void PhaseFit::PrintResidual(float *params) {
unsigned int nRead = signal.size();
unsigned int nData = nRead * nFlow;
float *tmp = new float[nData];
Evaluate(tmp,params);
float *y = new float[nData];
for(unsigned int iRead=0, iY=0; iRead<nRead; iRead++)
for(unsigned int iFlow=0; iFlow<nFlow; iFlow++)
y[iY++] = (float) signal[iRead][iFlow];
float res = CalcResidual(y, tmp, len);
cout << "Residual = " << setiosflags(ios::fixed) << setprecision(3) << res << "\n";
for(unsigned int iRead=0, iY=0; iRead<std::min(nRead,(unsigned int)2); iRead++) {
cout << "raw res [" << iRead << "]: ";
unsigned int iYsave = iY;
for(unsigned int iFlow=0; iFlow<nFlow; iFlow++,iY++)
cout << " " << setiosflags(ios::fixed) << setprecision(3) << y[iY] - tmp[iY];
cout << "\n";
iY = iYsave;
cout << "wt res [" << iRead << "]: ";
for(unsigned int iFlow=0; iFlow<nFlow; iFlow++,iY++)
cout << " " << setiosflags(ios::fixed) << setprecision(3) << residualWeight[iY] * (y[iY] - tmp[iY]);
cout << "\n";
}
delete [] y;
delete [] tmp;
}
/////////////////////////////////////
// Main Method //
/////////////////////////////////////
int main()
{
double tstart0, tstop0, ttime0;
double tstart0_imp, tstop0_imp, ttime0_imp;
std::cout<<"Reading file.."<<std::endl << std::endl;
// Read in the vector
std::vector<seed_t> seed_vector = ReadFile();
seed_vector = ExtendVector(seed_vector, 100);
const int seed_entries = seed_vector.size();
////////////////////////////
// Structs: Residual
////////////////////////////
float * residuals;
residuals = (float*) _mm_malloc(sizeof(float)*seed_entries,64);
std::cout<<"Structs: Residual" << std::endl;
int i = 0, num_threads, chunk;
#pragma omp parallel
#pragma omp master
num_threads = omp_get_num_threads();
chunk = seed_entries/num_threads;
tstart0 = dtime();
#pragma ivdep
{
//#pragma omp parallel for private(i)
#pragma omp parallel for simd private(i)
for(i = 0; i < seed_entries; i++)
{
residuals[i] = CalcResidual(seed_vector[i]);
}
}
tstop0 = dtime();
ttime0 = tstop0 - tstart0;
std::cout<<" Elapsed time: " << ttime0 <<"s"<< std::endl<<std::endl;
////////////////////////////
// Structs: Square Residual
////////////////////////////
float * square_residuals;
square_residuals = (float*) _mm_malloc(sizeof(float)*seed_entries,64);
std::cout<<"Structs: Residual Squared" << std::endl;
//int i, num_threads, chunk;
tstart0_imp = dtime();
#pragma ivdep
{
//#pragma omp parallel for private(i)
#pragma omp parallel for simd private(i)
for(i = 0; i < seed_entries; i++)
{
square_residuals[i] = CalcSquareResidual(seed_vector[i]);
}
}
tstop0_imp = dtime();
ttime0_imp = tstop0_imp - tstart0_imp;
std::cout<<" Elapsed time: " << ttime0_imp <<"s"<< std::endl<<std::endl;
////////////////////////////
// Error Checking
////////////////////////////
int tot_errs = 0;
for(int i = 0; i < seed_entries; i++)
{
float epsilon = 0.00001;
if(fabs(residuals[i] - sqrt(square_residuals[i])) > epsilon)
{
std::cout<<residuals[i] - sqrt(square_residuals[i]) << std::endl;
tot_errs++;
}
}
std::cout<<"Total errors for square method: " << tot_errs << std::endl;
_mm_free(residuals);
_mm_free(square_residuals);
return 0;
}
