On Tue, Nov 17, 2009 at 10:23 AM, Phil Endecott <spam_from_boost_dev@chezphil.org> wrote:
Do please have a look at the paper that I linked to before, and in particular the figure at the top of the second page:
http://www.mpi-inf.mpg.de/~kettner/pub/nonrobust_cgta_06.pdf
I have read this, and several other papers. I've even taken a few numerical analysis classes at the university.
1. Failures may not occur only in cases when things are very close. A problematic input may cause the output to be grossly wrong.
Lady luck has smiled on me thus far. ;-) And a small amount of data pre-processing goes a *long* way. Know your algorithms, and their pre/post-conditions.
2. You may not get a wrong answer; the program might instead segfault or loop forever.
This is not a FP-induced instability problem. This is a bug in your algorithm. Coincidentally, I just fixed a loop-forever bug in one of our geometry algorithms the other day. Jon