How to parallel computing in case the input is a matrix?

Question

I have a dataframe df_simulate containing 1000 rows and 10001 columns. I define a function var_sum that takes a pair of integers (m, n) as