Gustafson claims on his slide 2 that 64-bit IEEE-754 gives the wrong answer.
Yet, naive execution of the algorithm in APL, J, and C (gcc) all produce the correct answer using 64-bit reals. So, is he wrong? His penultimate slide mentions 80-bit precision, which is what X86 boxes use for IEEE-754 math internally, only converting back to 64-bit words when you have to store a result. The above systems use 80-bit precision for intermediate computations, so they get the correct result with 64-bit reals. Therefore, Gustafson's claim is correct, but I think it is somewhat misleading. Bob On 2017-02-12 06:26 PM, Raul Miller wrote:
Mmm... require it where, though? Thanks,
-- Robert Bernecky Snake Island Research Inc 18 Fifth Street Ward's Island Toronto, Ontario M5J 2B9 [email protected] tel: +1 416 203 0854 ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
