created using calling |
38 |
.Fn fec_new |
39 |
with the desired parameters. The code descriptor returned by the function |
40 |
must be passed to other functions, and destroyed calling |
41 |
.Fn fec_free |
42 |
.Pp |
43 |
Allowed values for k and n depend on a compile-time value |
44 |
of |
45 |
.Fa GF_BITS |
46 |
and must be k <= n <= 2^GF_BITS. |
47 |
Best performance is achieved with GF_BITS=8, although the code supports |
48 |
also GF_BITS=16. |
49 |
.Pp |
50 |
Encoding is done by calling |
51 |
.Fn fec_encode |
52 |
and passing it pointers to the code descriptor, the source and |
53 |
destination data packets, the index of the packet to be produced, |
54 |
and the size of the packet. |
55 |
|
56 |
.Pp Decoding is done calling |
57 |
.Fn fec_decode |
58 |
with pointers to the code, received packets, indexes of received |
59 |
packets, and packet size. Decoding is done in place, possibly |
60 |
shuffling the arrays passed as parameters. Decoding is deterministic |
61 |
as long as the received packets are different. The decoding procedure |
62 |
does some limited testing on this and returns if parameters are |
63 |
invalid. |
64 |
|
65 |
.Sh EXAMPLE |
66 |
.nf |
67 |
#include <fec.h> |
68 |
|
69 |
/* |
70 |
* example of sender code |
71 |
*/ |
72 |
void *code ; |
73 |
int n, k ; |
74 |
|
75 |
void *src[] ; |
76 |
void *pkt ; |
77 |
|
78 |
code = new_code (k, n ); |
79 |
|
80 |
for (i = 0 ; i < k ; i++ ) |
81 |
src[i] = .. pointer to i-th source packet .. |
82 |
for (each packet to transmit) { |
83 |
i = ... index of the packet ; |
84 |
fec_encode(code, src, pkt, i, size) ; |
85 |
.. use packet in pkt |
86 |
} |
87 |
fec_free(code) ; |
88 |
|
89 |
/* |
90 |
* example of receiver code |
91 |
*/ |
92 |
void *code ; |
93 |
int n, k ; |
94 |
|
95 |
void *data[] ; |
96 |
int *ix[] ; |
97 |
|
98 |
code = new_code (k, n ); |
99 |
|
100 |
for (i = 0 ; i < k ; i++ ) { |
101 |
... receive a new packet ... |
102 |
data[i] = .. pointer to i-th source packet .. |
103 |
ix[i] = .. index of i-th source packet .. |
104 |
} |
105 |
fec_decode(code, data, ix, size) ; |
106 |
/* |
107 |
* now data[] has pointers to the source packets |
108 |
*/ |
109 |
|
110 |
.SH BUGS |
111 |
Please direct bug reports to luigi@iet.unipi.it . |
112 |
.Sh "SEE ALSO" |