int main(int argc, char **argv)
{
double *A, *B, *C;
int N;
int i, j, k;
double elapsed;
// Input data
N = atoi(argv[1]);
A = (double *) malloc( N * N * sizeof(double) );
B = (double *) malloc( N * N * sizeof(double) );
C = (double *) malloc( N * N * sizeof(double) );
if((A == NULL) || (B == NULL) || (C == NULL) ){
printf("Running out of memory!\n"); exit(EXIT_FAILURE);
}
//Fill matrixes. Generate Identity like matrix for A and B , So C should result in an matrix with a single major diagonal
for(i=0; i < N; i++ ){
for(j=0; j < N; j++){
A[i+N*j] = (i==j)?i:0.0;
B[i+N*j] = (i==j)?1.0:0.0;
C[i+N*j] = 0.0;
}
}
int rows = N, columns = N;
int stride = N;
tick();
//cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, rows, columns, columns, 1.0, A, stride, B, stride, 1.0, C, stride);
for(i=0; i < N; i++){
for(j=0; j<N; j++){
C[j + i*N] = 0.0;
for(k=0; k<N; k++){
C[j + i*N] += A[k + i*N] * B[j + k*N];
}
}
}
elapsed = tack();
printf("%f sec\n", elapsed);
if( N < 30 )
{
printf("C ... \n");
for (i=0; i<N; i++) {
for (j=0; j<N; j++) {
printf("%3.1f ", C[i+N*j]);
}
printf("\n");
}
}
free(A);
free(B);
free(C);
exit(EXIT_SUCCESS);
}
int main(int argc, char *argv[])
{
long i;
graph G;
double runtime;
inputCheck(argc, argv);
if(N == 1){
generateTestGraph(&G);
} else {
generateGraph(N, randInit, &G, 0);
}
tick();
dijkstra(&G, 0, 0);
runtime = tack();
//
char *b;
b = malloc(G.N * 5);
if(b == NULL) {perror("malloc"); exit(EXIT_FAILURE); }
sprintf(b,"\nLowest distances!\nD=[");
for(i = 0; i<G.N; i++){
sprintf(&b[strlen(b)], "%d,", G.D[i]);
}
printf("%s]\n", b);
printf("Was working for [%f] sec.\n",runtime);
return EXIT_SUCCESS;
}
void test_get_diff(){
struct timestamp *ts;
time_t diff_sec;
long diff_nsec;
ts = init_timestamp();
/*test < 1sec*/
printf("Test in 500msec\n");
tick(ts);
usleep(500000); /*500000usec = 500msec*/
tack(ts);
diff_sec = get_diff_sec(ts);
diff_nsec = get_diff_nanosec(ts);
printf("start=%d:%ld end=%d:%ld\n", (int)(ts->start.tv_sec), ts->start.tv_nsec, (int)(ts->end.tv_sec), ts->end.tv_nsec);
printf("diff: %dsec + %ld nanosec\n", (int)diff_sec, diff_nsec);
/*test 1-2sec*/
printf("Test in 1500msec\n");
tick(ts);
usleep(1000000 + 500000); /*1500000usec = 1500msec*/
tack(ts);
diff_sec = get_diff_sec(ts);
diff_nsec = get_diff_nanosec(ts);
printf("start=%d:%ld end=%d:%ld\n", (int)(ts->start.tv_sec), ts->start.tv_nsec, (int)(ts->end.tv_sec), ts->end.tv_nsec);
printf("diff: %dsec + %ld nanosec\n", (int)diff_sec, diff_nsec);
/*test > 2sec*/
printf("Test in 3800msec\n");
tick(ts);
usleep(3000000 + 800000); /*3800000usec = 3800msec*/
tack(ts);
diff_sec = get_diff_sec(ts);
diff_nsec = get_diff_nanosec(ts);
printf("start=%d:%ld end=%d:%ld\n", (int)(ts->start.tv_sec), ts->start.tv_nsec, (int)(ts->end.tv_sec), ts->end.tv_nsec);
printf("diff: %dsec + %ld nanosec\n", (int)diff_sec, diff_nsec);
dispose_timestamp(ts);
}
int main(int argc, char** argv)
{
point* points;
distance solution;
double elapsedTime;
points = NULL;
inputCheck(argc, argv);
printf("Generating [%d] points\n", np);
if( generatePoints(np, &points) != EXIT_SUCCESS ){
printf("Generating Points failed!\n");
