int main(int argc, char* argv[]) {
if (argc != 2)
fprintf(stderr, "Usage: Requires number of threads.");
thread_count = atoi(argv[1]);
Lab3LoadInput(&A, &size);
x = CreateVec(size);
double start, end;
int i = 0;
GET_TIME(start);
# pragma omp parallel num_threads(thread_count) \
shared(A)
{
gaussian_elimination();
jordan_elimination();
# pragma omp for
for (i = 0; i < size; ++i) {
x[i] = A[i][size] / A[i][i];
}
}
GET_TIME(end);
Lab3SaveOutput(x, size, end-start);
printf("time is: %e\n", end-start);
DestroyVec(x);
DestroyMat(A, size);
return EXIT_SUCCESS;
}
int main(int argc, char* argv[])
{
int i, j, k, size;
double** Au;
double* X;
double temp, error, Xnorm;
int* index;
FILE* fp;
/*Load the datasize and verify it*/
Lab3LoadInput(&Au, &size);
if ((fp = fopen("data_output","r")) == NULL){
printf("Fail to open the result data file!\n");
return 2;
}
fscanf(fp, "%d\n\n", &i);
if (i != size){
printf("The problem size of the input file and result file does not match!\n");
return -1;
}
/*Calculate the solution by serial code*/
X = CreateVec(size);
index = malloc(size * sizeof(int));
for (i = 0; i < size; ++i)
index[i] = i;
if (size == 1)
X[0] = Au[0][1] / Au[0][0];
else{
/*Gaussian elimination*/
for (k = 0; k < size - 1; ++k){
/*Pivoting*/
temp = 0;
for (i = k, j = 0; i < size; ++i)
if (temp < Au[index[i]][k] * Au[index[i]][k]){
temp = Au[index[i]][k] * Au[index[i]][k];
j = i;
}
if (j != k)/*swap*/{
i = index[j];
index[j] = index[k];
index[k] = i;
}
/*calculating*/
for (i = k + 1; i < size; ++i){
temp = Au[index[i]][k] / Au[index[k]][k];
for (j = k; j < size + 1; ++j)
Au[index[i]][j] -= Au[index[k]][j] * temp;
}
}
/*Jordan elimination*/
for (k = size - 1; k > 0; --k){
for (i = k - 1; i >= 0; --i ){
temp = Au[index[i]][k] / Au[index[k]][k];
Au[index[i]][k] -= temp * Au[index[k]][k];
Au[index[i]][size] -= temp * Au[index[k]][size];
}
}
/*solution*/
for (k=0; k< size; ++k)
X[k] = Au[index[k]][size] / Au[index[k]][k];
}
/*compare the solution*/
error = 0;
Xnorm = 0;
for (i = 0; i < size; ++i){
fscanf(fp, "%lf\t", &temp);
error += (temp-X[i]) * (temp-X[i]);
Xnorm += X[i]*X[i];
}
error = sqrt(error);
Xnorm = sqrt(Xnorm);
printf("The relative error to the reference solution is %e\n", error / Xnorm);
if (error / Xnorm <= TOL)
printf("Congratulation!!! Your result is accepted!\n");
else
printf("Sorry, your result is wrong.\n");
fclose(fp);
DestroyVec(X);
DestroyMat(Au, size);
free(index);
return 0;
}
Geometry CreateGeometry(int N[3], double h[3], int Npml[3], int Nc, int LowerPML, double *eps, double *epsI, double *fprof, double wa, double y){
int i;
Geometry geo = (Geometry) malloc(sizeof(struct Geometry_s));
geo->Nc = Nc;
geo->LowerPML = LowerPML;
geo->interference = 0.0; // default no interference
