On Mon, Mar 17, 2014 at 5:12 PM, Jason Moore <[email protected]> wrote: > I have the basic understanding of floating point arthmetic and I've read > that essay at some point in my career. > > As to whether we output garbage with these methods, we typically look to > well defined benchmark problems. For example, the bicycle problem that runs > as a test in the sympy test suite checks the accuracy of the evaluation to > something like 14 decimal places against an independently derived system > from a number of other methods. That problem, in our sympy case, has tons of > minus's in the symbolic code but it gives the same answer as 10+ other > different implementations. Now if we didn't get the same answer, then we'd > need to be looking into things like floating point arithmetic. > > So, if we ever have a problem that gives dubious results, then we will > surely look into floating point arithmetic as a potential cause. But so far, > all of the problems we've tried to solve with this don't give garbage, or at > least don't give errors with respect to other's implementations.
Very cool, that's great news. Ondrej -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/sympy. To view this discussion on the web visit https://groups.google.com/d/msgid/sympy/CADDwiVDW0z0AgABqKp_3FW5XTY8C7pk2xtWxoL%2Bv1uOKyVn15Q%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
