In this paper, we consider DAG-based Distributed Ledger Technologies (DLTs), i.e., DLTs where each block can reference several previous blocks hence forming a Directed Acyclic Graph of Blocks (BDAG). Each block has a weight (usually a constant normalized to one) and our goal is to compute the heaviest sub-BDAG that does not contain conflicting blocks. First, we prove that computing such a sub-BDAG is NP-complete. Then, we show that the difficulty comes from concurrent conflicts and we present an optimal algorithm that is polynomial if the number of concurrent conflicts is bounded. Finally, we evaluate the performance of our algorithm on random BDAGs against an existing algorithm called GHOSTDAG and show that, in addition to being optimal, our algorithm is also more efficient in practice.