static void init_parameters(
int argc, char **argv,
int *n, int *iter,
int *inplace, int* patience
)
{
pfft_get_args(argc, argv, "-pfft_n", 3, PFFT_INT, n);
pfft_get_args(argc, argv, "-pfft_iter", 1, PFFT_INT, iter);
pfft_get_args(argc, argv, "-pfft_ip", 1, PFFT_INT, inplace);
pfft_get_args(argc, argv, "-pfft_pat", 1, PFFT_INT, patience);
}
static void init_parameters(
int argc, char **argv,
ptrdiff_t *n, int *np, ptrdiff_t *gc_below, ptrdiff_t *gc_above,
int* verbose
)
{
pfft_get_args(argc, argv, "-pfft_n", 3, PFFT_PTRDIFF_T, n);
pfft_get_args(argc, argv, "-pfft_np", 3, PFFT_INT, np);
pfft_get_args(argc, argv, "-pfft_gcb", 3, PFFT_PTRDIFF_T, gc_below);
pfft_get_args(argc, argv, "-pfft_gca", 3, PFFT_PTRDIFF_T, gc_above);
pfft_get_args(argc, argv, "-pfft_verb", 1, PFFT_INT, verbose);
}
static void init_parameters(
int argc, char **argv,
ptrdiff_t *n, int *np, int *loops, int *inplace,
unsigned *opt_flag, unsigned *tune_flag, unsigned *destroy_input_flag
)
{
int opt=0, tune=0, destroy_input=0;
pfft_get_args(argc, argv, "-pfft_n", 3, PFFT_PTRDIFF_T, n);
pfft_get_args(argc, argv, "-pfft_np", 2, PFFT_INT, np);
pfft_get_args(argc, argv, "-pfft_loops", 1, PFFT_INT, loops);
pfft_get_args(argc, argv, "-pfft_ip", 1, PFFT_INT, inplace);
pfft_get_args(argc, argv, "-pfft_opt", 1, PFFT_INT, &opt);
pfft_get_args(argc, argv, "-pfft_tune", 1, PFFT_INT, &tune);
pfft_get_args(argc, argv, "-pfft_di", 1, PFFT_INT, &destroy_input);
switch(opt){
case 1: *opt_flag = PFFT_MEASURE; break;
case 2: *opt_flag = PFFT_PATIENT; break;
case 3: *opt_flag = PFFT_EXHAUSTIVE; break;
}
if(destroy_input)
*destroy_input_flag = PFFT_DESTROY_INPUT;
if(tune)
*tune_flag = PFFT_TUNE;
}
static void init_parameters(
int argc, char **argv,
ptrdiff_t *N, ptrdiff_t *n, ptrdiff_t *local_M,
int *m, int *window, int *intpol, int *interlacing,
double *x_max, int *np
)
{
pfft_get_args(argc, argv, "-pnfft_local_M", 1, PFFT_PTRDIFF_T, local_M);
pfft_get_args(argc, argv, "-pnfft_N", 3, PFFT_PTRDIFF_T, N);
pfft_get_args(argc, argv, "-pnfft_n", 3, PFFT_PTRDIFF_T, n);
pfft_get_args(argc, argv, "-pnfft_np", 3, PFFT_INT, np);
pfft_get_args(argc, argv, "-pnfft_m", 1, PFFT_INT, m);
pfft_get_args(argc, argv, "-pnfft_window", 1, PFFT_INT, window);
pfft_get_args(argc, argv, "-pnfft_intpol", 1, PFFT_INT, intpol);
pfft_get_args(argc, argv, "-pnfft_interlacing", 1, PFFT_INT, interlacing);
pfft_get_args(argc, argv, "-pnfft_x_max", 3, PFFT_DOUBLE, x_max);
}
int main(int argc, char **argv)
{
int nthreads=1; /*number of threads to initialize openmp with*/
int runs=1; /*number of runs for testing*/
int np[2];
ptrdiff_t n[3];
ptrdiff_t alloc_local;
ptrdiff_t local_ni[3], local_i_start[3];
ptrdiff_t local_no[3], local_o_start[3];
double err;
pfft_complex *in, *out;
pfft_plan plan_forw=NULL, plan_back=NULL;
MPI_Comm comm_cart_2d;
/* Init OpenMP */
pfft_get_args(argc,argv,"-pfft_omp_threads",1,PFFT_INT,&nthreads);
pfft_get_args(argc,argv,"-pfft_runs",1,PFFT_INT,&runs);
pfft_plan_with_nthreads(nthreads);
/* Set size of FFT and process mesh */
n[0] = NNN;n[1] =NNN; n[2] =NNN;
np[0] = 1; np[1] = 1;
/* Initialize MPI and PFFT */
MPI_Init(&argc, &argv);
pfft_init();
pfft_plan_with_nthreads(nthreads);
pfft_printf(MPI_COMM_WORLD, "# %4d threads will be used for openmp (default is 1)\n", nthreads);
/* Create two-dimensional process grid of size np[0] x np[1], if possible */
if( pfft_create_procmesh_2d(MPI_COMM_WORLD, np[0], np[1], &comm_cart_2d) ){
pfft_fprintf(MPI_COMM_WORLD, stderr, "Error: This test file only works with %d processes.\n", np[0]*np[1]);
MPI_Finalize();
return 1;
}
/* Get parameters of data distribution */
alloc_local = pfft_local_size_dft_3d(n, comm_cart_2d, PFFT_TRANSPOSED_NONE,
local_ni, local_i_start, local_no, local_o_start);
/* Allocate memory */
in = pfft_alloc_complex(alloc_local);
out = pfft_alloc_complex(alloc_local);
