GCL and denormalized numbers



>>>>> "Camm" == Camm Maguire <camm at enhanced.com> writes:

    Camm> Greetings!  Is there a reason that maxima instructs GCL alone to use
    Camm> denormalized numbers in the f2cl stuff?

    Camm> (defun d1mach (i)
    Camm>   (ecase i
    Camm>     (1
    Camm>      #-gcl least-positive-normalized-double-float
    Camm>      #+gcl least-positive-double-float)
    Camm>     (2 most-positive-double-float)
    Camm>     (3 double-float-epsilon)
    Camm>     (4 (scale-float double-float-epsilon 1))
    Camm>     (5 (log (float (float-radix 1d0) 1d0) 10d0))))

Most likely because at the time this was written gcl didn't have
least-positive-normalized-double-float.  My copy of gcl 2.4 doesn't
have that.

    Camm> This will force -mieee-with-inexact on the alpha architecture, which
    Camm> slows the code considerably.

I would think that anyone still using an Alpha chip doesn't really
need speed anymore. :-)

Ray