void test_vmul_drv()
{
int n = 216 * 216 * 216;
int repeat = 20;
std::vector<double_complex> v1(n);
std::vector<double> v2(n);
mdarray<double_complex, 1> f1(n);
mdarray<double, 1> f2(n);
for (int i = 0; i < n; i++)
{
v1[i] = 1.0 / double_complex(i + 1, i + 1);
v2[i] = 2.0;
f1[i] = 1.0 / double_complex(i + 1, i + 1);
f2[i] = 2.0;
}
std::cout << "vector size: " << n << std::endl;
double t = kernel1(repeat, n, v1, v2);
std::cout << "kernel1 time: " << t << " speed: " << double(n * (2 * sizeof(double_complex) + sizeof(double)) * repeat) / t / (1 << 30) << " GBs" << std::endl;
t = kernel2(repeat, n, f1, f2);
std::cout << "kernel2 time: " << t << " speed: " << double(n * (2 * sizeof(double_complex) + sizeof(double)) * repeat) / t / (1 << 30) << " GBs" << std::endl;
t = kernel3(repeat, n, f1, v2);
std::cout << "kernel3 time: " << t << " speed: " << double(n * (2 * sizeof(double_complex) + sizeof(double)) * repeat) / t / (1 << 30) << " GBs" << std::endl;
}
/* sets the force due to the surface tension, according to equation (19)
* sets also the color field and its gradient
* ASSUMPTION: kernel(r,h)=kernel(-r,h)
* */
void setSurfaceT(System *sys, float(*kernel)(float*, float), float*(*kernel2)(float*, float), float(*kernel3)(float*, float)){
Particle *part1,*part2,*partM;
float r[3],h,coef,*coefGrad,coefLap,coef1,coef2,cfLap;
float norm,threshold=0.001;
int alpha;
partM=sys->parts+sys->nbParts;
h=sys->h;
for(part1=sys->parts;part1<partM;part1++){
part1->cf=0;
for(alpha=0;alpha<3;alpha++){
part1->fs[alpha]=0;
part1->gradcf[alpha]=0;
}
}
for(part1=sys->parts;part1<partM;part1++){
for(part2=part1;part2<partM;part2++){
r[0]=part1->pos[0]-part2->pos[0];
r[1]=part1->pos[1]-part2->pos[1];
r[2]=part1->pos[2]-part2->pos[2];
coef=kernel(r,h);
if(coef!=0){
coefGrad=kernel2(r,h);
coefLap=kernel3(r,h);
coef1=part2->mass/part2->ro;
coef2=part1->mass/part1->ro;
part1->cf+=coef1*coef;
for(alpha=0;alpha<3;alpha++)
part1->gradcf[alpha]+=coef1*coefGrad[alpha];
part1->fs[0]+=coef1*coefLap;
if(part2!=part1){
part2->cf+=coef2*coef;
for(alpha=0;alpha<3;alpha++)
part2->gradcf[alpha]-=coef2*coefGrad[alpha];
part2->fs[0]+=coef2*coefLap;
}
}
}
}
for(part1=sys->parts;part1<partM;part1++){
cfLap=part1->fs[0];
part1->fs[0]=0;
for(norm=0,alpha=0;alpha<3;alpha++)
norm+=part1->gradcf[alpha]*part1->gradcf[alpha];
norm=sqrt(norm);
if(norm>threshold){
cfLap*=-part1->sigma/norm;
for(alpha=0;alpha<3;alpha++)
part1->fs[alpha]=cfLap*part1->gradcf[alpha];
}
}
}
