double *B_a , *B_b ; //matrix blocks
size_t err;
//double alpha = 1, beta = 1; //C := alpha * a * b + beta * c
- size_t B_proc_col, B_proc_row; // Number of bloc(row or col) on one processor
+ size_t B_proc_row; // Number of bloc(row or col) on one processor
+#ifndef CYCLIC
+ size_t B_proc_col;
B_proc_col = k_b / Block_size; // Number of block on one processor
+#endif
B_proc_row = k_a / Block_size; // Number of block on one processor
//size_t lda = k_a, ldb = n, ldc = n;
XBT_DEBUG("position of B_b: %zu \n", pos_b);
}
end_time_intern = MPI_Wtime();
- communication_time += start_time_intern - end_time_intern;
+ communication_time += end_time_intern - start_time_intern;
MPI_Barrier(row_comm);
MPI_Barrier(col_comm);
c[i*ldc+j] += B_a[i*lda_local+k]*B_b[k*ldb_local+j];
end_time_intern = MPI_Wtime();
- computation_time += start_time_intern - end_time_intern;
+ computation_time += end_time_intern - start_time_intern;
}
MPI_Barrier(row_comm);
MPI_Barrier(col_comm);
end_time = MPI_Wtime();
- time = start_time - end_time;
- printf("communication time: %le nanoseconds, "
- "computation time: %le nanoseconds\n",
+ time = end_time - start_time ;
+ printf("communication time: %le seconds, "
+ "computation time: %le seconds\n",
communication_time, computation_time);