exit(EXIT_FAILURE);
}
tick();
mpiInit(argc, argv);
if( prepareMPIComm() )
{
if(mpi_id == 0)
{
printf("Starting search ...");
}
multiSearch(np, points, &solution);
elapsedTime = tack();
//printf("Found Solution a[%f,%f] , b[%f,%f] distance [%0.10f]\n", solution.a.x, solution.a.y, solution.b.x, solution.b.y, solution.d);
if(mpi_id == 0){
printf("Completed Search and found closest points at [%g, %g] , [%g, %g] with a distance of [%g]\n", mpi_id\
, solution.a.x ,solution.a.y, solution.b.x, solution.b.y
, solution.d);
printf("Operation took %f seconds \n", elapsedTime);
free(points);
}
} else {
}
mpiFinish();
exit(EXIT_SUCCESS);
}
void testScheduler(int nThreads, graph* G, char debug)
{
double runtime;
//Set max nThreads
omp_set_num_threads(nThreads);
if(mpi_id == 0) printf("Scheduler (Static, %d)", G->N/100 );
resetGraph(G);
omp_set_schedule(omp_sched_static, G->N/100);
if(mpi_id == 0) tick();
dijkstra(G, 0, debug);
if(mpi_id == 0){
runtime = tack();
printf("working for [%f] sec.\n",runtime);
}
if(mpi_id == 0) printf("Scheduler (dynamic, %d)", G->N/100 );
resetGraph(G);
omp_set_schedule(omp_sched_dynamic, G->N/100);
if(mpi_id == 0) tick();
dijkstra(G, 0, debug);
if(mpi_id == 0){
runtime = tack();
printf("working for [%f] sec.\n",runtime);
}
if(mpi_id == 0) printf("Scheduler (guided, %d)", G->N/100 );
resetGraph(G);
omp_set_schedule(omp_sched_guided, G->N/100);
if(mpi_id == 0) tick();
dijkstra(G, 0, debug);
if(mpi_id == 0){
runtime = tack();
printf("working for [%f] sec.\n",runtime);
}
}
static void
loop(
int *n,
char *b
)
{
int i;
if (n[0] == 0) {
*b = '\0';
puts(buf);
return;
}
for (i = 0; i < n[0]; i++)
loop(n+1, tack(b, i));
}
void test_tick_tack(void){
struct timestamp *ts;
time_t diff;
ts = init_timestamp();
assert(ts != NULL);
tick(ts);
usleep(1000000 + 500000);
tack(ts);
diff = ts->end.tv_sec - ts->start.tv_sec;
printf("start=%d:%ld end=%d:%ld\n", (int)(ts->start.tv_sec), ts->start.tv_nsec, (int)(ts->end.tv_sec), ts->end.tv_nsec);
dispose_timestamp(ts);
}
main (int argc, char *argv[])
{
struct em_file inputdata1;
struct em_file inputdata2;
struct em_file inputdata3;
struct em_file inputdata4;
struct em_file outputdata;
fftw_real *Vol_tmpl_sort, *Volume, *e3, *PointCorr, *sqconv;
fftw_complex *C3, *PointVolume, *PointSq;
rfftwnd_plan p3, pi3, r3, ri3;
fftw_real scale;
struct tm *zeit;
struct tm start;
char name[200];
int Rx_max, Ry_max, Rz_max;
int Rx_min, Ry_min, Rz_min;
int Vx_min, Vy_min, Vz_min;
int Vx_max, Vy_max, Vz_max;
float Phi, Psi, Theta, winkel_lauf;
float *Rot_tmpl, *Vol_tmpl;
int i, j, k, tmpx, tmpy, tmpz,lauf_pe, ksub;
int ijk;
int lauf, n;
float max, eps;
time_t lt;
float Ctmp, Ctmpim, Dtmp, Dtmpim;
int dim_fft;
int sub[3],range[3],range_sub[3],subc[3],offset[3],dimarray[3];
int FullVolume_dims[3];
int nr[3];
int area[3];
/* MPI Variablen */
int winkel_max, winkel_min;
int winkel_max_pe, winkel_min_pe;
int winkel_step_pe;
int Phi_max, Psi_max, Theta_max;
int Phi_min, Psi_min, Theta_min;
int Phi_step, Psi_step, Theta_step;
int Theta_winkel_start, Psi_winkel_start, Phi_winkel_start;
int Theta_winkel_nr, Psi_winkel_nr, Phi_winkel_nr;
