Re: [Sbcl-bugs] Question about getting NAN and INF bit patterns into SBCL

[Christophe R. <cs...@ca...>]

Steve Stevenson <st...@cl...> writes:

> This is a question rather than a bug ... I just want to have the
> correct bit patterns. Can I turn off the trap and compute it? Put it
> into hex? Really don't care how dirty it is.

[ sbcl-help is probably a better place ]

The dirtiest way is probably to use sb-kernel:make-single-float (and
make-double-float).  The functions aren't maximally logical -- the high
word argument of each of the functions is signed, not unsigned -- but
they construct a float out of 32-bit integer bit patterns in the obvious
way.

Cheers,

Christophe