That's not to say ECL can't be improved though.

How long does a naked for loop take?

Bill.

On 10 Sep, 06:50, William Stein <[email protected]> wrote:
> On Wed, Sep 9, 2009 at 10:38 PM, Bill Hart <[email protected]> 
> wrote:
>
> > This program only takes 0.68s in C using a pretty naive mpz program on
> > sage.math. I doubt the memory allocation is really relevant. The
> > interpreter overhead is by far the greatest component
>
> Thanks for looking into this.  I was also dubious because Python also
> has a memory pool for its integers.
>
> William
>
>
>
>
>
>
>
> > Bill.
>
> > On 9 Sep, 17:57, Nils Bruin <[email protected]> wrote:
> >> Inspired by a little experiment we did to see if there is room to
> >> improve to ECL's bignum performance, we ran a little experiment
> >> computing fibonacci numbers (we wanted to test addition because it was
> >> mainly ECL's memory management that was under consideration)
>
> >> with the following defs:
>
> >> def fibo(n):
> >>   a=1
> >>   b=1
> >>   c=0
> >>   for i in range(n):
> >>     c=a
> >>     a=a+b
> >>     b=c
> >>   return b
>
> >> def test(n,m):
> >>   for i in range(m):
> >>     _=fibo(n)
>
> >> sage: timeit("test(4000,4000)")
> >> 5 loops, best of 3: 6.99 s per loop
> >> sage: time test(4000,4000)
> >> CPU times: user 7.10 s, sys: 0.00 s, total: 7.10 s
> >> Wall time: 7.11 s
> >> sage: time test(4000,4000)
> >> CPU times: user 7.24 s, sys: 0.00 s, total: 7.24 s
> >> Wall time: 7.24 s
> >> sage: time test(4000,4000)
> >> CPU times: user 7.38 s, sys: 0.00 s, total: 7.38 s
> >> Wall time: 7.39 s
> >> sage: time test(4000,4000)
> >> CPU times: user 7.10 s, sys: 0.00 s, total: 7.10 s
> >> Wall time: 7.11 s
> >> sage: time test(4000,4000)
> >> CPU times: user 7.05 s, sys: 0.00 s, total: 7.05 s
> >> Wall time: 7.06 s
>
> >> In ECL, this took 14.8 sec (uncompiled) and 8.8 sec (compiled). Of
> >> course, this particular programming exercise was just to see how fast
> >> one can shove information into GMP and ECL at this point definitely
> >> doesn't claim to be particularly optimized for that particular task,
> >> but as you can see, straightforward python and sage do a good job.
>
> >> For comparison, Magma takes about 8.4 secs, as does the code above
> >> when I initialize a=int(1) etc.
>
> --
> William Stein
> Associate Professor of Mathematics
> University of Washingtonhttp://wstein.org
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send an email to [email protected]
To unsubscribe from this group, send an email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---

Reply via email to