A method for an integrated simulation linked with scheduling policies on a program-driven simulator

The performance of shared memory multiprocessor systems is dependent on the system software parameters such as the processor scheduling policies as well as on the hardware system parameters. Most existing simulators, however, lack the capability of integrated simulation of both hardware and software because they can run only a single benchmark at one time. We propose a new method on a program-driven simulator basis, which enables the concurrent executions of multiple parallel workloads, thus leads to support the modeling of various scheduling policies on shared memory multiprocessor systems. By using the proposed method, a user may measure and analyze the performance of the system under development more approximately to the real run-time environments