skaller wrote: > Any thoughts on how to add 'costing' information to Felix? > Chris King wrote: > but who chooses the constants?
This last point on the constants makes a big difference. I will come back to that. First, one has to decide what is really being 'costed'. Worst-case or average-case? Neither are fully satisfactory, for obvious reasons. How should costing be done? Well, there the answer is fairly well-understood now: series. Or more precisely, and asymptotic series. But constants matter a lot. For example, consider two algorithms which have cost n^3 and 10^12*n^2 respectively. The n^3 algorithm is to be chosen over the O(n^2) algorithm essentially *always*, as the cross-over is at n=10^12, which is likely bigger than what can realistically be done. These are extreme cases, but I have seen more subtle instances where the "fast" algorithm was never useful. Note that there are times when it is sub-dominant terms which allow you to choose between 2 algorithms. I guess what I am saying is that whatever you implement, the above considerations should be used in the *design* (ie design-for-change), as if Felix is successful, whatever first approximation of costing will need to evolve towards handling the above issues. Jacques ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Felix-language mailing list Felix-language@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/felix-language
