1
2gmap(1) Scotch user's manual gmap(1)
3
4
5
7 gmap, gpart - compute static mappings and partitions sequentially
8
10 gmap [options] [gfile] [tfile] [mfile] [lfile]
11
12 gpart [options] [nparts] [gfile] [mfile] [lfile]
13
15 The gmap program computes, in a sequential way, a static mapping of a
16 source graph onto a target graph.
17
18 The gpart program is a shortcut of gmap for computing unweighted parti‐
19 tions of a source graph.
20
21 Source graph file gfile can only be a centralized graph file. For gmap,
22 the target architecture file tfile describes either algorithmically-
23 coded topologies such as meshes and hypercubes, or decomposition-
24 defined architectures created by means of the amk_grf(1) program. The
25 resulting mapping is stored in file mfile. Eventual logging information
26 (such as the one produced by option -v) is sent to file lfile. When
27 file names are not specified, data is read from standard input and
28 written to standard output. Standard streams can also be explicitely
29 represented by a dash '-'.
30
31 When the proper libraries have been included at compile time, gmap and
32 gpart can directly handle compressed graphs, both as input and output.
33 A stream is treated as compressed whenever its name is postfixed with a
34 compressed file extension, such as in 'brol.grf.bz2' or '-.gz'. The
35 compression formats which can be supported are the bzip2 format
36 ('.bz2'), the gzip format ('.gz'), and the lzma format ('.lzma', on
37 input only).
38
40 -copt Choose default mapping strategy according to one or several
41 options among:
42
43 b enforce load balance as much as possible.
44
45 q privilege quality over speed (default).
46
47 s privilege speed over quality.
48
49 t enforce safety.
50
51 -h Display some help.
52
53 -mstrat
54 Use sequential mapping strategy strat (see Scotch user's manual
55 for more information).
56
57 -V Display program version and copyright.
58
59 -vverb Set verbose mode to verb. It is a set of one of more characters
60 which can be:
61
62 m mapping information.
63
64 s strategy information.
65
66 t timing information.
67
69 Target architectures represent graphs onto which source graphs are
70 mapped. In order to speed-up the obtainment of target architecture
71 topological properties during the computation of mappings, some classi‐
72 cal topologies are algorithmically coded into the mapper itself. These
73 topologies are consequently simply defined by their code name, followed
74 by their dimensional parameters:
75
76 cmplt dim
77 unweighted complete graph of size dim.
78
79 cmpltw dim w0 w1 ... wdim-1
80 weighted complete graph of size size and of respective loads w0,
81 w1, ..., wdim-1.
82
83 hcub dim
84 hypercube of dimension dim.
85
86 leaf hgt n0 w0 ... nhgt-1 whgt-1
87 tree-leaf graph of height hgt with (n0 times n1 times ...
88 nhgt-1) vertices, with inter-cluster link weights of w0, w1, ...
89 whgt-1.
90
91 mesh2D dimX dimY
92 2D mesh of dimX times dimY nodes.
93
94 mesh3D dimX dimY dimZ
95 23 mesh of dimX times dimY times dimZ nodes.
96
97 torus2D dimX dimY
98 2D torus of dimX times dimY nodes.
99
100 torus3D dimX dimY dimZ
101 3D torus of dimX times dimY times dimZ nodes.
102
103 Other target topologies can be created from their source graph descrip‐
104 tion by using the amk_grf(1) command. In this case, the target descrip‐
105 tion will begin with the code name deco.
106
108 Mappings are represented by as many lines as there are vertices in the
109 source graph. Each of these lines is made of two figures: the number of
110 the vertex (or its label if source graph vertices are labeled) and the
111 index of the target vertex to which it has been assigned. Target vertex
112 indices range from 0 to the number of vertices in the target architec‐
113 ture (that is, the number of parts) minus one.
114
115 This block of lines is always preceded by the number of such lines. In
116 most cases, since full mappings are requested, the number of lines is
117 equal to the number of vertices in the source graph.
118
120 Run gpart to compute a partition into 7 parts of graph 'brol.grf' and
121 save the resulting ordering to file 'brol.map'.
122
123 $ gpart 7 brol.grf brol.map
124
125 Run gmap to compute a partition, into 3 parts of respective weights 1,
126 2 and 4, of graph 'brol.grf' and save the resulting mapping to file
127 'brol.map'. The dash '-' standard file name is used so that the target
128 architecture description is read from the standard input, through the
129 pipe, as provided by the 'echo' shell command.
130
131 $ echo "cmpltw 3 1 2 4" | gmap brol.grf - brol.map
132
133
135 amk_grf(1), acpl(1), gmtst(1), dgmap(1).
136
137 Scotch user's manual.
138
140 Francois Pellegrini <francois.pellegrini@labri.fr>
141
142
143
144 August 03, 2010 gmap(1)