Content area
Abstract
Combinatorial problems are pervasive in our lives. From solving a Sudoku puzzle to quickly delivering Amazon orders, all these tasks requires finding a good combinations of elements, locations, and more. The ability to effectively solve combinatorial problems is essential, as it minimizes resource waste, prevents service delays, and reduces costs. For decades, Constraint Programming (CP) has been successfully employed to tackle combinatorial problems. However, as these problems grow larger and more complex each year, constraint solvers must continuously advance their capabilities to keep pace. Graphics Processing Units (GPUs) accelerate costly computations and have been successfully used across various domains. Their unprecedented power and accessibility prompt a (re)consideration of their use in CP. This thesis identifies and develops novel approaches to effectively and transparently harness the power of GPUs to advance CP. It introduces a shift in perspective that enables the achievement of meaningful results and paves the way for addressing increasingly complex and large-scale problems.
