Form is really slow and fast for me. What could this mean?

parisse parisse at mozart.ujf-grenoble.fr
Fri Mar 29 07:09:17 CET 2002


Richard Fateman wrote:

> It is not necessary to have two .sorts in there.
> I took out the first one, and it was just as fast.
> Why should this help?
> 
> I tried this:
> 
> Symbols x,y,z;
> Local q=(1+x+y+z)^20*(1+(1+z+x+y)^20)-(1+x+z+y)^20*(1+(1+z+x+y)^20);
> Print;
> .end
> 
> and it took about twice as long as the longest time.
> (190 seconds)
> 
> I also tried
> Local q=(1+x+y+z)^40
>   and that takes about .98 seconds
> Local q=(1+x+y+z)^40- (1+x+z+y)^40
>   and that takes about 1.7 seconds.
> 
> What is going on?


Taking the power is much faster simply because you do *not* apply
the chinese powering algorithm as a simple cost analysis shows
for any multivariate polynomial. Just make the 40* multiplication loop.







More information about the GiNaC-devel mailing list