1: /* This program illustrates use of parallel real FFT */
  2: static char help[]="This program illustrates the use of parallel real multi-dimensional fftw (without PETSc interface)";
  3: #include <petscmat.h>
  4: #include <fftw3.h>
  5: #include <fftw3-mpi.h>

  9: PetscInt main(PetscInt argc,char **args)
 10: {
 11:   ptrdiff_t      N0=2,N1=2,N2=2,N3=2,dim[4],N,D;
 12:   fftw_plan      bplan,fplan;
 13:   fftw_complex   *out;
 14:   double         *in1,*in2;
 15:   ptrdiff_t      alloc_local,local_n0,local_0_start;
 16:   ptrdiff_t      local_n1,local_1_start;
 17:   PetscInt       i,j,indx[100],n1;
 18:   PetscInt       size,rank,n,*in,N_factor;
 19:   PetscScalar    *data_fin,value1,one=1.0,zero=0.0;
 20:   PetscScalar    a,*x_arr,*y_arr,*z_arr,enorm;
 21:   Vec            fin,fout,fout1,x,y;
 22:   PetscRandom    rnd;

 25:   PetscInitialize(&argc,&args,(char*)0,help);
 26: #if defined(PETSC_USE_COMPLEX)
 27:   SETERRQ(PETSC_COMM_WORLD,PETSC_ERR_SUP, "This example requires real numbers. Your current scalar type is complex");
 28: #endif
 29:   MPI_Comm_size(PETSC_COMM_WORLD, &size);
 30:   MPI_Comm_rank(PETSC_COMM_WORLD, &rank);

 32:   PetscRandomCreate(PETSC_COMM_WORLD,&rnd);
 33:   D     =4;
 34:   dim[0]=N0;dim[1]=N1;dim[2]=N2;dim[3]=N3/2+1;


 37:   alloc_local = fftw_mpi_local_size_transposed(D,dim,PETSC_COMM_WORLD,&local_n0,&local_0_start,&local_n1,&local_1_start);

 39:   printf("The value alloc_local is %ld from process %d\n",alloc_local,rank);
 40:   printf("The value local_n0 is %ld from process %d\n",local_n0,rank);
 41:   printf("The value local_0_start is  %ld from process %d\n",local_0_start,rank);
 42:   printf("The value local_n1 is  %ld from process %d\n",local_n1,rank);
 43:   printf("The value local_1_start is  %ld from process %d\n",local_1_start,rank);

 45:   /* Allocate space for input and output arrays  */

 47:   in1=(double*)fftw_malloc(sizeof(double)*alloc_local*2);
 48:   in2=(double*)fftw_malloc(sizeof(double)*alloc_local*2);
 49:   out=(fftw_complex*)fftw_malloc(sizeof(fftw_complex)*alloc_local);


 52:   N=2*N0*N1*N2*(N3/2+1);N_factor=N0*N1*N2*N3;
 53:   n=2*local_n0*N1*N2*(N3/2+1);n1=local_n1*N0*2*N1*N2;

 55: /*    printf("The value N is  %d from process %d\n",N,rank); */
 56: /*    printf("The value n is  %d from process %d\n",n,rank); */
 57: /*    printf("The value n1 is  %d from process %d\n",n1,rank); */
 58:   /* Creating data vector and accompanying array with VeccreateMPIWithArray */
 59:   VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)in1,&fin);
 60:   VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)out,&fout);
 61:   VecCreateMPIWithArray(PETSC_COMM_WORLD,1,n,N,(PetscScalar*)in2,&fout1);

 63: /*    VecGetSize(fin,&size); */
 64: /*    printf("The size is %d\n",size); */

 66:   VecSet(fin,one);
 67: /*    VecAssemblyBegin(fin); */
 68: /*    VecAssemblyEnd(fin); */
 69: /*    VecView(fin,PETSC_VIEWER_STDOUT_WORLD); */


 72:   VecGetArray(fin,&x_arr);
 73:   VecGetArray(fout1,&z_arr);
 74:   VecGetArray(fout,&y_arr);

 76:   dim[3]=N3;

 78:   fplan=fftw_mpi_plan_dft_r2c(D,dim,(double*)x_arr,(fftw_complex*)y_arr,PETSC_COMM_WORLD,FFTW_ESTIMATE);
 79:   bplan=fftw_mpi_plan_dft_c2r(D,dim,(fftw_complex*)y_arr,(double*)z_arr,PETSC_COMM_WORLD,FFTW_ESTIMATE);

 81:   fftw_execute(fplan);
 82:   fftw_execute(bplan);

 84:   VecRestoreArray(fin,&x_arr);
 85:   VecRestoreArray(fout1,&z_arr);
 86:   VecRestoreArray(fout,&y_arr);

 88: /*    a = 1.0/(PetscReal)N_factor; */
 89: /*    VecScale(fout1,a); */

 91:   VecAssemblyBegin(fout1);
 92:   VecAssemblyEnd(fout1);

 94:   VecView(fout1,PETSC_VIEWER_STDOUT_WORLD);

 96:   fftw_destroy_plan(fplan);
 97:   fftw_destroy_plan(bplan);
 98:   fftw_free(in1); VecDestroy(&fin);
 99:   fftw_free(out); VecDestroy(&fout);
100:   fftw_free(in2); VecDestroy(&fout1);

102:   PetscFinalize();
103:   return 0;
104: }