int Theta_winkel_end, Psi_winkel_end, Phi_winkel_end;
int Theta_steps, Psi_steps, Phi_steps;
float Theta_winkel_rest_nr, Psi_winkel_rest_nr, Phi_winkel_rest_nr;
int in_max;
float rms_wedge, tempccf;
float *Ergebnis, *conv;
float cycles;
int cycle;
/* MPI Variablen Ende*/
if (argc < 15)
{
printf ("\n\n");
printf (" 'OSCAR' is an Optimized SCanning AlgoRithm for \n");
printf (" local correlation.\n");
printf (" All files in EM-V-int4 format !!!\n\n");
printf (" Input: Volume to be searched, Template mask for local \n ");
printf (" correlation, pointspread function and angular search \n");
printf (" range. \n");
printf (" Output: locally normalized X-Correlation Function Out.ccf.norm, \n");
printf (" non-normalized X-Correlation Function Out.ccf, and Out.ang \n");
printf (" with the corresponding angles.\n\n");
printf (" usage: oscar Volume Template Out ...\n");
printf (" ... Phi_min Phi_max Phi_step Psi_min Psi_max Psi_step The_min The_max The_step\n");
printf (" ... Poinspread-function mask-file dim_of_fft\n\n");
printf (" with Message Passing Interface (MPI)\n");
printf (" the total number of angles must be modulo\n");
printf (" of used processors!\n\n");
printf (" Linux: 1.'lamboot' to start MPI\n");
printf (" 2.'mpirun -np 2 oscar Volume Templ Out 30 180 30 30 180 30 30 180 30 Poinspread-function mask-file 256'\n\n");
printf (" In this version asymmetric masks can be used ! \n");
printf (" last revision , 11.11.03, Friedrich Foerster");
printf (" \n\n");
exit (1);
}
MPI_Init (&argc, &argv);
MPI_Comm_size (MPI_COMM_WORLD, &mysize);
MPI_Comm_rank (MPI_COMM_WORLD, &myrank);
/* Dimensionen auslesen */
// Dimension of fft
dim_fft = atoi (argv[15]);
nr[0]=1;
nr[1]=1;
nr[2]=1;
area[0]=dim_fft;
area[1]=dim_fft;
area[2]=dim_fft;
read_em_header(argv[1], &inputdata1); /* Searchvolume */
read_em (argv[2], &inputdata2); /* Template */
FullVolume_dims[0]=inputdata1.dims[0];
FullVolume_dims[1]=inputdata1.dims[1];
FullVolume_dims[2]=inputdata1.dims[2];
Rx_min = 1;
Ry_min = 1;
Rz_min = 1;
Rx_max = (inputdata2.dims[0]);
Ry_max = (inputdata2.dims[1]);
Rz_max = (inputdata2.dims[2]);
Vx_min = 1;
Vy_min = 1;
Vz_min = 1;
Vx_max = dim_fft;
Vy_max = dim_fft;
Vz_max = dim_fft;
p3 = rfftw3d_create_plan (Vx_max, Vy_max, Vz_max, FFTW_REAL_TO_COMPLEX,
FFTW_MEASURE | FFTW_IN_PLACE); /*FFTW_ESTIMATE FFTW_MEASURE */
pi3 = rfftw3d_create_plan (Vx_max, Vy_max, Vz_max, FFTW_COMPLEX_TO_REAL,
FFTW_MEASURE | FFTW_IN_PLACE);
r3 = rfftw3d_create_plan (Rx_max, Rx_max, Rx_max, FFTW_REAL_TO_COMPLEX,
FFTW_MEASURE | FFTW_IN_PLACE); /*FFTW_ESTIMATE FFTW_MEASURE */
ri3 = rfftw3d_create_plan (Rx_max, Rx_max, Rx_max, FFTW_COMPLEX_TO_REAL,
FFTW_MEASURE | FFTW_IN_PLACE);
if (myrank == 0)
{
printf("Plans for FFTW created \n");fflush(stdout);
}
Volume = (fftw_real *) calloc (Vx_max * Vx_max * 2 * (Vx_max / 2 + 1),sizeof (fftw_real) );
Rot_tmpl = (float *) malloc (sizeof (float) * Rx_max * Ry_max * Rz_max);
Vol_tmpl = (float *) malloc (sizeof (float) * Vx_max * Vy_max * Vz_max);
conv = (float *) malloc (sizeof (float) * Vx_max * Vy_max * Vz_max);
sqconv = (fftw_real *) calloc(Vz_max * Vy_max * 2 * (Vx_max / 2 + 1), sizeof (fftw_real));
if (!