/* Plan parallel forward FFT */
plan_forw = pfft_plan_dft_3d(
n, in, out, comm_cart_2d, PFFT_FORWARD, PFFT_TRANSPOSED_OUT| PFFT_MEASURE| PFFT_DESTROY_INPUT| PFFT_TUNE| PFFT_SHIFTED_IN);
/* Plan parallel backward FFT */
plan_back = pfft_plan_dft_3d(
n, out, in, comm_cart_2d, PFFT_BACKWARD, PFFT_TRANSPOSED_IN| PFFT_MEASURE| PFFT_DESTROY_INPUT| PFFT_TUNE| PFFT_SHIFTED_OUT);
/* Initialize input with random numbers */
pfft_init_input_complex_3d(n, local_ni, local_i_start,
in);
for(int i=0; i<runs; i++)
{
/* execute parallel forward FFT */
pfft_execute(plan_forw);
/* clear the old input */
/* pfft_clear_input_complex_3d(n, local_ni, local_i_start,
in);
*/
/* execute parallel backward FFT */
pfft_execute(plan_back);
/* Scale data */
ptrdiff_t l;
for(l=0; l < local_ni[0] * local_ni[1] * local_ni[2]; l++)
in[l] /= (n[0]*n[1]*n[2]);
}
pfft_print_average_timer_adv(plan_forw, MPI_COMM_WORLD);
pfft_print_average_timer_adv(plan_back, MPI_COMM_WORLD);
/* Print error of back transformed data */
err = pfft_check_output_complex_3d(n, local_ni, local_i_start, in, comm_cart_2d);
pfft_printf(comm_cart_2d, "Error after %d forward and backward trafos of size n=(%td, %td, %td):\n", runs, n[0], n[1], n[2]);
pfft_printf(comm_cart_2d, "maxerror = %6.2e;\n", err);
/* free mem and finalize */
pfft_destroy_plan(plan_forw);
pfft_destroy_plan(plan_back);
MPI_Comm_free(&comm_cart_2d);
pfft_free(in); pfft_free(out);
MPI_Finalize();
return 0;
}
int main(int argc, char **argv)
{
int rows=1,columns=1;
int nthreads=1;
int inplace=0;
int measure=0;
int runs=1;
pfft_get_args(argc,argv,"-pfft_test_runs",1,PFFT_INT,&runs);
pfft_get_args(argc,argv,"-pfft_omp_threads",1,PFFT_INT,&nthreads);
pfft_get_args(argc,argv,"-pfft_omp_rows",1,PFFT_INT,&rows);
pfft_get_args(argc,argv,"-pfft_omp_columns",1,PFFT_INT,&columns);
pfft_get_args(argc,argv,"-pfft_omp_inplace",1,PFFT_SWITCH,&inplace);
pfft_get_args(argc,argv,"-pfft_omp_measure",1,PFFT_SWITCH,&measure);
if(inplace!=1) inplace=0;
/* printf("Enter a number:");
scanf("%d",&nthreads);
printf("\n");*/
fftw_complex*m1 = (fftw_complex *) fftw_malloc(sizeof(fftw_complex)*rows*columns);
fftw_complex*m2 = (fftw_complex *) fftw_malloc(sizeof(fftw_complex)*rows*columns);
fftw_iodim howmany_dims[2];
howmany_dims[0].n=rows;
howmany_dims[0].is=columns;
howmany_dims[0].os=1;
howmany_dims[1].n=columns;
howmany_dims[1].is=1;
howmany_dims[1].os=rows;
const int howmany_rank = 2;
// init OMP threads for fftw
fftw_init_threads();
fftw_plan_with_nthreads(nthreads);
fftw_plan plan_transpose_outplace,plan_transpose_inplace,plansimple2d;
int fftw_flags=FFTW_ESTIMATE;
if(measure==1) fftw_flags=FFTW_MEASURE;
plan_transpose_outplace = fftw_plan_guru_dft(/*rank*/ 0,NULL,howmany_rank, howmany_dims,m1,m2,FFTW_FORWARD,fftw_flags);
plan_transpose_inplace = fftw_plan_guru_dft(/*rank*/ 0,NULL,howmany_rank, howmany_dims,m1,m1,FFTW_FORWARD,fftw_flags);
/* plansimple2d=fftw_plan_2d()*/
/*plansimple2d=fftw_plan_dft_2d(rows,columns,m1,m1,FFTW_FORWARD,FFTW_ESTIMATE);*/
/*printf("# rows=%d, columns=%d, threads=%d,inplace=%d, measure=%d\n",rows,columns,nthreads,inplace,measure);
initialize_matrix(m1,rows,columns);
printf("# start calculation...\n");*/
/* printf("# check");*/
/* printmatrix(m1,rows,columns);*/
struct timeval t1,t2;
gettimeofday(&t1,NULL);
int i;
for(i=0;i<runs;i++)
{
if(inplace==1) fftw_execute(plan_transpose_inplace);
if(inplace==0) fftw_execute(plan_transpose_outplace);
}
/*if(selectplan==3) fftw_execute(plansimple2d);*/
gettimeofday(&t2,NULL);
float seconds=t2.tv_sec-t1.tv_sec+(t2.tv_usec-t1.tv_usec)/1000000.0;
printf("# output: rows,cols,inplace,,measure,threads,runs,total time,average time\n");
printf("%d %d %d %d %d %d %.6f %6f\n",rows,columns,inplace,measure,nthreads,runs,seconds,seconds/runs);
/* printmatrix(m1,rows,columns);*/
free(m1);
free(m2);
return 0;
}
