next up previous
Next: Metric Assumption 2: Do Up: Interpretation of Results and Previous: Planner Assumption 3: Are

Metric Assumption 1: Do Alternative Platforms Lead to Different Performance?

In our experiments, performance did not vary as much as we expected. This result suggests that researchers in general are not developing for specific hardware/software configurations, but recent trends suggest otherwise, at least with regards to memory. Again, because these systems are research prototypes, it behooves the developer to be clear about his/her expectations and anyone subsequently using the system to accommodate those requests in their studies.
Recommendation 13: As with other factors in planner design, researchers must clearly state the hardware/software requirements for their planners, if the design is based on platform assumptions. Additionally, a careful study of memory versus time trade-offs should be undertaken, given the recent trends in memory exploitation.


©2002 AI Access Foundation and Morgan Kaufmann Publishers. All rights reserved.