A PERFORMANCE AND COST-ANALYSIS OF APPLYING SUPERSCALAR METHOD TO MAINFRAME COMPUTERS

This paper presents the results of evaluating the increased performanc e and cost of mainframe computers with superscalar architectures. Sinc e mainframe users demand object compatibility, we assume the same obje ct as that of nonsuperscalar machines. We compared four differently co nfigured superscalar machines based on Hitachi's high-end mainframe co mputer, the HITAC M-880, varying the multiplicity of operand accessibi lity and arithmetic capability. In estimating performance, we consider ed the effect of critical path delay on machine cycle time. For scient ific jobs, either dual operand accessibility or dual arithmetic capabi lity, or both, improved performance (MIPS) by 11-28% while increasing CPU hardware cost by 2-21%. For online transaction processing (OLTP), no configuration increased performance more than 4%. To make the super scalar architecture more effective for OLTP, it is important to reduce execution cycles per instruction (CPI), by reducing overhead caused b y sequential processes.