In this work, we seek to close the performance gap between exact inference in discrete graphical models and discrete-valued finite-domain probabilistic programs. The key idea behind existing state-of-the-art inference procedures in discrete graphical models is to compile the graphical model into a representation known as a weighted Boolean formula (WBF), which is a symbolic representation of the joint probability distribution over the graphical model’s random variables. This symbolic representation exposes key structural elements of the distribution, such as independences between random variables. Then, inference is performed via a weighted sum of the models of the WBF, a process known as weighted model counting (WMC). This WMC process exploits the independences present in the WBF, and is thus efficient. Inference via WMC is currently the state-of-the-art exact inference strategy for discrete graphical models, probabilistic logic programming languages, and probabilistic databases.