Exploring GPGPU workloads: Characterization methodology, analysis and microarchitecture evaluation implications