1GPMETIS(1) User Commands GPMETIS(1)
2
3
4
6 gpmetis - Partitions a graph into a specified number of parts.
7
9 gpmetis [options] graphfile nparts
10
12 Required parameters
13
14 graphfile
15 Stores the graph to be partitioned.
16
17 nparts The number of partitions to split the graph.
18
19 Optional parameters
20
21 -ptype=string
22
23 Specifies the scheme to be used for computing the k-way parti‐
24 tioning. The possible values are:
25
26 rb - Recursive bisectioning
27
28 kway - Direct k-way partitioning [default]
29
30 -ctype=string
31
32 Specifies the scheme to be used to match the vertices of the
33 graph during the coarsening. The possible values are:
34
35 rm - Random matching
36
37 shem - Sorted heavy-edge matching [default]
38
39 -iptype=string [applies only when -ptype=rb]
40
41 Specifies the scheme to be used to compute the initial parti‐
42 tioning of the graph. The possible values are:
43
44 grow - Grow a bisection using a greedy scheme [default for ncon=1]
45
46 random - Compute a bisection at random [default for ncon>1]
47
48 -objtype=string [applies only when -ptype=kway]
49
50 Specifies the objective that the partitioning routines will
51 optimize. The possible values are:
52
53 cut - Minimize the edgecut [default]
54
55 vol - Minimize the total communication volume
56
57 -no2hop
58
59 Specifies that the coarsening will not perform any 2-hop match‐
60 ings when the standard matching fails to sufficiently contract
61 the graph.
62
63 -contig [applies only when -ptype=kway]
64
65 Specifies that the partitioning routines should try to produce
66 partitions that are contiguous. Note that if the input graph is
67 not connected this option is ignored.
68
69 -minconn [applies only when -ptype=kway]
70
71 Specifies that the partitioning routines should try to minimize
72 the maximum degree of the subdomain graph, i.e., the graph in
73 which each partition is a node, and edges connect subdomains
74 with a shared interface.
75
76 -tpwgts=filename
77
78 Specifies the name of the file that stores the target weights
79 for each partition. By default, all partitions are assumed to be
80 of the same size.
81
82 -ufactor=int
83
84 Specifies the maximum allowed load imbalance among the parti‐
85 tions. A value of x indicates that the allowed load imbalance
86 is 1+x/1000. For ptype=rb, the load imbalance is measured as
87 the ratio of the 2*max(left,right)/(left+right), where left and
88 right are the sizes of the respective partitions at each bisec‐
89 tion. For ptype=kway, the load imbalance is measured as the
90 ratio of max_i(pwgts[i])/avgpwgt, where pwgts[i] is the weight
91 of the ith partition and avgpwgt is the sum of the total vertex
92 weights divided by the number of partitions requested. For
93 ptype=rb, the default value is 1 (i.e., load imbalance of
94 1.001). For ptype=kway, the default value is 30 (i.e., load
95 imbalance of 1.03).
96
97 -ubvec=string
98
99 Applies only for multi-constraint partitioning and specifies the
100 per constraint allowed load imbalance among partitions. The
101 required parameter corresponds to a space separated set of
102 floating point numbers, one for each of the constraints. For
103 example, for three constraints, the string can be "1.02 1.2
104 1.35" indicating a desired maximum load imbalance of 2%, 20%,
105 and 35%, respectively. The load imbalance is defined in a way
106 similar to ufactor. If supplied, this parameter takes priority
107 over ufactor.
108
109 -niter=int
110
111 Specifies the number of iterations for the refinement algorithms
112 at each stage of the uncoarsening process. Default is 10.
113
114 -ncuts=int
115
116 Specifies the number of different partitionings that it will
117 compute. The final partitioning is the one that achieves the
118 best edgecut or communication volume. Default is 1.
119
120 -nooutput
121
122 Specifies that no partitioning file should be generated.
123
124 -seed=int
125
126 Selects the seed of the random number generator.
127
128 -dbglvl=int
129
130 Selects the dbglvl.
131
132 -help
133
134 Prints this message.
135
136
137
138gpmetis 5.1.0 February 2019 GPMETIS(1)