Thank you, that was an interesting error. I have updated the rosettacode page, changing the definition of lcg so that the constants are represented in extended precision.
Thanks again, -- Raul On Fri, Jan 24, 2014 at 1:30 PM, Brian Schott <[email protected]> wrote: > Raul, > > No, that LCG does not do well for the bsd version (the ms version only has > 32,000 numbers it can generate, so it is impractical, but works as needed). > Consider the two examples below, one on J602 and one on J701. Judging by > the how the first few numbers do work, it may by possible to fix by just > adding precision to the algorithm, though. > > Thank you, > > J602: > > > x: rand_bsd 10 > 12345 1406932606 654583808 1358247936 2138638336 1459132416 1445521408 > 370866176 1896597568 1518859008 > > J701: > > rand_bsd 10 > 12345 1406932606 654583775 1449466924 229283573 1109335178 1051550459 > 1293799192 794471793 551188310 > > > > > > > On Fri, Jan 24, 2014 at 9:58 AM, Raul Miller <[email protected]> wrote: > >> Would http://rosettacode.org/wiki/Linear_congruential_generator#J >> satisfy your need for a "portable" "RNG" or do you have deeper needs? >> >> Thanks, >> >> -- >> Raul > > > -- > (B=) > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
