1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
|
.\" These macros were copied from MPlayer manpage, written by Gabucino,
.\" Diego Biurrun and Jonas Jermann.
..
.\" default indentation is 7, don't change!
.nr IN 7
.\" define indentation for suboptions
.nr SS 5
.\" add new suboption
.de IPs
.IP "\\$1" \n(SS
..
.\" begin of first level suboptions, end with .RE
.de RSs
.RS \n(IN+3
..
.TH TABLIX 1 2005-09-03 "Tomaz Solc" "Tablix User's Manual"
.SH NAME
tablix2_benchmark \- Tablix benchmark utility
.SH SYNOPSIS
.B tablix2_benchmark
.I benchmark
.B [
.I options
.B ]
\(dq
.I Tablix options
\(dq
.I file
.SH DESCRIPTION
Tablix is a powerful free software kernel for solving general timetabling problems. It uses a coarse-grained parallel genetic algorithm in combination with other techniques to construct sensible timetables from XML formatted problem descriptions. Tablix can run on a single host as well as on a heterogeneous parallel virtual machine using PVM3.
.P
.B tablix2_benchmark
is used to test the effect of different Tablix command line options on the final result. In most tests Tablix is started multiple times with the same input file to provide more reliable results. Each time Tablix is started, three variables are measured: the fitness value of the resulting timetable, serial number of the last generation and elapsed time (in seconds). When a test is finished, the average for each variable, as well as the maximum and minimum values are printed on standard output. The exact format depends on each test.
.P
The actual timetables that are generated during these tests are stored under a temporary file names in the current directory and are deleted after
.B tablix2_benchmark
exits.
.P
In some cases tests can take several days to complete. Because the genetic algorithm can in some cases go into an endless loop, it is always advisable to limit Tablix execution time with the
.B -t
option. This way at least the maximum execution time for a test can be calculated.
.B tablix2_benchmark
will detect when the time limit set by the
.B -t
option was reached. This information will be included in the final statistics.
.SH OPTIONS
The following tests are supported:
.TP
.B tablix2_benchmark --single \(dqOPTIONS\(dq file
Execute a single Tablix run using and print the results.
.B OPTIONS
parameter is mandatory and contains options that will be passed to the
.B tablix2
process.
.TP
.B tablix2_benchmark --multiple N \(dqOPTIONS\(dq FILE
Execute N Tablix runs and print the statistics.
.B OPTIONS
parameter is mandatory and contains options that will be passed to the
.B tablix2
process.
.TP
.B tablix2_benchmark --graph START STEP STOP \(dqOPTIONS\(dq FILE
Execute multiple Tablix runs. The character \(dq
.B N
\(dq in
.B OPTIONS
is substituted for an integer that is changed from
.B START
to
.B STOP
by
.B STEP
\&. Output is gnuplot friendly.
.TP
.B tablix2_benchmark --multi-graph M START STEP STOP \(dqOPTIONS\(dq FILE
Execute multiple Tablix runs. The character \(dq
.B N
\(dq in
.B OPTIONS
is substituted for an integer that is changed from
.B START
to
.B STOP
by
.B STEP
\&.
.B M
Tablix runs are executed for each value of
.B N
\&. Output is gnuplot friendly (use errorbars).
.SH EXAMPLES
The following test can be for example used to see the minimum grade, that can be reached with this configuration file:
.P
tablix2_benchmark --multiple 5 \(dq-t 90 -n 5\(dq sample.xml
.P
Tablix will be started five times. Each time with five computing nodes and with 90 minute time limit.
.P
.B
tablix2_benchmark
can also provide useful data for developers. Following test can be used to check the effect of the number of computing nodes on the total time required to obtain the result:
.P
tablix2_benchmark --multi-graph 5 1 1 10 \(dq-t 90 -n N\(dq sample.xml > stats.txt
.P
Tablix will be started 50 times. 5 times for each value of N (which in this case means the number of nodes). Ten values of N will be probed: from 1 to 10 with steps of 1. The output will be redirected to
.B stats.txt
\&. A graph of average computing time versus number of nodes can then be produced in
.B gnuplot
with the following command:
.P
plot \(dqtest.txt\(dq using ($1):($8):($9):($10) with yerrorbars
.SH BUGS
.B tablix2_benchmark
utility depends on a number of common UNIX programs (mktemp, grep, sed, awk, ...) that may not be available on all systems. The script is not currently capable of properly detecting if any required programs are missing.
.SH AUTHOR
Tomaz Solc (tomaz.solc@tablix.org)
.SH SEE ALSO
.BR tablix2 (1),
Tablix User's Manual,
Tablix modules HOWTO,
Tablix on Morphix HOWTO
|
