From: Dr. David Kirkby <david.kirkby@on...>  20090810 07:42:59

You probably see my previous posts on the fact that elliptic_e is giving incorrect results on SPARC hardware using ECL. If not, http://sagetrac.org/sage_trac/ticket/6716 has the details. Can anyone who knows the internals of Maxima tell me what standard libraries supplied with Solaris the functions 'elliptic_e' might be expected to call? Looking in a maths book, I see some formula with sin() in them, suggesting an incorrect result from sin(x) might mess the computation of an elliptic integral. A few quick checks with sin() on Sage showed no obvious differences from those computed in Mathematica. Is there anything else that could be suspected, and so narrow down the list of possible causes somewhat? Sage can certainly be eliminated, as the problem is seen when I call Maxima directly. Things that can't be totally eliminated, but some are very unlikly, include: * Hardware fault on the particular machine. * Design flaw in Sun UltraSPARC CPU (remember the Pentium floating point bug, which must have cost Intel a fortune?) * ECL * Solaris library bug * Maxima * gcc * I screwed up compiling Maxima or ECL. I hope to eliminate the possibility of a hardware fault by copying the binaries to another Sun. I believe ECL will build with Sun's compiler, so I can try to eliminate gcc as the culprit. Can anyone with a better maths background suggest other ways the possible range of issues could be narrowed down? 