Re: Further 64 bit bits.

[Dirk Herrmann]

On 14 Sep 2001, Marius Vollmer wrote:

> Rob Browning <address@hidden> writes:
> 
> > With respect to make check, is there any way to ask how many bits are
> > in a fixnum from the guile level?
> 
> This
> 
>     (inexact->exact (+ (/ (log (+ most-positive-fixnum 1)) (log 2)) 1))
> 
> ought to work.

I wonder if from the scheme level it is wise to expect fixnums to be
implemented this way, i. e. as a number that is represented with n
bits.  For example, an implementation of fixnums might not use the full
range of numbers [2^n - 1 ; -2^n], but might instead use certain numbers
within that range to indicate 'special cases' like positive infinity or
something.

Best regards,
Dirk Herrmann