/**
Handle the user-induced transformation from \a initial_selection_rect to \a new_selection_rect
*/
void PartEllipse::handleUserTransformation(const QRectF &initial_selection_rect, const QRectF &new_selection_rect) {
QList<QPointF> mapped_points = mapPoints(initial_selection_rect, new_selection_rect, saved_points_);
setRect(QRectF(mapFromScene(mapped_points.at(0)), mapFromScene(mapped_points.at(1))));
}
/**
Handle the user-induced transformation from \a initial_selection_rect to \a new_selection_rect
*/
void PartTerminal::handleUserTransformation(const QRectF &initial_selection_rect, const QRectF &new_selection_rect) {
QPointF mapped_point = mapPoints(initial_selection_rect, new_selection_rect, QList<QPointF>() << saved_position_).first();
setPos(mapped_point);
}
void derivs(int n,double time,double * x, double * der){
extern int max(int,int);
// For ease of use, create cart3d variables, associate them
// with locations in the given arrays, and write equations in
// terms of points rather than x and der.
cart3d pos;
cart3d vel;
cart3d xder;
cart3d vder;
mapPoints(n,3,x,der,&pos,&vel,&xder,&vder);
// BEGIN DEFINING ACCELERATIONS HERE, FOOLOWED BY DEFINITION
// OF XDER AS CURRENT VELOCITY.
double rad2,rad,diff,diffx,diffy,diffz,dcon;
int i;
double * shield_rad;
// precondition problem before force calculation,
// set up sheild radii, build trees, meshes.
shield_rad = new double[g_dynamic.npoints+1];
for(i=0;i<g_dynamic.npoints;i++) {
shield_rad[i]=g_dynamic.comp_s_rad((g_dynamic.gnorm*
g_dynamic.mass[i]),g_dynamic.time_step);
vder.x[i]=0.0;
vder.y[i]=0.0;
vder.z[i]=0.0;
}
//For MPI Version, need to pass once to each processor
// - g_dynamic.npoints
// - g_dynamic.gnorm
// - g_dynamic.mass
// - shield_rad
// - x
MPI_Request * message_request = new MPI_Request[g_mpi.size];
MPI_Request * number_request = new MPI_Request[g_mpi.size];
MPI_Request * numberper_request = new MPI_Request[g_mpi.size];
MPI_Request * gnorm_request = new MPI_Request[g_mpi.size];
MPI_Request * mass_request = new MPI_Request[g_mpi.size];
MPI_Request * srad_request = new MPI_Request[g_mpi.size];
MPI_Request * x_request = new MPI_Request[g_mpi.size];
// decide how to split up problem
int * number_per_cpu = new int[g_mpi.size];
int block = (g_dynamic.npoints)/(g_mpi.size);
for (int i=0;i<g_mpi.size;i++) {
number_per_cpu[i] = block;
if (i==g_mpi.size-1) {
//last client gets overflow
number_per_cpu[i]+=g_dynamic.npoints-g_mpi.size*block;
}
}
// tell clients to get ready for data and send it
for (int i=1;i<g_mpi.size;i++) {
MPI_Isend(0,0,MPI_INT,i,MPIDATA_DODERIVS,
MPI_COMM_WORLD,&message_request[i]);
MPI_Isend(&g_dynamic.npoints,1,MPI_INT,i,MPIDATA_PASSNUMBER,
MPI_COMM_WORLD,&number_request[i]);
MPI_Isend(&number_per_cpu[i],1,MPI_INT,i,MPIDATA_PASSNUMBERPER,
MPI_COMM_WORLD,&numberper_request[i]);
MPI_Isend(&g_dynamic.gnorm,1,MPI_DOUBLE,i,MPIDATA_PASSGNORM,
MPI_COMM_WORLD,&gnorm_request[i]);
MPI_Isend(g_dynamic.mass,g_dynamic.npoints,MPI_DOUBLE,i,
MPIDATA_PASSMASS,MPI_COMM_WORLD,&mass_request[i]);
MPI_Isend(shield_rad,g_dynamic.npoints,MPI_DOUBLE,i,
MPIDATA_PASSSHIELD,MPI_COMM_WORLD,&srad_request[i]);
MPI_Isend(x,g_dynamic.npoints*6,MPI_DOUBLE,i,
MPIDATA_PASSX,MPI_COMM_WORLD,&x_request[i]);
// have it wait here for a message that the work has been
// completed.
}
// determine number of processes, decide
// on which processes will solve which
// parts of problem
// For each body, compute accelerations.
for(i=0;i<number_per_cpu[0];i++){
for(int j=0;j<g_dynamic.npoints;j++){
if (i!=j) {
rad2=pow((pos.x[i]-pos.x[j]),2)+
pow((pos.y[i]-pos.y[j]),2)+
pow((pos.z[i]-pos.z[j]),2);
rad=sqrt(rad2);
dcon=g_dynamic.gnorm/(rad*rad2);
diffx=(pos.x[j]-pos.x[i])*dcon;
diffy=(pos.y[j]-pos.y[i])*dcon;
diffz=(pos.z[j]-pos.z[i])*dcon;
if (rad>shield_rad[j]) {
vder.x[i]+=diffx*g_dynamic.mass[j];
vder.y[i]+=diffy*g_dynamic.mass[j];
vder.z[i]+=diffz*g_dynamic.mass[j];
}
// if (rad>shield_rad[i]) {
// vder.x[j]-=diffx*g_dynamic.mass[i];
// vder.y[j]-=diffy*g_dynamic.mass[i];
// vder.z[j]-=diffz*g_dynamic.mass[i];
// }
}
}
xder.x[i]=vel.x[i];
xder.y[i]=vel.y[i];
xder.z[i]=vel.z[i];
}
for (int i=1;i<g_mpi.size;i++) {
//wait for processes to complete sending
MPI_Wait(&message_request[i], &g_mpi.status );
MPI_Wait(&number_request[i], &g_mpi.status );
MPI_Wait(&numberper_request[i], &g_mpi.status );
MPI_Wait(&gnorm_request[i], &g_mpi.status );
MPI_Wait(&mass_request[i], &g_mpi.status );
MPI_Wait(&srad_request[i], &g_mpi.status );
MPI_Wait(&x_request[i], &g_mpi.status );
}
for (int i=1;i<g_mpi.size;i++) {
// get replies
double * retval = new double[number_per_cpu[i]*6];
MPI_Recv(retval,number_per_cpu[i]*6,
MPI_DOUBLE,i,MPIDATA_DONEDERIVS,MPI_COMM_WORLD,
&g_mpi.status);
//compare replies
for (int ireturn=0;ireturn<number_per_cpu[i];ireturn++) {
int ireal = block*i + ireturn;
xder.x[ireal]=retval[ireturn];
xder.y[ireal]=retval[number_per_cpu[i]+ireturn];
xder.z[ireal]=retval[2*number_per_cpu[i]+ireturn];
vder.x[ireal]=retval[3*number_per_cpu[i]+ireturn];
vder.y[ireal]=retval[4*number_per_cpu[i]+ireturn];
vder.z[ireal]=retval[5*number_per_cpu[i]+ireturn];
}
delete retval;
}
// Clean up dynamically allocated arrays
delete shield_rad;
delete message_request;
delete number_request;
delete numberper_request;
delete gnorm_request;
delete mass_request;
delete srad_request;
delete x_request;
delete number_per_cpu;
}