>>>>> "Richard" == Richard Fateman <fateman@cs.berkeley.edu> writes:
Richard> Partly to be argumentative, but maybe it will illuminate
Richard> the portability philosophy... Do we guarantee that sin/cos/exp/log
Richard> are all the same? Or even decimal to binary conversion?
Of course not, but if someone says (sin pi) is 7, we beat them up to
fix it. If they're more than a few bits off, we complain to them to
fix it. Branch cuts are another problem, and I'm not even sure the
the Lisp branch cuts are the same as maxima's branch cuts for bigfloat
functions.
Richard> I would personally be very disappointed with a random() function
Richard> that was, say, 100 times slower than the native one. And that
Richard> might be the case. If someone wants a machine independent random,
Richard> provide one in the share library.
Richard> :)
Probably the best idea. I think the only lisp that might be hit hard
would be clisp. How hard, I don't know.
Ray