void rt_set_numthreads(SceneHandle voidscene, int numthreads) {
scenedef * scene = (scenedef *) voidscene;
#ifdef THR
if (numthreads > 0) {
scene->numthreads = numthreads;
}
else {
scene->numthreads = rt_thread_numprocessors();
}
/* force set of # kernel threads */
rt_thread_setconcurrency(scene->numthreads);
#else
scene->numthreads = 1;
#endif
scene->scenecheck = 1;
}
int calc_grid_energies_excl_mgrid(float* atoms, float* grideners, long int numplane, long int numcol, long int numpt, long int natoms, float gridspacing, unsigned char* excludepos, int maxnumprocs) {
// cionize_params* params, cionize_molecule* molecule, cionize_grid* grid) {
int i;
enthrparms *parms;
rt_thread_t * threads;
#if defined(THR)
int numprocs;
int availprocs = rt_thread_numprocessors();
if (params->maxnumprocs <= availprocs) {
numprocs = params->maxnumprocs;
} else {
numprocs = availprocs;
}
#else
int numprocs = 1;
#endif
printf("calc_grid_energies_excl_mgrid()\n");
/* DH: setup and compute long-range */
MgridParam mg_prm;
MgridSystem mg_sys;
Mgrid mg;
//MgridLattice *lattice;
int j, k;
int gridfactor;
double h, h_1;
int nspacings;
double *eh;
int ndim;
int ii, jj, kk;
int im, jm, km;
int ilo, jlo, klo;
double dx_h, dy_h, dz_h;
double xphi[4], yphi[4], zphi[4];
double t, en, c;
int koff, jkoff, index;
rt_timerhandle timer;
float totaltime;
memset(&mg_prm, 0, sizeof(mg_prm));
memset(&mg_sys, 0, sizeof(mg_sys));
/*
* set mgrid parameters
*
* make sure origin of mgrid's grid is at ionize's grid origin (0,0,0)
*
* length is (number of grid spacings) * (grid spacing),
* where the number of spacings is one less than number of grid points
*/
mg_prm.length = (numpt-1) * gridspacing;
mg_prm.center.x = 0.5 * mg_prm.length;
mg_prm.center.y = 0.5 * mg_prm.length;
mg_prm.center.z = 0.5 * mg_prm.length;
/* make sure domain and grid are both cubic */
if (numpt != numcol || numcol != numplane) {
printf("ERROR: grid must be cubic\n");
return -1;
}
/*
* grid used by mgrid needs spacing h >= 2
*
* determine grid factor: (2^gridfactor)*h_ionize = h_mgrid
*/
gridfactor = 0;
//nspacings = numpt - 1;
nspacings = numpt;
/* add one more spacing so that interpolation loop below will work */
h = gridspacing;
while (h < 2.0) {
h *= 2;
nspacings = ((nspacings & 1) ? nspacings/2 + 1 : nspacings/2);
gridfactor++;
}
mg_prm.nspacings = nspacings;
/* have to modify mgrid length */
mg_prm.length += h;
mg_prm.center.x = 0.5 * mg_prm.length;
mg_prm.center.y = 0.5 * mg_prm.length;
mg_prm.center.z = 0.5 * mg_prm.length;
h_1 = 1.0/h;
mg_prm.cutoff = CUTOFF;
mg_prm.boundary = MGRID_NONPERIODIC;
/* choice of splitting must be consistent with short-range below */
#if defined(CUBIC_TAYLOR2)
mg_prm.approx = MGRID_CUBIC;
mg_prm.split = MGRID_TAYLOR2;
#elif defined(QUINTIC1_TAYLOR3)
mg_prm.approx = MGRID_QUINTIC1;
mg_prm.split = MGRID_TAYLOR3;
#elif defined(HEPTIC1_TAYLOR4)
mg_prm.approx = MGRID_HEPTIC1;
mg_prm.split = MGRID_TAYLOR4;
/*
#elif defined(HERMITE_TAYLOR3)
mg_prm.approx = MGRID_HERMITE;
mg_prm.split = MGRID_TAYLOR3;
*/
#endif
mg_prm.natoms = natoms;
printf("natom = %d\n", mg_prm.natoms);
printf("mgrid center = %g %g %g\n",
mg_prm.center.x, mg_prm.center.y, mg_prm.center.z);
printf("mgrid length = %g\n", mg_prm.length);
printf("mgrid nspacings = %d\n", mg_prm.nspacings);
/* setup mgrid system */
mg_sys.f_elec = (MD_Dvec *) calloc(mg_prm.natoms, sizeof(MD_Dvec));
mg_sys.pos = (MD_Dvec *) calloc(mg_prm.natoms, sizeof(MD_Dvec));
mg_sys.charge = (double *) calloc(mg_prm.natoms, sizeof(double));
for (i = 0; i < mg_prm.natoms; i++) {
mg_sys.pos[i].x = atoms[4*i ];
mg_sys.pos[i].y = atoms[4*i + 1];
mg_sys.pos[i].z = atoms[4*i + 2];
mg_sys.charge[i] = atoms[4*i + 3];
}
/* setup mgrid solver and compute */
if (mgrid_param_config(&mg_prm)) {
printf("ERROR: mgrid_param_config() failed\n");
return -1;
}
printf("spacing = %g\n", mg_prm.spacing);
if (mgrid_init(&mg)) {
printf("ERROR: mgrid_init() failed\n");
return -1;
}
if (mgrid_setup(&mg, &mg_sys, &mg_prm)) {
printf("ERROR: mgrid_setup() failed\n");
