def test_boosted_frame_sim_twoproc():
"Test the example input script with two procs in `docs/source/example_input`"
temporary_dir = './tests/tmp_test_dir'
# Create a temporary directory for the simulation
# and copy the example script into this directory
if os.path.exists(temporary_dir):
shutil.rmtree(temporary_dir)
os.mkdir(temporary_dir)
shutil.copy('./docs/source/example_input/boosted_frame_script.py',
temporary_dir)
# Shortcut for the script file, which is repeatedly changed
script_filename = os.path.join(temporary_dir, 'boosted_frame_script.py')
# Read the script
with open(script_filename) as f:
script = f.read()
# Modify the script so as to enable finite order
script = replace_string(script, 'n_order = -1', 'n_order = 16')
script = replace_string(script, 'track_bunch = False',
'track_bunch = True')
with open(script_filename, 'w') as f:
f.write(script)
# Launch the script from the OS
response = os.system('cd %s; mpirun -np 2 python boosted_frame_script.py' %
temporary_dir)
assert response == 0
# Check that the particle ids are unique at each iterations
ts = OpenPMDTimeSeries(os.path.join(temporary_dir, 'lab_diags/hdf5'))
print('Checking particle ids...')
start_time = time.time()
for iteration in ts.iterations:
pid, = ts.get_particle(["id"], iteration=iteration)
assert len(np.unique(pid)) == len(pid)
end_time = time.time()
print("%.2f seconds" % (end_time - start_time))
# Suppress the temporary directory
shutil.rmtree(temporary_dir)
def check_theory_pml(show, boundaries):
"""
Check that the transverse E and B field are close to the high-gamma
theory for a gaussian bunch
"""
ts = OpenPMDTimeSeries(os.path.join(temporary_dir, 'diags/hdf5/'))
for iteration in ts.iterations:
compare_E(ts,
'x',
m=0,
iteration=iteration,
rtol=rtol0,
show=show,
boundaries=boundaries)
compare_E(ts,
'x',
m=1,
iteration=iteration,
rtol=rtol1,
show=show,
boundaries=boundaries)
def test_boosted_frame_sim_twoproc():
"Test the example input script with two procs in `docs/source/example_input`"
temporary_dir = './tests/tmp_test_dir'
# Create a temporary directory for the simulation
# and copy the example script into this directory
if os.path.exists( temporary_dir ):
shutil.rmtree( temporary_dir )
os.mkdir( temporary_dir )
shutil.copy('./docs/source/example_input/boosted_frame_script.py',
temporary_dir )
# Enter the temporary directory
os.chdir( temporary_dir )
# Read the script and check that the targeted lines are present
with open('boosted_frame_script.py') as f:
script = f.read()
# Modify the script so as to enable particle tracking
script = script.replace('track_bunch = False', 'track_bunch = True')
with open('boosted_frame_script.py', 'w') as f:
f.write(script)
# Launch the script from the OS
response = os.system( 'mpirun -np 2 python boosted_frame_script.py' )
assert response==0
# Check that the particle ids are unique at each iterations
ts = OpenPMDTimeSeries('./lab_diags/hdf5')
print('Checking particle ids...')
start_time = time.time()
for iteration in ts.iterations:
pid, = ts.get_particle(["id"], iteration=iteration)
assert len(np.unique(pid)) == len(pid)
end_time = time.time()
print( "%.2f seconds" %(end_time-start_time))
# Exit the temporary directory and suppress it
os.chdir('../../')
shutil.rmtree( temporary_dir )
from __future__ import absolute_import, division, print_function, unicode_literals
# OpenPMD imports
from opmd_viewer import OpenPMDTimeSeries
# read in the longitdunal electric field component
path_to_data = '../../rsfbpic/package_data'
field_name = 'E'
coord = 'z'
iteration = 280
mode = 0
theta = 0.
plot = False
# open the file; instantiate "time series" object
time_series = OpenPMDTimeSeries(path_to_data)
# read the specified field
ez, info_ez = time_series.get_field(iteration=iteration, \
field=field_name, \
coord=coord, \
m=mode, \
theta=theta, \
plot=plot)
print()
print("axes = ", info_ez.axes)
print()
print("rmin, rmax = ", info_ez.rmin, "; ", info_ez.rmax)
print("zmin, zmax = ", info_ez.zmin, "; ", info_ez.zmax)
print()
def GetTimeStepsInOpenPMDLocation(self, location):
ts = OpenPMDTimeSeries(location, check_all_files=False)
return ts.iterations
def add_particle_bunch_openPMD(sim,
q,
m,
ts_path,
z_off=0.,
species=None,
select=None,
iteration=None,
boost=None,
z_injection_plane=None,
initialize_self_field=True):
"""
Introduce a relativistic particle bunch in the simulation,
along with its space charge field, loading particles from an openPMD
timeseries.
