high performance computing on graphics processing units: hgpu.org

In this paper, we present an efficient diagnostic fault simulator based on a state-of-the-art graphics processing unit (GPU). Diagnostic fault simulation plays an important role to identify and locate the causes of circuit failures. However, today’s complex VLSI circuits pose ever higher computational demand for such simulators. Our GPU based diagnostic fault simulator (GDSim) is based on a novel two-stage simulation framework which exploits high computation efficiency on the GPU. The fault pair based simulation is proposed to overcome the limited capacity of GPU memory as well as achieve a substantial fine-grained parallelism. Multi-fault-signature and dynamic load balancing techniques are introduced for the best usage of computing resources on-board. Experimental results demonstrate a speedup of up to 121x (with average speedup of 38.43x) compared to a state-of-the-art CPU-based diagnostic fault simulator.