13.4 Bottom-Up (Results-Driven) Performance Modeling
Bottom-up modeling is based on using known data to predict system performance. To illustrate this approach, consider the small control system shown in Figure 13.13:
Over time, performance data has been collected on individual computing functions, such as the control algorithms. Using information like this makes it possible to predict the total control program timing, as shown in Figure 13.14. This technique can be used effectively in small systems or in subsystems of larger applications. However, the technique is not as foolproof as it may seem; the times actually produced in new designs depend on many factors, including:
- Processor type
- Clock rate
- Memory speeds and use of memory caches
- Location of code in memory
- Compiler efficiency
- Compiler optimization techniques: