n-body problem at shootout.alioth.debian.org

[bearophileHUGS at lycos.com]

Peter Maas:
> I have noticed that in the language shootout at shootout.alioth.debian.org
> the Python program for the n-body problem is about 50% slower than the Perl
> program. This is an unusual big difference. I tried to make the Python program
> faster but without success. Has anybody an explanation for the difference?
> It's pure math so I expected Perl and Python to have about the same speed.

I don't have a good answer for you, but if you profile that code you
find that most of the running time is used here, in this nested loop:

for j in xrange(i + 1, nbodies):
            b2 = bodies[j]

            dx = b_x - b2.x
            dy = b_y - b2.y
            dz = b_z - b2.z

            distance = sqrt(dx*dx + dy*dy + dz*dz)
            aux = dt / (distance*distance*distance)
            b_mass_x_mag = b_mass * aux
            b2_mass_x_mag = b2.mass * aux

            b.vx -= dx * b2_mass_x_mag
            b.vy -= dy * b2_mass_x_mag
            b.vz -= dz * b2_mass_x_mag
            b2.vx += dx * b_mass_x_mag
            b2.vy += dy * b_mass_x_mag
            b2.vz += dz * b_mass_x_mag

If you are using Python you can use Psyco, it helps a lot (473 s
instead of 2323 s, against 1622 s for the Perl version):
http://shootout.alioth.debian.org/sandbox/benchmark.php?test=nbody&lang=all
If you need speed, you can also use ShedSkin that compiles this Python
program producing an executable essentially fast as the C++ version
(about 16 s on the same computer, the Shootout autor isn't going to add
ShedSkin to the language list yet).

Bye,
bearophile