On 11/6/06, Christophe Raffalli <christophe.raffalli@univ-savoie.fr> wrote:
>
> Hi,
>
> Has anyone tried to patch OCaml to use long double (128 bits on intel/gcc) instead of double (64
> bits on intel gcc) ?

In reality they are not 128 bits, but just 80 bits long. The rest is
just to improve memory alignment.

I am not very authoritative to talk about patching ocaml, but judging
from the amount
of internal optimizations for handling of unboxed floating-point
numbers, changing it will be quite difficult.

--
Best regards,
 Alexander.