return -1;
}
timer = rt_timer_create();
rt_timer_start(timer);
if (mgrid_force(&mg, &mg_sys)) {
printf("ERROR: mgrid_force() failed\n");
return -1;
}
/* DH: end setup and compute long-range */
printf(" using %d processors\n", numprocs);
/* allocate array of threads */
threads = (rt_thread_t *) calloc(numprocs * sizeof(rt_thread_t), 1);
/* allocate and initialize array of thread parameters */
parms = (enthrparms *) malloc(numprocs * sizeof(enthrparms));
for (i=0; i<numprocs; i++) {
parms[i].threadid = i;
parms[i].threadcount = numprocs;
parms[i].atoms = atoms;
parms[i].grideners = grideners;
parms[i].numplane = numplane;
parms[i].numcol = numcol;
parms[i].numpt = numpt;
parms[i].natoms = natoms;
parms[i].gridspacing = gridspacing;
parms[i].excludepos = excludepos;
}
#if defined(THR)
/* spawn child threads to do the work */
for (i=0; i<numprocs; i++) {
rt_thread_create(&threads[i], energythread, &parms[i]);
}
/* join the threads after work is done */
for (i=0; i<numprocs; i++) {
rt_thread_join(threads[i], NULL);
}
#else
/* single thread does all of the work */
energythread((void *) &parms[0]);
#endif
/* DH: tabulate and cleanup long-range */
/* interpolate from mgrid potential lattice */
eh = (double *)(mg.egrid[0].data); /* mgrid's long-range potential lattice */
ndim = mg.egrid[0].ni; /* number of points in each dimension of lattice */
for (kk = 0; kk < numplane; kk++) {
for (jj = 0; jj < numcol; jj++) {
for (ii = 0; ii < numpt; ii++) {
/* distance between atom and corner measured in grid points */
dx_h = (ii*gridspacing) * h_1;
dy_h = (jj*gridspacing) * h_1;
dz_h = (kk*gridspacing) * h_1;
/* find closest mgrid lattice point less than or equal to */
im = ii >> gridfactor;
jm = jj >> gridfactor;
km = kk >> gridfactor;
#if defined(CUBIC_TAYLOR2)
ilo = im-1;
jlo = jm-1;
klo = km-1;
/* find t for x dimension and compute xphi */
t = dx_h - ilo;
xphi[0] = 0.5 * (1 - t) * (2 - t) * (2 - t);
t--;
xphi[1] = (1 - t) * (1 + t - 1.5 * t * t);
t--;
xphi[2] = (1 + t) * (1 - t - 1.5 * t * t);
t--;
xphi[3] = 0.5 * (1 + t) * (2 + t) * (2 + t);
/* find t for y dimension and compute yphi */
t = dy_h - jlo;
yphi[0] = 0.5 * (1 - t) * (2 - t) * (2 - t);
t--;
yphi[1] = (1 - t) * (1 + t - 1.5 * t * t);
t--;
yphi[2] = (1 + t) * (1 - t - 1.5 * t * t);
t--;
yphi[3] = 0.5 * (1 + t) * (2 + t) * (2 + t);
/* find t for z dimension and compute zphi */
t = dz_h - klo;
zphi[0] = 0.5 * (1 - t) * (2 - t) * (2 - t);
t--;
zphi[1] = (1 - t) * (1 + t - 1.5 * t * t);
t--;
zphi[2] = (1 + t) * (1 - t - 1.5 * t * t);
t--;
zphi[3] = 0.5 * (1 + t) * (2 + t) * (2 + t);
/* determine 64=4*4*4 eh grid stencil contribution to potential */
en = 0;
for (k = 0; k < 4; k++) {
koff = (k + klo) * ndim;
for (j = 0; j < 4; j++) {
jkoff = (koff + (j + jlo)) * ndim;
c = yphi[j] * zphi[k];
for (i = 0; i < 4; i++) {
index = jkoff + (i + ilo);
/*
ASSERT(&eh[index]
== mgrid_lattice_elem(&(mg.egrid[0]), i+ilo, j+jlo, k+klo));
*/
if (&eh[index] !=
mgrid_lattice_elem(&(mg.egrid[0]), i+ilo, j+jlo, k+klo)) {
printf("ndim=%d index=%d i+ilo=%d j+jlo=%d k+klo=%d\n"
"ia=%d ib=%d ni=%d\n"
"ja=%d jb=%d nj=%d\n"
"ka=%d kb=%d nk=%d\n",
ndim, index, i+ilo, j+jlo, k+klo,
mg.egrid[0].ia, mg.egrid[0].ib, mg.egrid[0].ni,
mg.egrid[0].ja, mg.egrid[0].jb, mg.egrid[0].nj,
mg.egrid[0].ka, mg.egrid[0].kb, mg.egrid[0].nk
);
abort();
}
en += eh[index] * xphi[i] * c;
}
}
}
/* end CUBIC */
#endif
//ENERGY(grid->eners[k*numcol*numpt + j*numpt + i]);
grideners[kk*numcol*numpt + jj*numpt + ii] += (float)en;
// (float) *((double *)mgrid_lattice_elem(lattice, i, j, k));
//ENERGY((float) *((double *)mgrid_lattice_elem(lattice, i, j, k)));
//ENERGY(grid->eners[k*numcol*numpt + j*numpt + i]*560.47254);
}
}
}
totaltime = rt_timer_timenow(timer);
printf("total time for mgrid: %.1f\n", totaltime);
rt_timer_destroy(timer);
/* cleanup mgrid */
mgrid_done(&mg);
free(mg_sys.f_elec);
free(mg_sys.pos);
free(mg_sys.charge);
/* DH: tabulate and cleanup long-range */
/* free thread parms */
free(parms);
free(threads);
return 0;
}
