Approximation algorithms for the m-dimensional 0–1 knapsack problem: Worst-case and probabilistic analyses