| Bill Allombert on Tue, 08 Sep 2015 22:10:21 +0200 |
[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]
| Re: Speed up {Flx,FpX,FpXQX}_divrem_basecase for suitable polynomials |
On Tue, Sep 08, 2015 at 11:47:30AM +0200, Peter Bruin wrote:
> Bonjour,
>
> When computing with non-prime finite fields, one is often free to choose
> a defining polynomial f (of some given degree n) over the prime field.
> If we write f in the form c*X^n + f1 with deg(f1) < n, then for division
> with remainder modulo f, it "should be" best to take deg(f1) as small as
> possible. However, the current code for division with remainder in PARI
> does not yet give an advantage in that case.
>
> The attached patch modifies the functions {Flx,FpX,FpXQX}_divrem_basecase
> and Flx_rem_basecase in order to reduce the complexity of computing the
> (quotient and) remainder of g by f from O((deg g - deg f)(deg f)) to
> O((deg g - deg f)(deg f1)).
>
> It passes the test suite and gives a noticeable speedup in my code for
> computing with curves over finite fields. For example, with this patch,
> multiplying two 10 x 10 matrices over a field of size 59^100 becomes
> about 25% faster when choosing a polynomial of the form x^100 + O(x^3)
> instead of an arbitrary polynomial of degree 100 over F_59; previously
> there was no visible difference.
Thanks, I have applied your patch.
I join a test suite:
p=59, deg 100: P1:288 P2:396
p=2^607-1, deg 10: P1:569 P2:608
p=2^607-1, deg 20: P1:1036 P2:1053
p=2^607-1, deg 30: P1:1460 P2:1677
p=2^607-1, deg 40: P1:512 P2:553
Cheers,
Bill
fun(p,P,N=1)=
{
a=ffgen(P*Mod(1,p),'a);
b=ffgen(ffinit(p,poldegree(P)),'b);
M1=matrix(10,10,i,j,random(a));M2=matrix(10,10,i,j,random(a));
M1b=matrix(10,10,i,j,random(b));M2b=matrix(10,10,i,j,random(b));
gettime();
for(i=1,N,M1*M2);
print1("P1:",gettime());
for(i=1,N,M1b*M2b);
print(" P2:",gettime());
}
fun(59,x^100+x+38,100)
fun(2^607-1,x^20+2,10)
fun(2^607-1,x^30+x+52,10)
fun(2^607-1,x^40+x+25,10)
fun(2^607-1,x^100+2*x+58,1)