(inputdata1.floatdata =
(float *) malloc (sizeof (float) * Vx_max * Vy_max * Vz_max)))
{
printf ("Memory allocation failure in inputdata1.floatdata!!!");
fflush (stdout);
exit (1);
}
if (!
(outputdata.floatdata =
(float *) malloc (sizeof (float) * Vx_max * Vy_max * Vz_max)))
{
printf ("Memory allocation failure in outputdata.floatdata!!!");
fflush (stdout);
exit (1);
}
if (!
(Vol_tmpl_sort = (fftw_real *) calloc (Vz_max*Vy_max*2*(Vx_max / 2 + 1), sizeof (fftw_real) )))
{
printf ("Memory allocation failure in Volume_tmpl_sort!!!");
printf ("Nx = %i, Ny = %i, Nz = %i, bytes = %i \n",2 *(Vx_max / 2 + 1),Vy_max, Vz_max, sizeof (fftw_real));
fflush (stdout);
exit (1);
}
Ergebnis = (float *) calloc (Vz_max * Vy_max * Vx_max, sizeof (float));
/* Winkelraum */
Phi_min = atof (argv[4]);
Phi_max = atof (argv[5]);
Phi_step = atof (argv[6]);
Psi_min = atof (argv[7]);
Psi_max = atof (argv[8]);
Psi_step = atof (argv[9]);
Theta_min = atof (argv[10]);
Theta_max = atof (argv[11]);
Theta_step = atof (argv[12]);
/* Pointspread Function*/
read_em (argv[13], &inputdata3);
/* mask function */
read_em (argv[14], &inputdata4);
Phi_steps = (Phi_max - Phi_min) / Phi_step + 1;
Psi_steps = (Psi_max - Psi_min) / Psi_step + 1;
Theta_steps = (Theta_max - Theta_min) / Theta_step + 1;
winkel_max = Phi_steps * Psi_steps * Theta_steps;
winkel_min = 0;
range[0]=dim_fft-1;
range[1]=dim_fft-1;
range[2]=dim_fft-1;
range_sub[0]=range[0]-Rx_max;
range_sub[1]=range[1]-Rx_max;
range_sub[2]=range[2]-Rx_max;
sub[0]=1;
sub[1]=1;
sub[2]=1;
cycles=(int)(FullVolume_dims[2]/(dim_fft-Rx_max)+0.5);
cycles=(int)(FullVolume_dims[1]/(dim_fft-Rx_max)+0.5)*cycles;
cycles=(int)(FullVolume_dims[0]/(dim_fft-Rx_max)+0.5)*cycles;
cycle=0;
if (myrank == 0)
{
printf ("\n oscar starts to run ... ");tack (&start);fflush (stdout);
/* prepare Output */
strcpy (name, argv[3]);
strcat (name, ".ccf");
printf ("\nCreate outputfile: %s ... \n", name);fflush(stdout);
create_em (name, FullVolume_dims);
strcpy (name, argv[3]);
strcat (name, ".ang");
printf ("Create outputfile: %s ... \n", name);fflush(stdout);
create_em (name, FullVolume_dims);
strcpy (name, argv[3]);
strcat (name, ".ccf.norm");
printf ("Create outputfile: %s ... \n", name);fflush(stdout);
create_em (name, FullVolume_dims);
}
for (sub[2]=1; sub[2] < FullVolume_dims[2]-Rz_max;sub[2]=sub[2]+dim_fft-Rz_max)
{
if (myrank == 0)
{
tack (&start);
printf ("%f%%..", (float) (cycle / cycles * 100));
fflush (stdout);
}
for (sub[1]=1; sub[1] < FullVolume_dims[1]-Ry_max;sub[1]=sub[1]+dim_fft-Ry_max)
{
for (sub[0]=1; sub[0] < FullVolume_dims[0]-Rx_max;sub[0]=sub[0]+dim_fft-Rx_max)
{
cycle=cycle+1;
subc[0]=sub[0];
subc[1]=sub[1];
subc[2]=sub[2];
if (sub[2] + range[2] > FullVolume_dims[2]) subc[2]=FullVolume_dims[2]-range[2]; /* we are at the corner ?!*/
if (sub[1] + range[1] > FullVolume_dims[1]) subc[1]=FullVolume_dims[1]-range[1]; /* we are at the corner ?!*/