Parameters
----------
sim : a Simulation object
The structure that contains the simulation.
q : float (in Coulomb)
Charge of the particle species
m : float (in kg)
Mass of the particle species
ts_path : string
The path to the directory where the openPMD files are.
For the moment, only HDF5 files are supported. There should be
one file per iteration, and the name of the files should end
with the iteration number, followed by '.h5' (e.g. data0005000.h5)
z_off: float (in meters)
Shift the particle positions in z by z_off. By default the initialized
phasespace is centered at z=0.
species: string
A string indicating the name of the species
This is optional if there is only one species
select: dict, optional
Either None or a dictionary of rules
to select the particles, of the form
'x' : [-4., 10.] (Particles having x between -4 and 10 microns)
'ux' : [-0.1, 0.1] (Particles having ux between -0.1 and 0.1 mc)
'uz' : [5., None] (Particles with uz above 5 mc)
iteration: integer (optional)
The iteration number of the openPMD file from which to extract the
particles.
boost : a BoostConverter object, optional
A BoostConverter object defining the Lorentz boost of
the simulation.
z_injection_plane: float (in meters) or None
When `z_injection_plane` is not None, then particles have a ballistic
motion for z<z_injection_plane. This is sometimes useful in
boosted-frame simulations.
`z_injection_plane` is always given in the lab frame.
initialize_self_field: bool, optional
Whether to calculate the initial space charge fields of the bunch
and add these fields to the fields on the grid (Default: True)
"""
# Import openPMD viewer
try:
from opmd_viewer import OpenPMDTimeSeries
except ImportError:
raise ImportError(
'The package `opmd_viewer` is required to restart from checkpoints.'
'\nPlease install it from https://github.com/openPMD/openPMD-viewer'
)
ts = OpenPMDTimeSeries(ts_path)
# Extract phasespace and particle weights
x, y, z, ux, uy, uz, w = ts.get_particle(
['x', 'y', 'z', 'ux', 'uy', 'uz', 'w'],
iteration=iteration,
species=species,
select=select)
# Convert the positions from microns to meters
x *= 1.e-6
y *= 1.e-6
z *= 1.e-6
# Shift the center of the phasespace to z_off
z = z - np.average(z, weights=w) + z_off
# Add the electrons to the simulation, and calculate the space charge
ptcl_bunch = add_particle_bunch_from_arrays(
sim,
q,
m,
x,
y,
z,
ux,
uy,
uz,
w,
boost=boost,
z_injection_plane=z_injection_plane,
initialize_self_field=initialize_self_field)
return ptcl_bunch
def run_simulation(gamma_boost, use_separate_electron_species):
"""
Run a simulation with a laser pulse going through a gas jet of ionizable
N5+ atoms, and check the fraction of atoms that are in the N5+ state.
Parameters
----------
gamma_boost: float
The Lorentz factor of the frame in which the simulation is carried out.
use_separate_electron_species: bool
Whether to use separate electron species for each level, or
a single electron species for all levels.
"""
# The simulation box
zmax_lab = 20.e-6 # Length of the box along z (meters)
zmin_lab = 0.e-6
Nr = 3 # Number of gridpoints along r
rmax = 10.e-6 # Length of the box along r (meters)
Nm = 2 # Number of modes used
# The particles of the plasma
p_zmin = 5.e-6 # Position of the beginning of the plasma (meters)
p_zmax = 15.e-6
p_rmin = 0. # Minimal radial position of the plasma (meters)
p_rmax = 100.e-6 # Maximal radial position of the plasma (meters)
n_atoms = 0.2 # The atomic density is chosen very low,
# to avoid collective effects
p_nz = 2 # Number of particles per cell along z
p_nr = 1 # Number of particles per cell along r
p_nt = 4 # Number of particles per cell along theta
# Boosted frame
boost = BoostConverter(gamma_boost)
# Boost the different quantities
beta_boost = np.sqrt(1. - 1. / gamma_boost**2)
zmin, zmax = boost.static_length([zmin_lab, zmax_lab])
p_zmin, p_zmax = boost.static_length([p_zmin, p_zmax])
n_atoms, = boost.static_density([n_atoms])
# Increase the number of particles per cell in order to keep sufficient
# statistics for the evaluation of the ionization fraction
if gamma_boost > 1:
p_nz = int(2 * gamma_boost * (1 + beta_boost) * p_nz)
# The laser
a0 = 1.8 # Laser amplitude
lambda0_lab = 0.8e-6 # Laser wavelength
# Boost the laser wavelength before calculating the laser amplitude
lambda0, = boost.copropag_length([lambda0_lab], beta_object=1.)
