long matrix::mul_ikj(const matrix *a, const matrix *b, matrix *r)
{
int width, height, h_div, h_mod;
a->size(&height, &width);
h_div = height / 16;
h_mod = height % 16;
for (int i = 0; i < height; i++) {
for (iter_row r_i(r, i); *r_i; r_i++)
**r_i = 0;
iter_row a_i(a, i);
for (int k = 0; k < width; k++) {
number n = **a_i; a_i++;
iter_row r_i(r, i);
iter_row b_i(b, k);
for (int j = 0; j < h_div; j++) {
(*r_i)[0] += (*b_i)[0] * n;
(*r_i)[1] += (*b_i)[1] * n;
(*r_i)[2] += (*b_i)[2] * n;
(*r_i)[3] += (*b_i)[3] * n;
(*r_i)[4] += (*b_i)[4] * n;
(*r_i)[5] += (*b_i)[5] * n;
(*r_i)[6] += (*b_i)[6] * n;
(*r_i)[7] += (*b_i)[7] * n;
(*r_i)[8] += (*b_i)[8] * n;
(*r_i)[9] += (*b_i)[9] * n;
(*r_i)[10] += (*b_i)[10] * n;
(*r_i)[11] += (*b_i)[11] * n;
(*r_i)[12] += (*b_i)[12] * n;
(*r_i)[13] += (*b_i)[13] * n;
(*r_i)[14] += (*b_i)[14] * n;
(*r_i)[15] += (*b_i)[15] * n;
r_i += 16; b_i += 16;
}
for (int j = 0; j < h_mod; j++) {
**r_i += **b_i * n;
r_i++; b_i++;
}
}
}
return (long)width * width * height * 2;
}
long matrix::mul_ijk(const matrix *a, const matrix *b, matrix *r)
{
int width, height;
a->size(&height, &width);
for (int i = 0; i < height; i++) {
iter_row r_i(r, i);
for (int j = 0; j < height; j++) {
iter_row a_i(a, i);
iter_col b_i(b, j);
double n = 0;
for (int k = 0; k < width; k++) {
n += **a_i * **b_i;
a_i++; b_i++;
}
**r_i = n;
r_i++;
}
}
return (long)width * width * height * 2;
}
void updateVerletList(const string &verletStringId,
Particles & particles,
Grid & grid,
double radius)
{
double radiusSquared = radius*radius;
int verletId = particles.getVerletId(verletStringId);
const unordered_map<int, GridPoint*> &gridpoints = grid.gridpoints();
const mat & R = particles.r();
const vector<int> &mygridPoints = grid.myGridPoints();
particles.clearVerletList(verletId);
#ifdef USE_OPENMP
#pragma omp parallel for
#endif
for(int i=0; i<mygridPoints.size(); i++)
{
double dx, dy, dz;
int gridId = mygridPoints.at(i);
const GridPoint & gridPoint = *gridpoints.at(gridId);
for(const pair<int, int> & idCol_i:gridPoint.particles())
{
int id_i = idCol_i.first;
vector<int> verletList;
const vec & r_i = R.col(idCol_i.second);
for(const pair<int, int> & idCol_j:gridPoint.particles())
{
int id_j = idCol_j.first;
if(id_i == id_j)
continue;
const vec & r_j = R.col(idCol_j.second);
dx = r_i(0) - r_j(0);
dy = r_i(1) - r_j(1);
dz = r_i(2) - r_j(2);
double drSquared = dx*dx + dy*dy + dz*dz;
if(drSquared < radiusSquared)
{
verletList.push_back(id_j);
}
}
// Neighbouring cells
const vector<GridPoint*> & neighbours = gridPoint.neighbours();
for(const GridPoint *neighbour:neighbours)
{
for(const pair<int, int> & idCol_j:neighbour->particles())
{
const vec & r_j = R.col(idCol_j.second);
dx = r_i(0) - r_j(0);
dy = r_i(1) - r_j(1);
dz = r_i(2) - r_j(2);
double drSquared = dx*dx + dy*dy + dz*dz;
if(drSquared < radiusSquared)
{
verletList.push_back(idCol_j.first);
}
}
}
#ifdef USE_OPENMP
#pragma omp critical
{
particles.setVerletList(id_i, verletList, verletId);
}
#else
particles.setVerletList(id_i, verletList, verletId);
#endif
}
}
}