if (sub[0] + range[0] > FullVolume_dims[0]) subc[0]=FullVolume_dims[0]-range[0]; /* we are at the corner ?!*/
read_em_subregion (argv[1], &inputdata1,subc,range);
read_em_subregion (argv[1], &outputdata,subc,range);
/* Umsortieren der Daten */
lauf = 0;
for (k = 0; k < Vz_max; k++)
{
for (j = 0; j < Vy_max; j++)
{
for (i = 0; i < Vx_max; i++)
{
/* square - needed for normalization */
sqconv[i + 2 * (Vx_max / 2 + 1) * (j + Vy_max * k)] = inputdata1.floatdata[lauf]*inputdata1.floatdata[lauf];
Volume[i + 2 * (Vx_max / 2 + 1) * (j + Vy_max * k)] = inputdata1.floatdata[lauf];
inputdata1.floatdata[lauf] = -1.0; /* kleine Zahl wg Max-Op , hier kommen die CCFs rein*/
outputdata.floatdata[lauf] = -1.0; /* hier kommen die Winkel rein*/
lauf++;
}
}
}
rfftwnd_one_real_to_complex (p3, &Volume[0], NULL); /* einmalige fft von Suchvolumen */
rfftwnd_one_real_to_complex (p3, &sqconv[0], NULL); /* FFT of square*/
winkel_step_pe = (int) winkel_max / mysize;
winkel_min_pe = myrank * winkel_step_pe;
winkel_max_pe = winkel_min_pe + winkel_step_pe;
Theta_winkel_nr = (int) winkel_min_pe / (Psi_steps * Phi_steps);
Theta_winkel_rest_nr = winkel_min_pe - Theta_winkel_nr * (Psi_steps * Phi_steps);
Psi_winkel_nr = (int) Theta_winkel_rest_nr / (Phi_steps);
Psi_winkel_rest_nr = Theta_winkel_rest_nr - Psi_winkel_nr * (Phi_steps);
Phi_winkel_nr = (int) Psi_winkel_rest_nr;
Theta = Theta_winkel_nr * Theta_step + Theta_min;
Phi = Phi_winkel_nr * Phi_step + Phi_min - Phi_step;
Psi = Psi_winkel_nr * Psi_step + Psi_min;
eps = 0.001;
n = 0;
//Friedrich -> Zaehlung der voxels
n = countvoxel(inputdata4.dims[0], inputdata4.floatdata, eps);
eps = 0.001;
for (winkel_lauf = winkel_min_pe; winkel_lauf < winkel_max_pe;winkel_lauf++)
{
if (Phi < Phi_max)
Phi = Phi + Phi_step;
else
{
Phi = Phi_min;
Psi = Psi + Psi_step;
}
if (Psi > Psi_max)
{
Psi = Psi_min;
Theta = Theta + Theta_step;
}
tom_rotate3d (&Rot_tmpl[0], &inputdata2.floatdata[0], Phi, Psi, Theta, Rx_max, Ry_max, Rz_max);
/*calculate Ref variance */
rms_wedge = energizer (Rx_min, Rx_max, n, &Rot_tmpl[0], &inputdata3.floatdata[0], &inputdata4.floatdata[0], r3, ri3);
pastes (&Rot_tmpl[0], &Vol_tmpl[0], 1, 1, 1, Rx_max, Ry_max, Rz_max, Vx_max);
scale = 1.0 / ((double)Vx_max * (double)Vy_max * (double)Vz_max * ((double) rms_wedge) );
//printf("hippo1: scale = %.10f \n",scale);
sort4fftw(&Vol_tmpl_sort[0],&Vol_tmpl[0],Vx_max, Vy_max, Vz_max);
rfftwnd_one_real_to_complex (p3, &Vol_tmpl_sort[0], NULL);
PointVolume = (fftw_complex *) & Volume[0];
C3 = (fftw_complex *) & Vol_tmpl_sort[0];
/* Correlation */
correl(&PointVolume[0], &C3[0], Vx_max, Vy_max, Vz_max, scale);
/* back to real space */
rfftwnd_one_complex_to_real (pi3, &C3[0], NULL);
PointCorr = (fftw_real *) & C3[0];
/* Umsortieren der Daten */
sortback4fftw( &PointCorr[0], &Ergebnis[0], Vx_max, Vy_max, Vz_max);
// crossen
cross(&Ergebnis[0], Vx_max);