# Duration and initial position of the laser
ctau = 10. * lambda0
z0 = -2 * ctau
# Calculate laser amplitude
omega = 2 * np.pi * c / lambda0
E0 = a0 * m_e * c * omega / e
B0 = E0 / c
def laser_func(F, x, y, z, t, amplitude, length_scale):
"""
Function that describes a Gaussian laser with infinite waist
"""
return( F + amplitude * math.cos( 2*np.pi*(z-c*t)/lambda0 ) * \
math.exp( - (z - c*t - z0)**2/ctau**2 ) )
# Resolution and number of timesteps
dz = lambda0 / 16.
dt = dz / c
Nz = int((zmax - zmin) / dz) + 1
N_step = int(
(2. * 40. * lambda0 + zmax - zmin) / (dz * (1 + beta_boost))) + 1
# Get the speed of the plasma
uz_m, = boost.longitudinal_momentum([0.])
v_plasma, = boost.velocity([0.])
# The diagnostics
diag_period = N_step - 1 # Period of the diagnostics in number of timesteps
# Initial ionization level of the Nitrogen atoms
level_start = 2
# Initialize the simulation object, with the neutralizing electrons
# No particles are created because we do not pass the density
sim = Simulation(Nz,
zmax,
Nr,
rmax,
Nm,
dt,
zmin=zmin,
v_comoving=v_plasma,
use_galilean=False,
boundaries='open',
use_cuda=use_cuda)
# Add the charge-neutralizing electrons
elec = sim.add_new_species(q=-e,
m=m_e,
n=level_start * n_atoms,
p_nz=p_nz,
p_nr=p_nr,
p_nt=p_nt,
p_zmin=p_zmin,
p_zmax=p_zmax,
p_rmin=p_rmin,
p_rmax=p_rmax,
continuous_injection=False,
uz_m=uz_m)
# Add the N atoms
ions = sim.add_new_species(q=0,
m=14. * m_p,
n=n_atoms,
p_nz=p_nz,
p_nr=p_nr,
p_nt=p_nt,
p_zmin=p_zmin,
p_zmax=p_zmax,
p_rmin=p_rmin,
p_rmax=p_rmax,
continuous_injection=False,
uz_m=uz_m)
# Add the target electrons
if use_separate_electron_species:
# Use a dictionary of electron species: one per ionizable level
target_species = {}
level_max = 6 # N can go up to N7+, but here we stop at N6+
for i_level in range(level_start, level_max):
target_species[i_level] = sim.add_new_species(q=-e, m=m_e)
else:
# Use the pre-existing, charge-neutralizing electrons
target_species = elec
level_max = None # Default is going up to N7+
# Define ionization
ions.make_ionizable(element='N',
level_start=level_start,
level_max=level_max,
target_species=target_species)
# Set the moving window
sim.set_moving_window(v=v_plasma)
# Add a laser to the fields of the simulation (external fields)
sim.external_fields = [
ExternalField(laser_func, 'Ex', E0, 0.),
ExternalField(laser_func, 'By', B0, 0.)
