Fault injection on GPGPU application

Today, with the development of GPU computing techniques in terms of architectures and hardware and software support, people realized that intensive computing workload could be ported to GPU device. Applications could exploit GPUs’ characteristics for parallel computing and gain a significantly high speedup comparing to CPU architecture. However, failures are still unavoidable. People have already put a big amount of effort on detecting, simulating, modifying and analyzing these failures and errors for CPU applications during past years. Based on what I have experienced on CPU platform, I am very interested in characterizing the frequency of register errors and memory errors on GPU platform and make a reasonable comparison with CPU. This project is trying to figure out the impact of soft errors to GPGPU applications by injecting faults into GPGPU applications. We use GPGPU-Sim as the simulator to run the test application and we use AES as the test application. In addition, we also provide a CPU version of test to compare the result of fault injection with GPU.