Actual source code: ex144.c
petsc-3.8.4 2018-03-24
1: /* This program illustrates use of parallel real FFT */
2: static char help[]="This program illustrates the use of parallel real 2D fft using fftw without PETSc interface";
3: #include <petscmat.h>
4: #include <fftw3.h>
5: #include <fftw3-mpi.h>
7: int main(int argc,char **args)
8: {
9: const ptrdiff_t N0=2056,N1=2056;
10: fftw_plan bplan,fplan;
11: fftw_complex *out;
12: double *in1,*in2;
13: ptrdiff_t alloc_local,local_n0,local_0_start;
14: ptrdiff_t local_n1,local_1_start;
15: PetscInt i,j;
16: PetscMPIInt size,rank;
17: int n,N,N_factor,NM;
18: PetscScalar one=2.0,zero=0.5;
19: PetscScalar two=4.0,three=8.0,four=16.0;
20: PetscScalar a,*x_arr,*y_arr,*z_arr,enorm;
21: Vec fin,fout,fout1;
22: Vec ini,final;
23: PetscRandom rnd;
24: PetscErrorCode ierr;
25: PetscInt *indx3,tempindx,low,*indx4,tempindx1;
27: PetscInitialize(&argc,&args,(char*)0,help);if (ierr) return ierr;
28: MPI_Comm_size(PETSC_COMM_WORLD, &size);
29: MPI_Comm_rank(PETSC_COMM_WORLD, &rank);
31: PetscRandomCreate(PETSC_COMM_WORLD,&rnd);
33: alloc_local = fftw_mpi_local_size_2d_transposed(N0,N1/2+1,PETSC_COMM_WORLD,&local_n0,&local_0_start,&local_n1,&local_1_start);
34: #if defined(DEBUGGING)
35: printf("The value alloc_local is %ld from process %d\n",alloc_local,rank);
36: printf("The value local_n0 is %ld from process %d\n",local_n0,rank);
37: printf("The value local_0_start is %ld from process %d\n",local_0_start,rank);
38: /* printf("The value local_n1 is %ld from process %d\n",local_n1,rank); */
39: /* printf("The value local_1_start is %ld from process %d\n",local_1_start,rank); */
40: /* printf("The value local_n0 is %ld from process %d\n",local_n0,rank); */
41: #endif
43: /* Allocate space for input and output arrays */
44: in1=(double*)fftw_malloc(sizeof(double)*alloc_local*2);
45: in2=(double*)fftw_malloc(sizeof(double)*alloc_local*2);
46: out=(fftw_complex*)fftw_malloc(sizeof(fftw_complex)*alloc_local);
48: N = 2*N0*(N1/2+1);
49: N_factor = N0*N1;
50: n = 2*local_n0*(N1/2+1);
52: /* printf("The value N is %d from process %d\n",N,rank); */
53: /* printf("The value n is %d from process %d\n",n,rank); */
54: /* printf("The value n1 is %d from process %d\n",n1,rank);*/
55: /* Creating data vector and accompanying array with VeccreateMPIWithArray */
56: VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)in1,&fin);
57: VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)out,&fout);
58: VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)in2,&fout1);
60: /* Set the vector with random data */
61: VecSet(fin,zero);
62: /* for (i=0;i<N0*N1;i++) */
63: /* { */
64: /* VecSetValues(fin,1,&i,&one,INSERT_VALUES); */
65: /* } */
67: /* VecSet(fin,one); */
68: i =0;
69: VecSetValues(fin,1,&i,&one,INSERT_VALUES);
70: i =1;
71: VecSetValues(fin,1,&i,&two,INSERT_VALUES);
72: i =4;
73: VecSetValues(fin,1,&i,&three,INSERT_VALUES);
74: i =5;
75: VecSetValues(fin,1,&i,&four,INSERT_VALUES);
76: VecAssemblyBegin(fin);
77: VecAssemblyEnd(fin);
79: VecSet(fout,zero);
80: VecSet(fout1,zero);
82: /* Get the meaningful portion of array */
83: VecGetArray(fin,&x_arr);
84: VecGetArray(fout1,&z_arr);
