 |
Mesh Oriented datABase
(version 5.5.1)
An array-based unstructured mesh library
|
|
Mesh Oriented datABase
(version 5.5.1)
An array-based unstructured mesh library
|
1 /** @example GenLargeMesh.cpp \n
2 * \brief Create a large structured mesh, partitioned \n
3 *
4 * It shows how to create a mesh on the fly, on multiple processors
5 * Each processor will create its version of a block mesh, partitioned
6 * as AxBxC blocks. Each block will be with blockSize^3 hexahedrons, and
7 * will get a different PARALLEL_PARTITION tag
8 *
9 * The number of tasks will be MxNxK, and it must match the mpi size
10 * Each task p will generate its mesh at location (m,n,k), and it is
11 * lexicographic ordering: rank = m + n * M + k * M * N
12 *
13 * By default M=1, N=1, K=1, so by default it should be launched on 1 proc
14 * By default, blockSize is 4, and A=2, B=2, C=2, so each task will generate locally
15 * blockSize^3 x A x B x C hexahedrons (value = 64x8 = 512 hexas, in 8 partitions)
16 * (if -t, multiple by 6 for total number of cells/tets)
17 * The total number of partitions will be A*B*C*M*N*K (default 8)
18 *
19 * Each part in partition will get a proper tag, and the numbering is in
20 * lexicographic ordering; x direction varies first, then y, then z.
21 * The same principle is used for global id numbering of the nodes and cells.
22 * (x varies first)
23 *
24 * The vertices will get a proper global id, which will be used to resolve the
25 * shared entities
26 * The output will be written in parallel, and we will try sizes as big as we can
27 * (up to a billion vertices, because we use int for global ids)
28 *
29 * Within each partition, the hexas entity handles will be contiguous, and also the
30 * vertices; The global id will be determined easily, for a vertex, but the entity
31 * handle space will be more interesting to handle, within a partition (we want
32 * contiguous handles within a partition). After merging the vertices, some fragmentation
33 * will occur in the vertices handle space, within each partition.
34 *
35 * To run: ./GenLargeMesh
36 *
37 * When launched on more procs, you have to make sure
38 * num procs = M*N*K
39 *
40 * So you can launch with
41 * mpiexec -np 8 ./GenLargeMesh -M 2 -N 2 -K 2
42 *
43 * We also added -q option; it works now only for hexa mesh, it will generate
44 * quadratic hex27 elements
45 *
46 * -t option will generate tetrahedrons instead of hexahedra. Each hexahedra is
47 * decomposed into 6 tetrahedrons.
48 *
49 * -f option will also generate all edges and faces in the model.
50 * -w will use a newer merging method locally. Merging is necessary to merge
51 * vertices on the local task, and the new method does not use a searching tree,
52 * but rather the global id set on the vertices in a consistent manner
53 *
54 * -d and -i options can be used to add some artificial tags on the model;
55 * you can have multiple -d and -i options; -i <tag_name> will set an integer
56 * tag with name tag_name on the vertices; -d < tag_name2> will generate
57 * double tags on cells (3d elements). You can have multiple tags, like
58 * -i tag1 -i tag2 -i tag3 -d tag4
59 *
60 * -x, -y, -z options will control the geometric dimensions of the final mesh, in
61 * x, y and z directions.
62 *
63 * -o <out_file> controls the name of the output file; it needs to have extension h5m,
64 * because the file is written in parallel.
