Performance Estimates
Complexity analysis = theoretical performance estimation.
It assists us to choose the most appropriate algorithm.
We can also consider an experimental approach to performance:
- determine the critical (frequent/expensive) operations in the system
- estimate the likely number of occurrences on classes of data
- combine the two to produce a cost estimate
E.g. a program reads 100 disk pages and performs 100,000 adds per page.
Cost = 100(Cost(disk) + 105.Cost(add)) = 102(10-1s +105.10-6s)
Need to know relative costs of various operations.
|