Optimising Speed 1 (time)

• Examine time/space tradeoffs

• Chose: throughput / CPU time / wall-clock time?

  • Have you CPU to burn?

• Chose: average or worst-case values?

  • Randomised algorithms and hash tables vs brute-force

• Implement bounded work and data queues to avoid overflows and sluggish recovery from activity spikes

Previous slide

Next slide

Back to first slide

View graphic version

EXAMPLE: Web browser caching to eliminate latency and bandwidth delays; memory functions and (caching) lazy-evaluation mechanisms.

Bounded Queues of work: (eg collapse and discard pointer/mount events) so that you don’t queue up lots of (possibly redundant) work and lose user responsiveness. In many cases only the last value of some input parameter is important. To see how not to do this, look at Excel!

CHECKLIST:

• Do you know what takes most time to re-fetch or re-compute? Can you cache it?