Bill Allombert on Sat, 28 Sep 2002 10:56:24 +0200


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

Re: t_INT API


On Fri, Sep 27, 2002 at 12:33:57PM -0700, Ilya Zakharevich wrote:
> On Fri, Sep 27, 2002 at 01:40:49PM +0200, Bill Allombert wrote:
> > I propose to use the following interface:
> > #define int_MSW(x) 
> > x being a t_INT, return a pointer to the most significant word of an t_INT.
> > 
> > #define int_LSW(x) 
> > x being a t_int, return a pointer to the least significant word of an t_INT.
> > 
> > #define int_nextW(x) 
> > x pointing to a mantissa word, return the next (more significant) mantissa word.
> > 
> > #define int_precW(x)
> > x pointing to a mantissa word, return the previous (less significant) mantissa word.
> 
> [Do not see any usage for next/prec.]  I need a way to access the word
> number n.

next/prec are needed for left-right binary style algorithms.
And I agree, we need a way to access word n.
maybe:
#define int_word(x,n)

One question though:
this system is fine if we only have to support PARI and current GMP format.
However should we had a provision for other formats, e.g. formats
that use less than 32 bits per words like "GMP with nails" ???
In this case we could only provide a macro to access bits individualy.
But that maybe slower and an overkill.

Cheers,
Bill.