class SequentialIterationSuite:
"""
An example benchmark that times the performance of various kinds
of iterating over dictionaries in Python.
"""
timeout=1000
params = ['python', 'njit', 'njit_parallel']
def setup(self, engine):
np.random.seed(0)
self.sim = Simulation({}, 108, 100, 1e-6, file_path=None, m=1, T=0, dx=1, save_every_x_iters=1, gpu=False, shape='fcc', engine=engine)
def time_run(self, engine):
self.sim.run()
def test_run(simulation_params):
np.random.seed(4)
d = Simulation(**simulation_params).run()
df = d.diagnostic_df()
for key, tolerance in tolerances.items():
fitting = np.allclose(df.iloc[0][key],
df.iloc[-1][key],
atol=tolerance)
if not fitting:
df.plot("t", ["kinetic_energy", "potential_energy"])
plt.show()
break
if not fitting:
raise ValueError()
shutil.rmtree(os.path.dirname(simulation_params["file_path"]))
return d
def test_gpu_run(simulation_params):
simulation_params["N"] = 256
simulation_params["file_path"] = "/tmp/nbody_gpu_test_run/data{0:08d}.h5"
simulation_params["gpu"] = True
d = Simulation(**simulation_params).run().diagnostic_values
for key in ["kinetic_energy", "potential_energy"]:
assert np.isclose(d[min(d)][key], d[max(d)][key])
shutil.rmtree(os.path.dirname(gpu_simulation_params["file_path"]))
return d
def simulation_numba(request):
np.random.seed(0)
sim = Simulation(
{},
108,
100,
1e-6,
file_path=None,
m=1,
T=0,
dx=1,
save_every_x_iters=1,
gpu=False,
shape="fcc",
engine=request.param,
)
sim.run()
return sim
def simulation_python():
np.random.seed(0)
sim = Simulation(
{},
108,
100,
1e-6,
file_path=None,
m=1,
T=0,
dx=1,
save_every_x_iters=1,
gpu=False,
shape="fcc",
engine="python",
)
sim.run()
return sim
def setup(n):
np.random.seed(0)
sim = Simulation({},
n,
100,
1e-6,
file_path=None,
m=1,
T=0,
dx=1,
save_every_x_iters=1,
gpu=False,
shape='fcc')
return sim
def unstable_simulation(request):
sim = Simulation(
{},
2,
100,
request.param,
file_path=None,
m=1,
T=0,
dx=1,
save_every_x_iters=1,
gpu=False,
)
sim.r = np.array([[0, 0, 0], [1, 0, 0]], dtype=float)
sim.extrapolate_old_r()
sim.run()
return sim
def test_well_initialized():
np.random.seed(4)
d = Simulation(**simulation_params)
# 32 unique positions
assert np.unique(d.r, axis=0).shape[0] == d.N
def setup(self, engine):
np.random.seed(0)
self.sim = Simulation({}, 108, 100, 1e-6, file_path=None, m=1, T=0, dx=1, save_every_x_iters=1, gpu=False, shape='fcc', engine=engine)