Empirical Study of Normalized Timing Requirements for Various Standard Operations

摘要

Simulations are notorious for being among the most expensive of all computational tasks. In many engineering, military and scientific computer installations processing time is devoted almost exclusively to simulation. The execution of one simulation can take many hours of computer time. When the simulation model is probablistic or if sensitivity analysis is necessary, many replications may be needed to accurately determine the results. Substantive cost improvements can be achieved by designing simulators with efficiency of operation emphasized from the outset. This report contains normalized timing data that can help programmers make design tradeoffs which will lead to more efficient code. (ERA citation 12:044796)