Fast Makespan Estimation for GPU Threads on a Single Streaming Multiprocessor
Ref: CISTER-TR-130406 Publication Date: 14, Apr, 2013
Fast Makespan Estimation for GPU Threads on a Single Streaming MultiprocessorRef: CISTER-TR-130406 Publication Date: 14, Apr, 2013
Graphics Processing Units (GPUs) are widely used to unload the CPUs, liberate other resources of a given computer system, and provide an alternative to multiprocessor computers as a means of processing computationally expensive parallel tasks. The recent trend of utilizing GPUs in embedded systems necessitates the development of timing analysis techniques for finding the joint worst-case execution time for a group of GPU threads of the same parallel application, on a streaming multiprocessor. The state-of-the-art approaches for computing the exact maximum makespan of GPU threads running on a single streaming multiprocessor are intractable and even pessimistic approximations usually take a long time to complete. We therefore develop a technique for finding an estimate of the maximum makespan using metaheuristics. Its simplicity, flexibility and ability for massive parallelization, determine a potential of usage for soft real-time systems.