magistraleinformaticanetworking:spm:exercise1617spm

Write e program that generates two N by N matrixes A and B and computes C, the result of the multiplication of A by B. The computation has to be performed by M threads, each working on a partition of C. Check the correctness of the result against result computed sequentially. Plot scalability curve for the program.

- use
int atoi(char *)

to convert the command line arguments to integers

int main(int argc, char * argv[]) { int n = atoi(argv[1]); int m = atoi(argv[2]); ... }

- use the
std::chrono

library to get time before and after the parallel computation and to compute the elapsed time:

std::chrono::time_point<std::chrono::system_clock> start, end; start = std::chrono::system_clock::now(); ... end = std::chrono::system_clock::now(); std::chrono::duration<double> elapsed_seconds = end-start;

- use plain
rand

to generate random numbers in the two matrixes:

srand(123); for(....) { ... = ((float) (rand() % MAX)) / ((float) RAND_MAX);

Write a program that generates an N by N matrix A and then executed a number Niter of iterations where at each iteration A[i][j] is substituted by the average of A[i][j] and the four neighbours (north, south, east and west). At the borders, the missing elements in the neighbourhood have to be assumed to be 0. All the new A[i][j] values must be computed summing up values **at the current iteration** for the neighbours.
Computation of the single iteration should be parallelized using threads.

magistraleinformaticanetworking/spm/exercise1617spm.txt · Ultima modifica: 12/10/2016 alle 15:36 (7 anni fa) da Marco Danelutto