]
# Add a particle diagnostic
sim.diags = [
ParticleDiagnostic(
diag_period,
{"ions": ions},
particle_data=["position", "gamma", "weighting", "E", "B"],
# Test output of fields and gamma for standard
# (non-boosted) particle diagnostics
write_dir='tests/diags',
comm=sim.comm)
]
if gamma_boost > 1:
T_sim_lab = (2. * 40. * lambda0_lab + zmax_lab - zmin_lab) / c
sim.diags.append(
BackTransformedParticleDiagnostic(zmin_lab,
zmax_lab,
v_lab=0.,
dt_snapshots_lab=T_sim_lab / 2.,
Ntot_snapshots_lab=3,
gamma_boost=gamma_boost,
period=diag_period,
fldobject=sim.fld,
species={"ions": ions},
comm=sim.comm,
write_dir='tests/lab_diags'))
# Run the simulation
sim.step(N_step, use_true_rho=True)
# Check the fraction of N5+ ions at the end of the simulation
w = ions.w
ioniz_level = ions.ionizer.ionization_level
# Get the total number of N atoms/ions (all ionization levels together)
ntot = w.sum()
# Get the total number of N5+ ions
n_N5 = w[ioniz_level == 5].sum()
# Get the fraction of N5+ ions, and check that it is close to 0.32
N5_fraction = n_N5 / ntot
print('N5+ fraction: %.4f' % N5_fraction)
assert ((N5_fraction > 0.30) and (N5_fraction < 0.34))
# When different electron species are created, check the fraction of
# each electron species
if use_separate_electron_species:
for i_level in range(level_start, level_max):
n_N = w[ioniz_level == i_level].sum()
assert np.allclose(target_species[i_level].w.sum(), n_N)
# Check consistency in the regular openPMD diagnostics
ts = OpenPMDTimeSeries('./tests/diags/hdf5/')
last_iteration = ts.iterations[-1]
w, q = ts.get_particle(['w', 'charge'],
species="ions",
iteration=last_iteration)
# Check that the openPMD file contains the same number of N5+ ions
n_N5_openpmd = np.sum(w[(4.5 * e < q) & (q < 5.5 * e)])
assert np.isclose(n_N5_openpmd, n_N5)
# Remove openPMD files
shutil.rmtree('./tests/diags/')
# Check consistency of the back-transformed openPMD diagnostics
if gamma_boost > 1.:
ts = OpenPMDTimeSeries('./tests/lab_diags/hdf5/')
last_iteration = ts.iterations[-1]
w, q = ts.get_particle(['w', 'charge'],
species="ions",
iteration=last_iteration)
# Check that the openPMD file contains the same number of N5+ ions
n_N5_openpmd = np.sum(w[(4.5 * e < q) & (q < 5.5 * e)])
assert np.isclose(n_N5_openpmd, n_N5)
# Remove openPMD files
shutil.rmtree('./tests/lab_diags/')
import numpy as np
from os.path import basename
import matplotlib.pyplot as plt
from opmd_viewer import OpenPMDTimeSeries
from opmd_viewer.addons import LpaDiagnostics
from scipy.ndimage.filters import gaussian_filter
from scipy import fftpack
from scipy.signal import savgol_filter
from mpl_toolkits import mplot3d
from matplotlib import animation
from mpl_toolkits.mplot3d import Axes3D
from matplotlib import cm
ts_laser = LpaDiagnostics('./diags/hdf5/', check_all_files=False)
ts_particles = OpenPMDTimeSeries('./diags/hdf5/', check_all_files=False)
fields = {'cmap': 'coolwarm'}
particles = {'cmap': 'jet'}
SimulationName = 'TrailingPulse_PulseTrain'
#Make directories for plots
Path = 'Plots_' + SimulationName
if not os.path.exists(Path):
os.makedirs(Path)
if not os.path.exists(Path + '/EzLineout'):
os.makedirs(Path + '/EzLineout')
if not os.path.exists(Path + '/ExLineout'):
os.makedirs(Path + '/ExLineout')
if not os.path.exists(Path + '/ExLineoutTrailingPulse'):
def test_cpu_gpu_deposition(show=False):
"Function that is run by py.test, when doing `python setup.py test`"
# Skip this test if cuda is not installed
if not cuda_installed:
return
# Perform deposition for a few timesteps, with both the CPU and GPU
for hardware in ['cpu', 'gpu']:
if hardware == 'cpu':
use_cuda = False
elif hardware == 'gpu':
use_cuda = True
# Initialize the simulation object
sim = Simulation(Nz,
zmax,
Nr,
rmax,
Nm,
dt,
p_zmin,
p_zmax,
p_rmin,
p_rmax,
p_nz,
p_nr,
p_nt,
n_e,
zmin=zmin,
use_cuda=use_cuda)
# Tweak the velocity of the electron bunch
gamma0 = 10.