/* 3rd: divide */
lauf = 0;
for (k = 0 ; k < Vz_max ; k++)
{
for (j = 0; j < Vy_max; j++)
{
for (i = 0; i < Vx_max; i++)
{
if (inputdata1.floatdata[lauf] < Ergebnis[lauf] )
{
inputdata1.floatdata[lauf] = Ergebnis[lauf];
outputdata.floatdata[lauf] = (int) winkel_lauf;
}
lauf++;
}
}
}
} /* Ende winkel_lauf */
//FF
MPI_Barrier (MPI_COMM_WORLD);
/* Ergebnisse einsammeln (myrank 0)*/
if (myrank == 0)
{
for (lauf_pe = 1; lauf_pe < mysize; lauf_pe++)
{
MPI_Recv (&Ergebnis[0], Vx_max * Vy_max * Vz_max, MPI_FLOAT, lauf_pe,
99, MPI_COMM_WORLD, &status);
MPI_Recv (&conv[0], Vx_max * Vy_max * Vz_max, MPI_FLOAT,
lauf_pe, 98, MPI_COMM_WORLD, &status);
/* use conv as temporary memory for angles */
for (lauf = 0; lauf < Vx_max * Vy_max * Vz_max; lauf++)
{
if (inputdata1.floatdata[lauf] < Ergebnis[lauf])
{
inputdata1.floatdata[lauf] = Ergebnis[lauf];
outputdata.floatdata[lauf] = conv[lauf];
}
}
}
/*Ergebnisse eingesammelt */
}
// myrank > 0: Ergebnisse senden
else
{
MPI_Send (inputdata1.floatdata, Vx_max * Vy_max * Vz_max, MPI_FLOAT, 0,
99, MPI_COMM_WORLD);
MPI_Send (outputdata.floatdata, Vx_max * Vy_max * Vz_max, MPI_FLOAT, 0,
98, MPI_COMM_WORLD);
}
MPI_Barrier (MPI_COMM_WORLD);
// nicht normalisiertes Volumen und Winkel rausschreiben
subc[0]=subc[0]+Rx_max/2;
subc[1]=subc[1]+Rx_max/2;
subc[2]=subc[2]+Rx_max/2;
if (myrank==0)
{
offset[0]=Rx_max/2;
offset[1]=Rx_max/2;
offset[2]=Rx_max/2;
dimarray[0]=dim_fft;
dimarray[1]=dim_fft;
dimarray[2]=dim_fft;
strcpy (name, argv[3]);
strcat (name, ".ccf");
write_em_subsubregion (name, &inputdata1,subc,range_sub,offset,dimarray);
strcpy (name, argv[3]);
strcat (name, ".ang");
write_em_subsubregion (name, &outputdata,subc,range_sub,offset,dimarray);
/* ------------------- normalization - here only PE 0 ---------- */
pastes (&inputdata4.floatdata[0], &Vol_tmpl[0], 1, 1, 1, Rx_max, Ry_max, Rz_max, Vx_max); /* paste mask into zero volume*/
/* 1st local mean */
sort4fftw(&Vol_tmpl_sort[0], &Vol_tmpl[0], Vx_max, Vy_max, Vz_max);
rfftwnd_one_real_to_complex (p3, &Vol_tmpl_sort[0], NULL);
C3 = (fftw_complex *) & Vol_tmpl_sort[0];
/* Convolution of volume and mask */
scale = 1.0 / ((double)Vx_max * (double)Vy_max * (double)Vz_max );
convolve( &PointVolume[0], &C3[0], Vx_max, Vy_max, Vz_max, scale);
rfftwnd_one_complex_to_real (pi3, &C3[0], NULL);
PointCorr = (fftw_real *) & C3[0];
/* Umsortieren der Daten */
sortback4fftw( &PointCorr[0], &conv[0], Vx_max, Vy_max, Vz_max);
/* 2nd : convolution of square and resorting*/
pastes (&inputdata4.floatdata[0], &Vol_tmpl[0], 1, 1, 1, Rx_max, Ry_max, Rz_max, Vx_max); /* paste mask into zero volume*/
sort4fftw( &Vol_tmpl_sort[0], &Vol_tmpl[0], Vx_max, Vy_max, Vz_max);
rfftwnd_one_real_to_complex (p3, &Vol_tmpl_sort[0], NULL);
C3 = (fftw_complex *) & Vol_tmpl_sort[0];
PointSq = (fftw_complex *) & sqconv[0];// set pointer to FFT of square
convolve( &PointSq[0], &C3[0], Vx_max, Vy_max, Vz_max, scale);
