Bug#724320: gmp: x32: sizeof(mp_limb_t)!=sizeof(void*) is not supported by GAP and PARI

Bill Allombert Bill.Allombert at math.u-bordeaux1.fr
Mon Sep 30 17:47:07 CEST 2013


On Sat, Sep 28, 2013 at 09:40:07PM +0200, Vincent Lefevre wrote:
> On 2013-09-28 10:38:24 -0500, Steve M. Robbins wrote:
> > Hi,
> > 
> > On September 28, 2013 10:26:33 AM Vincent Lefevre wrote:
> > > On 2013-09-28 09:03:53 +0200, Marc Glisse wrote:
> > 
> > > > the whole point of creating x32 was to benefit from the speed advantages
> > > > of amd64 without having a pointer size of 64 bits. Those speed advantages
> > > > include using more registers, but also using hardware 64 bit long long.
> > > > The speed difference between 32 and 64 bits [ ... ] so it doesn't make
> > > > sense to me to penalize x32
> > >
> > > I agree. PARI is not the only software that uses GMP.
> > 
> > OK, so I may not have been completely clear, but what I was asking is which 
> > makes most sense for Debian's GMP package: 8 bytes or 4 bytes.  I think what 
> > Marc and Vincent are saying is that 8 bytes is the better choice.  Did I read 
> > that correctly?
> 
> Yes, I think that 8 bytes is the better choice if this leads to
> faster code and doesn't break other software than PARI.

It does break GAP at least.

Cheers,
-- 
Bill. <ballombe at debian.org>

Imagine a large red swirl here. 


More information about the gmp-discuss mailing list