One-dimensional gravitational N-body code
A one-dimensional gravitational N-body code.
wendy solves the one-dimensional gravitational N-body problem to machine precision with an efficient algorithm [O(log N) / particle-collision]. Alternatively, it can solve the problem with approximate integration, but with exact forces.
Jo Bovy (University of Toronto): bovy - at - astro - dot - utoronto - dot - ca
Install the latest release using
pip install wendy
or clone/fork/download the repository and install using
sudo python setup.py install
or locally using
python setup.py install --user
The behavior of the parallel sorting algorithm used when setting
sort='parallel' in the approximate version of the N-body code (
approx=True) is controlled by a few compilation-time variables:
PARALLEL_SERIAL_SORT_SWITCH, which sets the length of an array below which the serial sort is used,
PARALLEL_SERIAL_MERGE_SWITCH, which sets the length of an array below which a serial merge is used (as part of the
mergesort sorting algorithm used), and
PARALLEL_SORT_NUM_THREADS, the number of threads used in the parallel sorting algorithm. By default, these are set to
PARALLEL_SORT_NUM_THREADS=32, which appear to work well. Significant speed-ups can be obtained by optimizing these for your system and specific problem. They can be set to different values by running, e.g.,
export CFLAGS="$CFLAGS -D PARALLEL_SERIAL_SORT_SWITCH=10 -D PARALLEL_SERIAL_MERGE_SWITCH=10 -D PARALLEL_SORT_NUM_THREADS=2"
before compiling the code (if you are trying to change them, make sure to force a re-compilation by removing the
wendy.nbody to initialize a generator object for initial (x,v) with masses m. The generator then returns the state of the system at equally-spaced time intervals:
g= wendy.nbody(x,v,m,0.05) # delta t = 0.05 next_x, next_v= next(g) # at t=0.05 next_x, next_v= next(g) # at t=0.10 ...
The generator initialization with
wendy.nbody has options to (a) solve the problem exactly or not using
approx=, (b) include an external harmonic oscillator potential
omega^2 x^2 / 2 with
omega= (both for exact and approximate solutions), and (c) include an arbitrary external force
ext_force=, only for the approximate solution).
- Phase mixing and violent relaxation in one dimension: example notebook (run locally to see movies, or view on nbviewer)
- Adiabatic vs. non-adiabatic energy injection for an exponential disk: example notebook (run locally to see movies, or view on nbviewer)
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.