Re: [Chicken-users] 32 bit integers?

[Kon Lovett]

On May 22, 2007, at 12:33 PM, Dan Muresan wrote:

> > >
<snip>
> Actually, I haven't found the 31-bit format documented anywhere  
> (maybe I missed it). I thought that at least normal C int's would  
> always fit in a fixnum. It was an unpleasant surprise to see an  
> expression like (logior flags ...) fail, when "flags" was promoted  
> to a flonum.
>
> > If you use one of integer unsigned-integer integer32 unsigned- 
> > integer32 integer64 long unsigned-long as a foreign parameter/ 
> > return type Chicken will automatically convert a flonum integer to/ 
> > from the native type.
>
> Yes, but flonum's can't be used in some contexts:
>
> * bit ops
>
> * modulo-2^32 arithmetic (i.e. C-like 32-bit wraparound arithmetic)
>
> I don't see a way to hack these using flonums. Bignums would solve  
> the problem but they're slow.

The bitwise-* procedures & arithmetic-shift are supposed to work w/  
fixnums & integer flonums. The magnitude of a flonum is converted to  
a C unsigned int (on 32 bit machines, u_int64_t otherwise), operated  
upon, then the result is returned as a fixnum or flonum, depending on  
the precision of the operation result.

The modulo-2^32 is a problem though.

Sorry,
Kon

>
<snip>