85: VecGetArray(fout,&y_arr);
87: fplan=fftw_mpi_plan_dft_r2c_2d(N0,N1,(double*)x_arr,(fftw_complex*)y_arr,PETSC_COMM_WORLD,FFTW_ESTIMATE);
88: bplan=fftw_mpi_plan_dft_c2r_2d(N0,N1,(fftw_complex*)y_arr,(double*)z_arr,PETSC_COMM_WORLD,FFTW_ESTIMATE);
90: fftw_execute(fplan);
91: fftw_execute(bplan);
93: VecRestoreArray(fin,&x_arr);
94: VecRestoreArray(fout1,&z_arr);
95: VecRestoreArray(fout,&y_arr);
97: /* VecView(fin,PETSC_VIEWER_STDOUT_WORLD); */
98: VecCreate(PETSC_COMM_WORLD,&ini);
99: VecCreate(PETSC_COMM_WORLD,&final);
100: VecSetSizes(ini,local_n0*N1,N0*N1);
101: VecSetSizes(final,local_n0*N1,N0*N1);
102: VecSetFromOptions(ini);
103: VecSetFromOptions(final);
105: if (N1%2==0) {
106: NM = N1+2;
107: } else {
108: NM = N1+1;
109: }
110: /*printf("The Value of NM is %d",NM); */
111: VecGetOwnershipRange(fin,&low,NULL);
112: /*printf("The local index is %d from %d\n",low,rank); */
113: PetscMalloc1(local_n0*N1,&indx3);
114: PetscMalloc1(local_n0*N1,&indx4);
115: for (i=0;i<local_n0;i++) {
116: for (j=0;j<N1;j++) {
117: tempindx = i*N1 + j;
118: tempindx1 = i*NM + j;
120: indx3[tempindx]=local_0_start*N1+tempindx;
121: indx4[tempindx]=low+tempindx1;
122: /* printf("index3 %d from proc %d is \n",indx3[tempindx],rank); */
123: /* printf("index4 %d from proc %d is \n",indx4[tempindx],rank); */
124: }
125: }
127: PetscMalloc2(local_n0*N1,&x_arr,local_n0*N1,&y_arr); /* arr must be allocated for VecGetValues() */
128: VecGetValues(fin,local_n0*N1,indx4,(PetscScalar*)x_arr);
129: VecSetValues(ini,local_n0*N1,indx3,x_arr,INSERT_VALUES);
131: VecAssemblyBegin(ini);
132: VecAssemblyEnd(ini);
134: VecGetValues(fout1,local_n0*N1,indx4,y_arr);
135: VecSetValues(final,local_n0*N1,indx3,y_arr,INSERT_VALUES);
136: VecAssemblyBegin(final);
137: VecAssemblyEnd(final);
138: PetscFree2(x_arr,y_arr);
140: /*
141: VecScatter vecscat;
142: IS indx1,indx2;
143: for (i=0;i<N0;i++) {
144: indx = i*NM;
145: ISCreateStride(PETSC_COMM_WORLD,N1,indx,1,&indx1);
146: indx = i*N1;
147: ISCreateStride(PETSC_COMM_WORLD,N1,indx,1,&indx2);
148: VecScatterCreate(fin,indx1,ini,indx2,&vecscat);
149: VecScatterBegin(vecscat,fin,ini,INSERT_VALUES,SCATTER_FORWARD);
150: VecScatterEnd(vecscat,fin,ini,INSERT_VALUES,SCATTER_FORWARD);
151: VecScatterBegin(vecscat,fout1,final,INSERT_VALUES,SCATTER_FORWARD);
152: VecScatterEnd(vecscat,fout1,final,INSERT_VALUES,SCATTER_FORWARD);
153: }
154: */
156: a = 1.0/(PetscReal)N_factor;
157: VecScale(fout1,a);
158: VecScale(final,a);
161: /* VecView(ini,PETSC_VIEWER_STDOUT_WORLD); */
162: /* VecView(final,PETSC_VIEWER_STDOUT_WORLD); */
163: VecAXPY(final,-1.0,ini);
165: VecNorm(final,NORM_1,&enorm);
166: if (enorm > 1.e-10) {
167: PetscPrintf(PETSC_COMM_WORLD," Error norm of |x - z| = %e\n",enorm);
168: }
170: /* Execute fftw with function fftw_execute and destory it after execution */
171: fftw_destroy_plan(fplan);
172: fftw_destroy_plan(bplan);
173: fftw_free(in1); VecDestroy(&fin);
174: fftw_free(out); VecDestroy(&fout);
175: fftw_free(in2); VecDestroy(&fout1);
177: VecDestroy(&ini);
178: VecDestroy(&final);
180: PetscRandomDestroy(&rnd);
181: PetscFree(indx3);
182: PetscFree(indx4);
183: PetscFinalize();
184: return ierr;
185: }