|
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
| • |
Time
to compute the force
|
|
|
|
on a
particle p behaves like
|
|
|
|
O(log
n) instead of O(n)
|
|
|
| • |
Experimental
results:
|
|
|
Number
of 2-body
|
|
|
|
interactions
calculated for a
|
|
|
5000
particle system is
|
|
|
about
1/10 the number
|
|
|
|
needed
for the
|
|
|
|
straightforward
algorithm
|
|
|
|
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
| • |
Reference:
|
|
|
|
[Barnes
& Hut, Nature 1986]
|
|
| • |
Note
that this is still an
|
|
|
|
expensive
calculation
|
|
|
|
– |
O(n log n) time needed for
|
|
|
|
|
force
calculations for one
|
|
|
|
time
step
|
|
|
|
– |
Ideally,
a galaxy simulation
|
|
|
|
should
use millions of stars
|
|
|
|