High performance simulations of random walks in various environments. The costly functions are written in C++ and then ported to Python using PyBind11.
Installation is automated in bash with the setup.sh script.
Dependences:
Numerical data is handled using the npquad numpy extension. Floating point data is returned as numpy arrays with dtype=np.quad. Note that quad precision support is limited so downcasting to np.float64 after all calculations are done is recommended. Some helper functions are located in /pysrc/fileIO.py and /pysrc/quadMath.py.
- pyDiffusion
- Contains Python code which wraps the generated C++ library.
- src
- Contains the C++ library to simulate many particles diffusing on a 1D lattice in a space-time random field.
- runFiles
- Python and Bash scripts that measure statistics about systems of diffusing particles. The code is written to run on a HPC using Slurm.
import sys
sys.path.append("pysrc")
from pydiffusionPDF as DiffusionPDF
import matplotlib
matplotlib.use("Agg")
from matplotlib import pyplot as plt
# Specify some constants like the number of particles, beta, and number of
# timesteps to evolve the system
nParticles = 1e50
beta = 1
num_of_timesteps = 10_000
# Initialize the system with parameters and other key word arguments
d = DiffusionPDF(
nParticles,
beta=beta,
occupancySize=num_of_timesteps,
probDistFlag=False,
)
# Evolve the system to the specified number of timesteps
d.evolveToTime(num_of_timesteps)
# Get the rightmost edge and the time
maxEdge = d.maxDistance
time = d.time
# Plot the rightmost edge over time and save
fig, ax = plt.subplots()
ax.set_xlabel("Time")
ax.set_ylabel("Distance to Center")
ax.set_xscale("log")
ax.set_yscale("log")
ax.plot(time, maxEdge)
plt.show()