5
## Options for the flredecomp program
8
## Select which partitioner to use when flredecomping.
9
## Graph partitioners available are ParMETIS, PT-Scotch
10
## and Zoltan PHG. The Zoltan PHG hypergraph partitoner
12
## Default is ParMETIS graph partitioner.
13
element final_partitioner {
14
## Use the ParMETIS graph partitioner. ParMETIS setup to match as
15
## closely as possible the setup used previously by Sam.
19
## Use the PT-Scotch graph partitioner.
23
## Use the Zoltan PHG partitioner.
25
## Select the partitioning method you would like used by Zoltan PHG.
26
## Currently hypergraph partitioning is the simplest implementation
27
## and can produce non-contiguous partitions for certain problems.
35
## Turn on more verbose output for use when debugging Zoltan.
36
element zoltan_debug {
37
## Turn on graph checking.
38
## When using ParMETIS or PT-Scotch options for turning on
39
## graph checking are provided by Zoltan.
40
## 1 - on process checking,
41
## 2 - full checking (very slow)
42
element graph_checking {
45
## Print out a dump file of the edge counts.
46
## Edge counts for each owned node are calculated in zoltan_cb_get_num_edges.
47
## This option dumps the edge count for each owned node.
48
## Dump is to the current directory, in a file called edge_counts_*.dat
49
## One dump file is created for each rank.
50
element dump_edge_counts {
53
## Print out a dump file of node sizes.
54
## Zoltan needs to be told how much data is associated with each node when
55
## doing phase one migration.
56
## Here we dump the size calculated by zoltan_cb_pack_node_sizes for each
58
## Dump is to the current directory, in a file called node_sizes_*.dat
59
## One dump file is created for each rank.
60
element dump_node_sizes {
63
## Print out a dump file of halo node sizes.
64
## Zoltan needs to be told how much data is associated with each halo node when
65
## doing phase two migration.
66
## Here we dump the size calculated by zoltan_cb_pack_halo_node_sizes for each
68
## Dump is to the current directory, in a file called halo_node_sizes_*.dat
69
## One dump file is created for each rank.
70
element dump_halo_node_sizes {
73
## Print out a dump file of field sizes.
74
## Zoltan needs to be told how much data is associated with the fields for each
75
## element when we're transfering fields.
76
## Here we dump the size calculated by zoltan_cb_pack_field_sizes for each
78
## Dump is to the current directory, in a file called field_sizes_*.dat
79
## One dump file is created for each rank.
80
element dump_field_sizes {