rfftwnd_one_complex_to_real (pi3, &C3[0], NULL);
PointCorr = (fftw_real *) &C3[0];
//FF
lauf = 0;
for (k = 0; k < Vz_max; k++)
{
for (j = 0; j < Vy_max; j++)
{
for (i = 0; i < Vx_max; i++)
{
conv[lauf] = sqrt(PointCorr[i + 2 * (Vx_max / 2 + 1) * (j + Vy_max * k)] - conv[lauf]*conv[lauf]/((float) n) ) ;/*local variance*/
lauf++;
}
}
}
cross(&conv[0], Vx_max);
/* perform division */
for (lauf = 0; k < Vz_max*Vy_max*Vz_max; lauf++)
{
if (conv[lauf] > eps)
{
inputdata1[lauf].floatdata = inputdata1[lauf].floatdata/conv[lauf];
}
else
{
inputdata1[lauf].floatdata = 0;
}
}
strcpy (name, argv[3]);
strcat (name, ".ccf.norm");
write_em_subsubregion (name, &inputdata1,subc,range_sub,offset,dimarray);
}
MPI_Barrier (MPI_COMM_WORLD);
}
} /* these are the new brackets from the subregion_read , SN */
}
free(Ergebnis);
free(inputdata1.floatdata);
free(inputdata2.floatdata);
free(inputdata3.floatdata);
free(inputdata4.floatdata);
rfftwnd_destroy_plan(p3);
rfftwnd_destroy_plan(pi3);
rfftwnd_destroy_plan(r3);
rfftwnd_destroy_plan(ri3);
free(Volume);
free(sqconv);
free(conv);
free(Rot_tmpl);
free(Vol_tmpl_sort);
free(outputdata.floatdata);
if (myrank==0)
{
printf ("oscar finished. ");
tack (&start); fflush(stdout);
}
MPI_Finalize();
/* end main */
}
int main(int argc, char **argv)
{
int N;
int nThreads;
int nColumns;
int i,j,k;
double *A,*Bi,*C,*Ci;
int BiRows, BiColumns;
CompressedMatrix *cBi;
CompressedMatrix *cCi;
double elapsed;
char printDebug;
//************ Check Input **************/
if(argc < 3){
printf("Usage: %s MaxtrixSize NumberOfThreads\n" , argv[0] );
exit(EXIT_FAILURE);
}
N = atoi(argv[1]);
if( N <= 1){
printf("MatrixSize must be bigger than 1!");
exit(EXIT_FAILURE);
}
nThreads = atoi(argv[2]);
if( nThreads <= 1){
printf("NumberOfThreads must be bigger than 1!");
exit(EXIT_FAILURE);
}
omp_set_num_threads(nThreads);
omp_set_schedule(omp_sched_dynamic, N/10);
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &mpi_id);
MPI_Comm_size(MPI_COMM_WORLD, &mpi_size);
nColumns = N / mpi_size; //For the moment depend on N being a multiple the number of MPI nodes
//************ Prepare Matrix **************/
A = (double *) malloc( N*N * sizeof(double) );
if((A == NULL) ){
printf("Running out of memory!\n"); exit(EXIT_FAILURE);
}
// if(mpi_id != 0){
// MPI_Finalize();
// exit(0);
// }
if(mpi_id == 0)
{
printDebug = 0;
if(printDebug) printf("[%d] Generating A ...",mpi_id);
//Fill matrixes. Generate Identity like matrix for A and B , So C should result in an matrix with a single major diagonal
for(i=0; i < N; i++ ){
for(j=0; j < N; j++){
A[i+N*j] = (i==j)?i:0.0;
// //Sparse Matrix with 10% population
// A[i+N*j] = rand()%10;
// if(A[i+N*j] == 0)
// A[i+N*j] = rand()%10;
// else
// A[i+N*j] = 0;
}
}
// printMatrix(A, N, nColumns);
// cA = compressMatrix(A, N, nColumns);
// printCompressedMatrix(cA);
// uncompressMatrix(cA, &Bi, &i, &j);
