212
edits
Changes
→Solution to first Kernel problem
The problem was resulting because of this calculation, '''u[m * nx + it] = un[m * nx + it - 1] - c*dt / dx*(un[m * nx + it - 1] - un[(m - 1) * nx + it - 1]);''' Perhaps using the c, dt & dx values incorporated to many trips to global memory which caused a hang in the operation and CUDA automatically crashed the driver. To solve this problem a scalar variable (local variable) was created to store this value in registered memory for each thread to access. '''double total = c*dt / dx;'''... Now the program executes with an argument of 2000 2000 and yeilds similar results to the original program. Unfortunately a new problem has risen, when the argument is raised above 2000 the program once again crashes and I am stuck with no solution currently...
''' __global__ void Calculate (double* u, double* un,int nx, int c, double dx, double dt) { double total = c*dt / dx; for (int it = 1; it <= nx - 1; it++) { for (int k = 0; k <= nx - 1; k++) { un[k * nx + it - 1] = u[k * nx + it - 1]; } for (int m = 1; m <= nx - 1; m++) { u[0 * nx + it] = un[1 * nx + it - 1]; u[m * nx + it] = un[m * nx + it - 1] - total * ( un[m * nx + it - 1] - un[(m - 1) * nx + it - 1] ); } }
}'''