65 *
66 * -k will keep the edges and faces that are generated as part of resolving shared entities
67 * (by default these edges and faces are removed); when -f option is used, the -k option is
68 * enabled too (so no faces and edges are deleted)
69 *
70 */
71
72 #include "moab/Core.hpp"
73 #include "moab/ProgOptions.hpp"
74 #include "moab/MeshGeneration.hpp"
75 #ifdef MOAB_HAVE_MPI
76 #include "moab/ParallelComm.hpp"
77 #include "MBParallelConventions.h"
78 #endif
79
80 #include <ctime>
81 #include <iostream>
82 #include <vector>
83
84 using namespace moab;
85 using namespace std;
86
87 int main( int argc, char** argv )
88 {
89
90 bool nosave = false;
91
92 #ifdef MOAB_HAVE_MPI
93 MPI_Init( &argc, &argv );
94 #endif
95 MeshGeneration::BrickOpts bopts;
96 // default options
97 bopts.A = bopts.B = bopts.C = 2;
98 bopts.M = bopts.N = bopts.K = 1;
99 bopts.blockSize = 4;
100 bopts.xsize = bopts.ysize = bopts.zsize = 1.;
101 bopts.ui = CartVect( 1., 0, 0. );
102 bopts.uj = CartVect( 0., 1., 0. );
103 bopts.uk = CartVect( 0., 0., 1. );
104 bopts.newMergeMethod = bopts.quadratic = bopts.keep_skins = bopts.tetra = false;
105 bopts.adjEnts = bopts.parmerge = false;
106 bopts.GL = 0;
107
108 ProgOptions opts;
109
110 opts.addOpt< int >( string( "blockSize,b" ), string( "Block size of mesh (default=4)" ), &bopts.blockSize );
111 opts.addOpt< int >( string( "xproc,M" ), string( "Number of processors in x dir (default=1)" ), &bopts.M );
112 opts.addOpt< int >( string( "yproc,N" ), string( "Number of processors in y dir (default=1)" ), &bopts.N );
113 opts.addOpt< int >( string( "zproc,K" ), string( "Number of processors in z dir (default=1)" ), &bopts.K );
114
115 opts.addOpt< int >( string( "xblocks,A" ), string( "Number of blocks on a task in x dir (default=2)" ), &bopts.A );
116 opts.addOpt< int >( string( "yblocks,B" ), string( "Number of blocks on a task in y dir (default=2)" ), &bopts.B );
117 opts.addOpt< int >( string( "zblocks,C" ), string( "Number of blocks on a task in x dir (default=2)" ), &bopts.C );
118
119 opts.addOpt< double >( string( "xsize,x" ), string( "Total size in x direction (default=1.)" ), &bopts.xsize );
120 opts.addOpt< double >( string( "ysize,y" ), string( "Total size in y direction (default=1.)" ), &bopts.ysize );
121 opts.addOpt< double >( string( "zsize,z" ), string( "Total size in z direction (default=1.)" ), &bopts.zsize );
122
123 opts.addOpt< void >( "newMerge,w", "use new merging method", &bopts.newMergeMethod );
124
125 opts.addOpt< void >( "quadratic,q", "use hex 27 elements", &bopts.quadratic );
126
127 opts.addOpt< void >( "keep_skins,k", "keep skins with shared entities", &bopts.keep_skins );
128
129 opts.addOpt< void >( "tetrahedrons,t", "generate tetrahedrons", &bopts.tetra );
130
131 opts.addOpt< void >( "faces_edges,f", "create all faces and edges", &bopts.adjEnts );
132
133 opts.addOpt< int >( string( "ghost_layers,g" ), string( "Number of ghost layers (default=0)" ), &bopts.GL );
134
135 vector< string > intTagNames;
136 string firstIntTag;
137 opts.addOpt< string >( "int_tag_vert,i", "add integer tag on vertices", &firstIntTag );
138
139 vector< string > doubleTagNames;
140 string firstDoubleTag;
141 opts.addOpt< string >( "double_tag_cell,d", "add double tag on cells", &firstDoubleTag );
142
143 string outFileName = "GenLargeMesh.h5m";
144 opts.addOpt< string >( "outFile,o", "Specify the output file name string (default GenLargeMesh.h5m)",
145 &outFileName );
146
147 #ifdef MOAB_HAVE_HDF5_PARALLEL
148 bool readb = false;
149 opts.addOpt< void >( "readback,r", "read back the generated mesh", &readb );
150
151 bool readAndGhost = false;
152 opts.addOpt< void >( "readAndGhost,G", "read back the generated mesh and ghost one layer", &readAndGhost );
153 #endif
154
155 opts.addOpt< void >( "parallel_merge,p", "use parallel mesh merge, not vertex ID based merge", &bopts.parmerge );
156
157 opts.addOpt< void >( "no_save,n", "do not save the file", &nosave );
158
159 opts.parseCommandLine( argc, argv );
160
161 opts.getOptAllArgs( "int_tag_vert,i", intTagNames );
162 opts.getOptAllArgs( "double_tag_cell,d", doubleTagNames );
163
