posted on 1993-01-01, 00:00authored byPhilip Koopman
While the cache memory designed into advanced processors can significantly speed up the average performance of many programs, it also causes performance varations that surprise system designers and cause problems during product integration and deployment. This paper gives a description of cache memory behavior for real-time embedded systems, using the example of real data collected from an 80486 CPU interrupt service routine. While vendor-supported tools for predicting and bounding worst-case CPU delays are still in their infancy, there are some coping strategies that will reduce problems while minimizing risk.