Re: index calculus vs pollard rho

[Georgi Guninski <guninski@guninski.com>]

On Thu, Apr 03, 2014 at 03:22:02PM +0200, Pascal Molin wrote:
> The following znlog uses index calculus on a 46 bits subgroup, but p itself
> is large,
> this is slow (and memory-demanding)
> 
> *gp* > p=nextprime(2^120); znlog(Mod(3,p),Mod(2,p),p-1)
> 
> time = 51,617 ms.
> 
> %21 = 391862826185609110238504885400229618
> 
> 
> while the same is easier to compute with pollard
> 
> *gp* > p=nextprime(2^120); znlog(Mod(3,5*p)^4,Mod(2,5*p)^4,p-1)
> 
> time = 5,615 ms.
> 

Are you sure this is not pure luck and this
works for sufficiently man 120 bit primes?

This appears slow for me:

p=nextprime(2^120-2^60); znlog(Mod(3,5*p)^(4),Mod(2,5*p)^(4),p-1)

> %20 = 391862826185609110238504885400229618
> 
> 
> I do not know how the choice of algorithm is made, maybe there is an issue
> here.
> -- 
> Pascal