Bill Allombert on Fri, 27 Jan 2017 11:17:54 +0100


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

Re: Getting a very large polynomial into pari/gp


On Sun, Jan 01, 2017 at 03:47:11AM +0000, Neill Clift wrote:
> On 12/31/2016 9:05 AM, Bill Allombert wrote:
> > The libpari library provides a set of functions to deal with polynomials
> > over F_p that implement the half-GCD algorithm but they are not directly
> > available in GP.
> >
> > Try:
> > install("FpX_gcd","GGG")
> > p = 2^64-59;
> > a = Pol(read("a.gp"));
> > b = Pol(read("b.gp"));
> > FpX_gcd(a,b,p)
> >
> > and actually since p is less than 2^64 there is another function that
> > use a more compact representation of polynomials (Vecsmall), but the
> > above function will do the conversions for you.
> > Read about FpX, FpX_gcd, Flx and Flx_gcd in the documentation.
> While the other gcd ran overnight this one failed in an hour.
> 
> (18:03) gp > FpX_gcd(a,b,p)
>    ***   at top-level: FpX_gcd(a,b,p)
>    ***                 ^--------------
>    *** FpX_gcd: not enough memory
> (19:08) gp >
> 
> This seems to be different from the stack size messages I have got 
> (stack size set to 64G).

After investigating with Neill we found this was due to a problem with
GMP 6.0.0 with LLP64 system when using more than 2GB of memory,
so I have rebuild the Pari installer with GMP 6.1.2 which fix this problem:

<http://pari.math.u-bordeaux.fr/pub/pari/windows/snapshots/Pari64-2-9-1_gmp612.exe>

Please test that it does not introduce other issues.
If all go well, I will use this GMP build for the next stable releases.

Cheers,
Bill.