A performance comparison of the IBM RS/6000 and the Astronautics ZS-1

A performance evaluation method for comparing and improving concurrent uniprocessor architectures is introduced, and a detailed case study of its application to two machines is reported. The architectures of the machines, the IBM RS/6000 and the Astronautics ZS-1, are described. The measured performance is given, and a model that yields hard bounds on a machine's performance on a benchmark is presented. The sources of performance loss, i.e. the difference between the model-derived bounds and the measured performance, are examined, and areas for improving the architectures are determined.<
>