We propose a new method for accelerating the computation of a concurrency relation, that is all pairs of places in a Petri net that can be marked together. Our approach relies on a state space abstraction, that involves a mix between structural reductions and linear algebra, and a new data-structure that is specifically designed for our task. Our algorithms are implemented in a tool, called Kong, that we test on a large collection of models used during the 2020 edition of the Model Checking Contest. Our experiments show that the approach works well, even when a moderate amount of reductions applies.