"John Randall" <[EMAIL PROTECTED]> wrote: > Dan is right (although one can speculate why it goes wrong for 10000 to > 19999).
J seems to store extended integers internally in base-10000 form (as an array of integers, each of which stores 4 decimal digits). This makes it easy for decimal display output, and also lets arithmetic arithmetic to be done without overflow (even when multiplying chunks). Perhaps it is estimating how many 4-digit chunks the result will take, and gets the result wrong in the range 10000^n ... <:2*10000^n. (I suspect that 1=1^1r2 also exhibits this problem, but nobody ever noticed it because the result just happens to also be correct). This problem seems independent of the denominator of the power, while the (erroneous) result is then raised to the numerator: 10000^3r91 1000000000000 -- Mark D. Niemiec <[EMAIL PROTECTED]> ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
