Mike wrote: [snipped] > items that specifically referred to the floating point Error in the first batch of Pentium chips
Sorry, I should have narrowed the search then. This is not specific to that error, it is a consequence of the FP number system representation. It arises in certain circumstances, but ones that are not uncommon. I just wanted to flag that people could be more aware of what really happens at a hardware/number representation level when they want to divide / compare / subtract similar numbers, so they can watch for pitfalls. As far as possible, avoid floating point divides. They're ugly, and they're slow. I spent too long designing alternative representations that avoid this, so its one of my soap boxes when people complain that compares don't work as they expect. > other occurrences of 'catastrophic loss of precision' that can be tested using modern Pentiums Absolutely. I don't have any code to demo this, but will try to find some slacktime to rustle something up. Any of the URIs below will probably give you the idea. > Some URLs would be helpful here. Thanks. Sure OK, here goes: (shout if you need more) http://www.cse.msu.edu/~cse320/Documents/FloatingPoint.pdf http://www.cs.princeton.edu/introcs/91float/ (and scroll down to catastrophic cancellation) http://cs-www.bu.edu/faculty/djy/cs210/unit06.html (see section 'Cancellation Error') http://www.math.uu.se/~warwick/summer04/material/reading/rump2.pdf (if you really want the theory) http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/393090 (from activestate, a Python example) ed c _______________________________________________ Perl-Win32-Users mailing list Perl-Win32-Users@listserv.ActiveState.com To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs
