Re: [Maxima-discuss] bfloat accuracy, %pi

[Raymond T. <toy...@gm...>]

>>>>> "Richard" == Richard Fateman <fa...@be...> writes:

    Richard> I think Stavros' example may be related to a change in the parser!

Don't make us hold our breath!  What makes you think it's the parser?

    Richard> I responded to him privately; I think the bfloat problem is not
    Richard> changed depending on version;
    Richard> also the formula for relative error in the makelist is
    Richard> not right. There is a big than expected error near
    Richard> special points like n*%pi/2.

I do remember fpsin tried hard to add more bits to %pi so that the
result would be accurate to fpprec digits.  But that bit of code seems
unchanged in a long time. (I think; I didn't actually check.)

--
Ray