We present the memory size, computational time, and technique aspects of de
nsity matrix renormalization group (DMRG) algorithm. We show how to estimat
e the memory size and computational time before starting a large scale DMRG
calculation. We propose an implementation of the Hamiltonian wavefunction
multiplication and a wavefunction initialization in DMRG with block matrix
data structure. One-dimensional Heisenberg model is used to illustrate our
study.