for(i=0; i<3; i++){
geo->h[i] = h[i];
geo->Npml[i] = Npml[i];
}
CreateGrid(&geo->gN, N, geo->Nc, 2);
CreateGrid(&geo->gM, N, 1, 1); // 3/3/14: set M = N as per Steven
CreateVec(2*Nxyzc(geo)+2, &geo->vepspml);
int manual_epspml = 0;
PetscOptionsGetInt(PETSC_NULL,PETSC_NULL,"-manual_epspml", &manual_epspml, NULL);
if(manual_epspml == 0){
Vecfun pml;
CreateVecfun(&pml,geo->vepspml);
for(i=pml.ns; i<pml.ne; i++){
Point p;
CreatePoint_i(&p, i, &geo->gN);
project(&p, 3);
dcomp eps_geoal;
eps_geoal = pmlval(xyzc(&p), N, geo->Npml, geo->h, geo->LowerPML, 0);
setr(&pml, i, p.ir? cimag(eps_geoal) : creal(eps_geoal) );
}
DestroyVecfun(&pml);
}
CreateVec(Mxyz(geo), &geo->vMscratch[0]);
for(i=0; i<SCRATCHNUM; i++){
geo->vNhscratch[i] = 0; // allows checking whether vN created or not
if(i>0)VecDuplicate(geo->vMscratch[0], &geo->vMscratch[i]);
}
double *scratch;
int ms, me;
VecGetOwnershipRange(geo->vMscratch[0], &ms, &me);
if( !manual_epspml){
VecGetArray(geo->vMscratch[0], &scratch);
for(i=ms; i<me;i++)
scratch[i-ms] = eps[i-ms];
VecRestoreArray(geo->vMscratch[0], &scratch);
}
CreateVec(2*Nxyzc(geo)+2, &geo->vH);
VecDuplicate(geo->vH, &geo->veps);
VecDuplicate(geo->vH, &geo->vIeps);
for(i=0; i<SCRATCHNUM; i++) VecDuplicate(geo->vH, &geo->vscratch[i]);
VecSet(geo->vH, 1.0);
if( !manual_epspml){
VecShift(geo->vMscratch[0], -1.0); //hack, for background dielectric
InterpolateVec(geo, geo->vMscratch[0], geo->vscratch[1]);
VecShift(geo->vscratch[1], 1.0);
VecPointwiseMult(geo->veps, geo->vscratch[1], geo->vepspml);
if(epsI != NULL){ // imaginary part of passive dielectric
VecGetArray(geo->vMscratch[0], &scratch);
for(i=ms; i<me; i++){
scratch[i-ms] = epsI[i-ms];
}
VecRestoreArray(geo->vMscratch[0], &scratch);
InterpolateVec(geo, geo->vMscratch[0], geo->vscratch[1]);
VecPointwiseMult(geo->vscratch[1], geo->vscratch[1], geo->vepspml);
TimesI(geo, geo->vscratch[1], geo->vscratch[2]);
VecAXPY(geo->veps, 1.0, geo->vscratch[2]);
}
}
if(manual_epspml){
char epsManualfile[PETSC_MAX_PATH_LEN];
PetscOptionsGetString(PETSC_NULL,PETSC_NULL,"-epsManualfile", epsManualfile, PETSC_MAX_PATH_LEN, NULL);
FILE *fp = fopen(epsManualfile, "r");
ReadVectorC(fp, 2*Nxyzc(geo)+2, geo->veps);
// 07/11/15: if manual_epspml, then directly read in the Nxyzcr+2 vector
fclose(fp);
}
TimesI(geo, geo->veps, geo->vIeps);
// vIeps for convenience only, make sure to update it later if eps ever changes!
geo->D = 0.0;
geo->wa = wa;
geo->y = y;
VecDuplicate(geo->veps, &geo->vf);
VecDuplicate(geo->vMscratch[0], &geo->vfM);
VecGetArray(geo->vfM, &scratch);
for(i=ms; i<me;i++)
scratch[i-ms] = fprof[i-ms];
VecRestoreArray(geo->vfM, &scratch);
InterpolateVec(geo, geo->vfM, geo->vf);
return geo;
}
