Using SystemC to Model and Simulate a Many-Core Architecture for LU Decomposition

Designing efficient many-core architectures with hundreds of cores is a very challenging task due to the complexity and size of the design space. System-level simulation can help designers exploring the design space of many-core architectures. In this paper, we use SystemC to model a many-core architecture and run a parallel LU decomposition algorithm, an important linear algebra kernel that is widely used in both scientific and engineering applications. The design is parameterized permitting to adapt the model to various hardware constraints. The simulation of the model outputs the number of communications and the number of clock cycles required to complete the algorithm. Given the obtained simulation times, SystemC can be used to efficiently model many-core architectures. Also, the results show the scalability of the architecture and are according to the theoretical formulations.