164 if( bopts.adjEnts ) bopts.keep_skins = true; // Do not delete anything
165
166 Interface* mb = new( std::nothrow ) Core;
167 if( NULL == mb )
168 {
169 #ifdef MOAB_HAVE_MPI
170 MPI_Finalize();
171 #endif
172 return 1;
173 }
174
175 int rank = 0, size = 1;
176
177 #ifdef MOAB_HAVE_MPI
178 MPI_Comm_rank( MPI_COMM_WORLD, &rank );
179 MPI_Comm_size( MPI_COMM_WORLD, &size );
180 #endif
181
182 EntityHandle fileset;
183 ErrorCode rval = mb->create_meshset( MESHSET_SET, fileset );MB_CHK_ERR( rval );
184 #ifdef MOAB_HAVE_MPI
185 ParallelComm* pc = new ParallelComm( mb, MPI_COMM_WORLD );
186 MeshGeneration* mgen = new MeshGeneration( mb, pc, fileset );
187 #else
188 MeshGeneration* mgen = new MeshGeneration( mb, 0, fileset );
189 #endif
190
191 clock_t tt = clock();
192
193 rval = mgen->BrickInstance( bopts );MB_CHK_ERR( rval );
194
195 Range all3dcells;
196 rval = mb->get_entities_by_dimension( fileset, 3, all3dcells );MB_CHK_ERR( rval );
197
198 if( 0 == rank )
199 {
200 cout << "generate local mesh: " << ( clock() - tt ) / (double)CLOCKS_PER_SEC << " seconds" << endl;
201 tt = clock();
202 cout << "number of elements on rank 0: " << all3dcells.size() << endl;
203 cout << "Total number of elements " << all3dcells.size() * size << endl;
204 cout << "Element type: " << ( bopts.tetra ? "MBTET" : "MBHEX" )
205 << " order:" << ( bopts.quadratic ? "quadratic" : "linear" ) << endl;
206 }
207 Range verts;
208 rval = mb->get_entities_by_dimension( 0, 0, verts );MB_CHK_SET_ERR( rval, "Can't get all vertices" );
209
210 if( !nosave )
211 {
212 #ifdef MOAB_HAVE_HDF5_PARALLEL
213 rval = mb->write_file( outFileName.c_str(), 0, ";;PARALLEL=WRITE_PART;CPUTIME;", &fileset, 1 );MB_CHK_SET_ERR( rval, "Can't write in parallel" );
214 #else
215 // should be a vtk file, actually, maybe make sure of that
216 rval = mb->write_file( outFileName.c_str(), 0, "", &fileset, 1 );MB_CHK_SET_ERR( rval, "Can't write in serial" );
217 #endif
218 if( 0 == rank )
219 {
220 cout << "write file " << outFileName << " in " << ( clock() - tt ) / (double)CLOCKS_PER_SEC << " seconds"
221 << endl;
222 tt = clock();
223 }
224 }
225 // delete the mesh that we already have in-memory
226 size_t nLocalVerts = verts.size();
227 size_t nLocalCells = all3dcells.size();
228
229 mb->delete_mesh();
230
231 #ifdef MOAB_HAVE_HDF5_PARALLEL
232 if( !nosave && readb )
233 {
234 // now recreate a core instance and load the file we just wrote out to verify
235 Core mb2;
236 std::string read_opts( "PARALLEL=READ_PART;PARTITION=PARALLEL_PARTITION;PARALLEL_RESOLVE_"
237 "SHARED_ENTS;CPUTIME;" );
238 if( readAndGhost ) read_opts += "PARALLEL_GHOSTS=3.0.1;";
239 rval = mb2.load_file( outFileName.c_str(), 0, read_opts.c_str() );MB_CHK_SET_ERR( rval, "Can't read in parallel" );
240 if( 0 == rank )
241 {
242 cout << "read back file " << outFileName << " with options: \n"
243 << read_opts << " in " << ( clock() - tt ) / (double)CLOCKS_PER_SEC << " seconds" << endl;
244 tt = clock();
245 }
246 moab::Range nverts, ncells;
247 rval = mb2.get_entities_by_dimension( 0, 0, nverts );MB_CHK_SET_ERR( rval, "Can't get all vertices" );
248 rval = mb2.get_entities_by_dimension( 0, 3, ncells );MB_CHK_SET_ERR( rval, "Can't get all 3d cells elements" );
249
250 if( readAndGhost && size > 1 )
251 {
252 // filter out the ghost nodes and elements, for comparison with original mesh
253 // first get the parallel comm
254 ParallelComm* pcomm2 = ParallelComm::get_pcomm( &mb2, 0 );
255 if( NULL == pcomm2 ) MB_SET_ERR( MB_FAILURE, "can't get parallel comm." );
256 rval = pcomm2->filter_pstatus( nverts, PSTATUS_GHOST, PSTATUS_NOT );MB_CHK_SET_ERR( rval, "Can't filter ghost vertices" );
257 rval = pcomm2->filter_pstatus( ncells, PSTATUS_GHOST, PSTATUS_NOT );MB_CHK_SET_ERR( rval, "Can't filter ghost cells" );
258 }
259 if( nverts.size() != nLocalVerts && ncells.size() != nLocalCells )
260 {
261 MB_SET_ERR( MB_FAILURE, "Reading back the output file led to inconsistent number of entities." );
262 }
263
264 // delete the mesh that we already have in-memory
265 mb2.delete_mesh();
266 }
267 #endif
268
269 #ifdef MOAB_HAVE_MPI
270 MPI_Finalize();
271 #endif
272 return 0;
273 }
1.9.1.