ExaFMM 1
Fast-multipole Method for exascale systems
|
00001 /* 00002 Copyright (C) 2011 by Rio Yokota, Simon Layton, Lorena Barba 00003 00004 Permission is hereby granted, free of charge, to any person obtaining a copy 00005 of this software and associated documentation files (the "Software"), to deal 00006 in the Software without restriction, including without limitation the rights 00007 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 00008 copies of the Software, and to permit persons to whom the Software is 00009 furnished to do so, subject to the following conditions: 00010 00011 The above copyright notice and this permission notice shall be included in 00012 all copies or substantial portions of the Software. 00013 00014 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 00015 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 00016 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 00017 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 00018 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 00019 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 00020 THE SOFTWARE. 00021 */ 00022 #include "parallelfmm.h" 00023 #ifdef VTK 00024 #include "vtk.h" 00025 #endif 00026 00027 int main() { 00028 const int numBodies = 10000; 00029 const int numTarget = 100; 00030 IMAGES = 0; 00031 THETA = 1 / sqrtf(4); 00032 Bodies bodies(numBodies); 00033 Bodies jbodies; 00034 Cells cells, jcells; 00035 ParallelFMM<Laplace> FMM; 00036 FMM.initialize(); 00037 if( MPIRANK == 0 ) FMM.printNow = true; 00038 00039 FMM.startTimer("Set bodies "); 00040 FMM.random(bodies,MPIRANK+1); 00041 FMM.stopTimer("Set bodies ",FMM.printNow); 00042 00043 FMM.startTimer("Set domain "); 00044 FMM.setGlobDomain(bodies); 00045 FMM.stopTimer("Set domain ",FMM.printNow); 00046 00047 FMM.octsection(bodies); 00048 00049 #ifdef TOPDOWN 00050 FMM.topdown(bodies,cells); 00051 #else 00052 FMM.bottomup(bodies,cells); 00053 #endif 00054 00055 FMM.commBodies(cells); 00056 00057 #ifndef VTK 00058 if( IMAGES != 0 ) { 00059 FMM.startTimer("Set periodic "); 00060 jbodies = FMM.periodicBodies(bodies); 00061 FMM.stopTimer("Set periodic ",FMM.printNow); 00062 FMM.eraseTimer("Set periodic "); 00063 } else { 00064 jbodies = bodies; 00065 } 00066 FMM.startTimer("Direct sum "); 00067 Bodies bodies2 = bodies; 00068 bodies2.resize(numTarget); 00069 FMM.initTarget(bodies2); 00070 for( int i=0; i!=MPISIZE; ++i ) { 00071 FMM.shiftBodies(jbodies); 00072 FMM.evalP2P(bodies2,jbodies); 00073 if(FMM.printNow) std::cout << "Direct loop : " << i+1 << "/" << MPISIZE << std::endl; 00074 } 00075 FMM.stopTimer("Direct sum ",FMM.printNow); 00076 #endif 00077 00078 FMM.resetTimer(); 00079 FMM.initTarget(bodies); 00080 FMM.evalP2M(cells); 00081 FMM.evalM2M(cells,cells); 00082 jcells = cells; 00083 if( MPISIZE != 1 ) { 00084 #pragma omp parallel sections num_threads(2) 00085 { 00086 #pragma omp section 00087 { 00088 FMM.downward(cells,jcells,false); 00089 } 00090 #pragma omp section 00091 { 00092 FMM.updateBodies(); 00093 } 00094 } 00095 jbodies = bodies; 00096 jcells = cells; 00097 FMM.commCells(jbodies,jcells); 00098 FMM.eraseLocalTree(jcells); 00099 } 00100 FMM.downward(cells,jcells); 00101 if(FMM.printNow) FMM.writeTime(); 00102 if(FMM.printNow) FMM.writeTime(); 00103 00104 #ifndef VTK 00105 real diff1 = 0, norm1 = 0, diff2 = 0, norm2 = 0, diff3 = 0, norm3 = 0, diff4 = 0, norm4 = 0; 00106 bodies.resize(numTarget); 00107 FMM.evalError(bodies,bodies2,diff1,norm1,diff2,norm2); 00108 MPI_Datatype MPI_TYPE = FMM.getType(diff1); 00109 MPI_Reduce(&diff1,&diff3,1,MPI_TYPE,MPI_SUM,0,MPI_COMM_WORLD); 00110 MPI_Reduce(&norm1,&norm3,1,MPI_TYPE,MPI_SUM,0,MPI_COMM_WORLD); 00111 MPI_Reduce(&diff2,&diff4,1,MPI_TYPE,MPI_SUM,0,MPI_COMM_WORLD); 00112 MPI_Reduce(&norm2,&norm4,1,MPI_TYPE,MPI_SUM,0,MPI_COMM_WORLD); 00113 if(FMM.printNow) FMM.printError(diff3,norm3,diff4,norm4); 00114 00115 #else 00116 for( B_iter B=jbodies.begin(); B!=jbodies.end(); ++B ) B->ICELL = 0; 00117 for( C_iter C=jcells.begin(); C!=jcells.end(); ++C ) { 00118 Body body; 00119 body.ICELL = 1; 00120 body.X = C->X; 00121 body.SRC = 0; 00122 jbodies.push_back(body); 00123 } 00124 00125 int Ncell = 0; 00126 vtkPlot vtk; 00127 if( MPIRANK == 0 ) { 00128 vtk.setDomain(FMM.getR0(),FMM.getX0()); 00129 vtk.setGroupOfPoints(jbodies,Ncell); 00130 } 00131 for( int i=1; i!=MPISIZE; ++i ) { 00132 FMM.shiftBodies(jbodies); 00133 if( MPIRANK == 0 ) { 00134 vtk.setGroupOfPoints(jbodies,Ncell); 00135 } 00136 } 00137 if( MPIRANK == 0 ) { 00138 vtk.plot(Ncell); 00139 } 00140 #endif 00141 FMM.finalize(); 00142 }