The modeling of physical mechanisms governing damage and fracture under extreme loading conditions has made remarkable progress in recent decades, driven by various industrial demands. However, numerous scientific questions remain unanswered. Particularly in the context of spallation, discussions persist around the interactions between pores and their spatial distribution. Our study aims to shed new light on these inquiries using large-scale Molecular Dynamics (MD) simulations, encompassing billions of atoms. To achieve this, we have developed an in-situ clustering tool based on connected component labeling, capable of running on multi-billion-particle simulations to detect aggregates or pores. This tool provides the number of pores, their volumes, and positions at each time step. We applied this approach to piston-driven shocks in tantalum, as well as in a model brittle material with adjustable ductility. For each material, we can analyze the statistics of pore volume and spatial repartition. Additionally, we can extract the spall fracture surface from the simulation and analyze the spatial correlations of roughness fluctuations, comparing them to experimental data.The distribution of pore centers, along with the roughness studies, reveals spatial correlations. The length over which these correlations exist is likely indicative of interactions between pores, which are often overlooked in modeling.