// printMatrix(Bi, i, j);
//
// MPI_Finalize();
// exit(0);
tick();
if(printDebug) printf("[%d] Broadcasting A ...",mpi_id);
MPI_Bcast( A, N*N, MPI_DOUBLE, 0, MPI_COMM_WORLD);
if(printDebug) printf("[%d] Generating B ...",mpi_id);
double* B; CompressedMatrix* cB;
B = (double *) malloc( N*N * sizeof(double) );
for(i=0; i < N; i++ ){
for(j=0; j < N; j++){
B[j+N*i] = (i==j)?1.0:0.0;
}
}
if(printDebug) printf("[%d] Compressing and distributing Bi ...",mpi_id);
cB = compressMatrix(B, N, N);
for(i=1; i < mpi_size; i++){
mpiSendCompressedMatrix(cB, i*nColumns, (i+1)*nColumns, i);
}
//Fake shorten cB
free(B);
cB->columns = nColumns;
uncompressMatrix(cB, &Bi, &BiRows, &BiColumns);
Ci = MatrixMultiply(A, N, N, Bi, nColumns);
if(printDebug) printf("[%d] Ci = A x Bi ...", mpi_id);
if(printDebug) printMatrix(Ci, N, nColumns);
cCi = compressMatrix(Ci, N, nColumns);
if(printDebug) printf("cCi ...\n");
if(printDebug) printCompressedMatrix(cCi);
MPI_Barrier(MPI_COMM_WORLD);
if(printDebug) printf("[%d] Receiving Ci fragments ...\n", mpi_id);
CompressedMatrix** Cii;
Cii = (CompressedMatrix**) malloc(sizeof(CompressedMatrix*) * mpi_size);
if(Cii == NULL){ perror("malloc"); exit(EXIT_FAILURE); }
Cii[0] = cCi;
for(i=1; i < mpi_size; i++){
Cii[i] = mpiRecvCompressedMatrix(N,nColumns, i);
}
if(printDebug) printf("[%d] Joining Cii ...\n", mpi_id);
CompressedMatrix *cC;
cC = joinCompressedMatrices(Cii, mpi_size);
if(printDebug) printCompressedMatrix(cC);
elapsed = tack();
printf("[%d] C ...\n", mpi_id);
uncompressMatrix(cC, &C, &i,&j);
if(i <= 20){
printMatrix(C, i,j);
} else {
if(i < 1000){
printf("C is too big, only printing first diagonal %d.\n[",j);
for(k=0; (k < i) && (k < j); k++){
printf("%3.2f ",C[k + k*j]);
}
printf("]\n");
} else {
printf("C is just too big!");
}
}
printf("Took [%f] seconds!\n",elapsed);
} else {
printDebug = 0;
if(printDebug) printf("[%d] Waiting for A ...",mpi_id);
MPI_Bcast( A, N*N, MPI_DOUBLE, 0, MPI_COMM_WORLD);
if(printDebug) printf("[%d] Received A ...\n", mpi_id);
if(printDebug) printMatrix(A, N, N);
if(printDebug) printf("[%d] Waiting for Bi ...",mpi_id);
cBi = mpiRecvCompressedMatrix(N, nColumns, 0);
uncompressMatrix(cBi, &Bi, &BiRows, &BiColumns);
if(printDebug) printf("[%d] Received Bi ...",mpi_id);
if(printDebug) printMatrix(Bi,BiRows, BiColumns);
assert( (BiRows == N) && "Number or Rows in Bi is not right!");
assert( (BiColumns == nColumns) && "Number or Columns in Bi is not right!");
Ci = MatrixMultiply(A, N, N, Bi, BiColumns);
if(printDebug) printf("[%d] Ci = A x Bi ...", mpi_id);
if(printDebug) printMatrix(Ci, N, nColumns);
cCi = compressMatrix(Ci, N, nColumns);
if(printDebug) printCompressedMatrix(cCi);
MPI_Barrier(MPI_COMM_WORLD);
if(printDebug) printf("[%d] Returning Ci ...\n", mpi_id);
mpiSendCompressedMatrix(cCi, 0, nColumns, 0);
}
MPI_Finalize();
// NxM = NxN * NxM
exit(EXIT_SUCCESS);
}
