[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

CMUCL floats

I really appreciate your response way back when about float functions.
I really hope that you do implement that extension that you mentioned.
As far as PCL goes, I have implemented my own version of CLOS
(starting from the one in the MOP book) that is much more efficient
than PCL.  This would allow me to make the optimizations allowing for
funcalling float arg methods.  

The solution to all of these problems is to have a 64-bit machine.
The 64-bits would be divided into 32-bit data and 32-bit tag.  The tag
would actually be a class pointer (ala CLOS).  This would speed up all
the existing immediates and allow single-float immediates.  Plus it
allows all data-structures to be objects.  I'm a super CLOS fanatic
and think CommonLisp should be integrated into CLOS.  I truly think
that CLOS could be implemented efficiently enough to make this
feasible.  For instance, certain methods could be inlined if the
proper type information was present (making floating point and fixnum
math efficient).  What do you think?  Am I deluding myself?

By the way, I am super pumped up about CMUCL and want to contribute to
the effort!  I think it is fantastic and am very much behind it.  This
summer I plan to spend a large amount of time working on it.  Have you
set up any mechanism for contributed code?

-- jonathan