// jscott: added 3d Fourier transform - cost 3N^2
void idFFT::FFT3D( cpxFloat_t *data, int N, int ISI )
{
cpxFloat_t *orig;
int i, j;
orig = data;
// Transform each slice
for( j = 0; j < N; j++ )
{
// 1D transform
data = orig + j * N * N;
for( i = 0; i < N; i++ )
{
FFT1D( data, N, ISI, 1 );
data += N;
}
// 1D transform
data = orig + j * N * N;
for( i = 0; i < N; i++ )
{
FFT1D( data, N, ISI, N );
data++;
}
}
// Transform the volume
data = orig;
for( j = 0; j < N * N; j++ )
{
FFT1D( data, N, ISI, N * N );
data++;
}
}
// RAVEN BEGIN
// jscott: added 2d Fourier transform - cost 2N
void idFFT::FFT2D( cpxFloat_t *data, int N, int ISI )
{
cpxFloat_t *orig;
int i;
orig = data;
// 1D horizontal transform
for( i = 0; i < N; i++ )
{
FFT1D( data, N, ISI, 1 );
data += N;
}
// 1D vertical transform
data = orig;
for( i = 0; i < N; i++ )
{
FFT1D( data, N, ISI, N );
data++;
}
}
void SlaveStart()
{
long i;
long MyNum;
double *upriv;
long initdone;
long finish;
long l_transtime=0;
long MyFirst;
long MyLast;
LOCK(Global->idlock);
MyNum = Global->id;
Global->id++;
UNLOCK(Global->idlock);
BARINCLUDE(Global->start);
/* POSSIBLE ENHANCEMENT: Here is where one might pin processes to
processors to avoid migration */
BARRIER(Global->start, P);
upriv = (double *) malloc(2*(rootN-1)*sizeof(double));
if (upriv == NULL) {
fprintf(stderr,"Proc %ld could not malloc memory for upriv\n",MyNum);
exit(-1);
}
for (i=0;i<2*(rootN-1);i++) {
upriv[i] = umain[i];
}
MyFirst = rootN*MyNum/P;
MyLast = rootN*(MyNum+1)/P;
TouchArray(x, trans, umain2, upriv, MyFirst, MyLast);
BARRIER(Global->start, P);
/* POSSIBLE ENHANCEMENT: Here is where one might reset the
statistics that one is measuring about the parallel execution */
if ((MyNum == 0) || (dostats)) {
CLOCK(initdone);
}
/* perform forward FFT */
FFT1D(1, M, N, x, trans, upriv, umain2, MyNum, &l_transtime, MyFirst,
MyLast, pad_length, test_result, dostats);
/* perform backward FFT */
if (test_result) {
FFT1D(-1, M, N, x, trans, upriv, umain2, MyNum, &l_transtime, MyFirst,
MyLast, pad_length, test_result, dostats);
}
if ((MyNum == 0) || (dostats)) {
CLOCK(finish);
Global->transtimes[MyNum] = l_transtime;
Global->totaltimes[MyNum] = finish-initdone;
}
if (MyNum == 0) {
Global->finishtime = finish;
Global->initdonetime = initdone;
}
}
void SlaveStart()
{
long i;
long MyNum;
double *upriv;
//int a = 2*(rootN-1)*sizeof(double);
//double upriv[a];
long initdone;
long finish;
long l_transtime=0;
long MyFirst;
long MyLast;
BARRIER(Global->start, P);
LOCK(Global->idlock);
MyNum = Global->id;
Global->id++;
UNLOCK(Global->idlock);
BARINCLUDE(Global->start);
BARRIER(Global->start, P);
//upriv = (double *) malloc(2*(rootN-1)*sizeof(double));
upriv = (double *) our_malloc(2*(rootN-1)*sizeof(double));
if (upriv == NULL) {
fprintf(stderr,"Proc %ld could not malloc memory for upriv\n",MyNum);
exit(-1);
}
for (i=0;i<2*(rootN-1);i++) {
upriv[i] = umain[i];
}
MyFirst = rootN*MyNum/P;
MyLast = rootN*(MyNum+1)/P;
TouchArray(x, trans, umain2, upriv, MyFirst, MyLast);
BARRIER(Global->start, P);
if ((MyNum == 0) || (dostats)) {
CLOCK(initdone);
}
//printf("\nentrando em forward FFT\n");
/* perform forward FFT */
FFT1D(1, M, N, x, trans, upriv, umain2, MyNum, &l_transtime, MyFirst,
MyLast, pad_length, test_result, dostats);
/* perform backward FFT */
if (test_result) {
FFT1D(-1, M, N, x, trans, upriv, umain2, MyNum, &l_transtime, MyFirst,
MyLast, pad_length, test_result, dostats);
}
if ((MyNum == 0) || (dostats)) {
CLOCK(finish);
Global->transtimes[MyNum] = l_transtime;
Global->totaltimes[MyNum] = finish-initdone;
}
if (MyNum == 0) {
Global->finishtime = finish;
Global->initdonetime = initdone;
}
join_point(&myJoinPoint);
}
