CSP II—a universal computer architecture simulation system for performance evaluation

• Authors: Harold Y. Iwata;Melvin M. Cutler
• Venue: ANSS '75 Proceedings of the 3rd symposium on Simulation of computer systems
• Year: 1975

CSP II is a system through which computers can be quickly and flexibly simulated at the functional level, combining the accuracy of a detailed simulation with the simplicity of a high-level simulation. The simulation experiences which motivated the authors' design of CSP II are described, as are the objectives that CSP II was designed to achieve. Strongest of the latter are the capabilities to vary the workload, instruction set, architecture, and timing without recoding the simulation program. Since simulation languages and systems typically offer complete timing flexibility and some flexibility in specifying architecture and workload, emphasis is placed on how CSP II's table-driven approach achieves this total flexibility without significant inefficiency. The array of statistics-gathering mechanisms built into CSP II is described, from a detailed scoreboard of processor state changes to lumped parameters describing computer performance within a simulation run. The great emphasis placed on workload specification capability by the CSP II design is discussed, concluding with the description of the routines written to support workload specification, measurement, and generation. The advantages of the CSP II approach to workload specification are noted, including the ability to generate a workload synthetically. The use of a single workload specification to compare two or more computers, possibly with different instruction sets, is described-a unique capability of the CSP II system.