Multiplication of multidiagonal matrix by dense matrix - CUDA version

We consider a matrix A where must of the diagonals in the matrix have all the elements zero. Matrix A is multiplied by a dense matrix B to obtain a dense matrix C. All the matrices are square NxN. Matrix A is stored in the same form as matrices B and C, but it is not necessary to consider in the multiplication the zero diagonals of matrix A.


A number of problems is solved. For each problem the function to parallelize has:

Input parameters:

-int N: number of rows and columns of the matrices

-double *A: multidiagonal matrix, stored in dense format

-double *B: dense matrix

Output parameter:

-double C: result matrix, C=A*B

Files

For more instructions: general instructions.