sim.ptcl[0].ux = sim.ptcl[0].x
sim.ptcl[0].uy = sim.ptcl[0].y
sim.ptcl[0].uz = np.sqrt(gamma0**2 - sim.ptcl[0].ux**2 -
sim.ptcl[0].uy**2 - 1)
sim.ptcl[0].inv_gamma = 1. / gamma0 * np.ones_like(sim.ptcl[0].x)
sim.ptcl[0].m = 1e10
# Add a field diagnostic
sim.diags = [
FieldDiagnostic(diag_period,
sim.fld,
fieldtypes=['rho', 'J'],
comm=sim.comm,
write_dir=os.path.join('tests', hardware))
]
### Run the simulation
sim.step(N_step)
# Check that the results are identical
ts_cpu = OpenPMDTimeSeries('tests/cpu/hdf5')
ts_gpu = OpenPMDTimeSeries('tests/gpu/hdf5')
for iteration in ts_cpu.iterations:
for field, coord in [('rho', ''), ('J', 'x'), ('J', 'z')]:
# Jy is not tested because it is zero
print('Testing %s at iteration %d' % (field + coord, iteration))
F_cpu, info = ts_cpu.get_field(field, coord, iteration=iteration)
F_gpu, info = ts_gpu.get_field(field, coord, iteration=iteration)
tolerance = 1.e-13 * (abs(F_cpu).max() + abs(F_gpu).max())
if not show:
assert np.allclose(F_cpu, F_gpu, atol=tolerance)
else:
if not np.allclose(F_cpu, F_gpu, atol=tolerance):
plot_difference(field, coord, iteration, F_cpu, F_gpu,
info)
# Remove the files used
shutil.rmtree('tests/cpu')
shutil.rmtree('tests/gpu')
def test_boosted_output(gamma_boost=10.):
"""
# TODO
Parameters
----------
gamma_boost: float
The Lorentz factor of the frame in which the simulation is carried out.
"""
# The simulation box
Nz = 500 # Number of gridpoints along z
zmax_lab = 0.e-6 # Length of the box along z (meters)
zmin_lab = -20.e-6
Nr = 10 # Number of gridpoints along r
rmax = 10.e-6 # Length of the box along r (meters)
Nm = 2 # Number of modes used
# Number of timesteps
N_steps = 500
diag_period = 20 # Period of the diagnostics in number of timesteps
dt_lab = (zmax_lab - zmin_lab) / Nz * 1. / c
T_sim_lab = N_steps * dt_lab
# Move into directory `tests`
os.chdir('./tests')
# Initialize the simulation object
sim = Simulation(
Nz,
zmax_lab,
Nr,
rmax,
Nm,
dt_lab,
0,
0, # No electrons get created because we pass p_zmin=p_zmax=0
0,
rmax,
1,
1,
4,
n_e=0,
zmin=zmin_lab,
initialize_ions=False,
gamma_boost=gamma_boost,
v_comoving=-0.9999 * c,
boundaries='open',
use_cuda=use_cuda)
sim.set_moving_window(v=c)
# Remove the electron species
sim.ptcl = []
# Add a Gaussian electron bunch
# Note: the total charge is 0 so all fields should remain 0
# throughout the simulation. As a consequence, the motion of the beam
# is a mere translation.
N_particles = 3000
add_elec_bunch_gaussian(sim,
sig_r=1.e-6,
sig_z=1.e-6,
n_emit=0.,
gamma0=100,
sig_gamma=0.,
Q=0.,
N=N_particles,
zf=0.5 * (zmax_lab + zmin_lab),
boost=BoostConverter(gamma_boost))
sim.ptcl[0].track(sim.comm)
# openPMD diagnostics
sim.diags = [
BoostedParticleDiagnostic(zmin_lab,
zmax_lab,
v_lab=c,
dt_snapshots_lab=T_sim_lab / 3.,
Ntot_snapshots_lab=3,
gamma_boost=gamma_boost,
period=diag_period,
fldobject=sim.fld,
species={"bunch": sim.ptcl[0]},
comm=sim.comm)
]
# Run the simulation
sim.step(N_steps)
# Check consistency of the back-transformed openPMD diagnostics:
# Make sure that all the particles were retrived by checking particle IDs
ts = OpenPMDTimeSeries('./lab_diags/hdf5/')
ref_pid = np.sort(sim.ptcl[0].tracker.id)
for iteration in ts.iterations:
pid, = ts.get_particle(['id'], iteration=iteration)
pid = np.sort(pid)
assert len(pid) == N_particles
assert np.all(ref_pid == pid)
# Remove openPMD files
shutil.rmtree('./lab_diags/')
os.chdir('../')
def LoadOpenPMDData(self):
"""OpenPMD Loader"""
# First check whether openPMD is installed
if not openpmd_installed:
raise RunTimeError(
"You need to install openPMD-viewer, e.g. with:\n"
"pip install openPMD-viewer")
# Scan the folder using openPMD-viewer
ts = OpenPMDTimeSeries(self._dataLocation, check_all_files=False)
# TODO: Change hasNonISUnits to False once unit reading is implemented
# Register the available fields
if ts.avail_fields is not None:
for field in ts.avail_fields:
# Vector field
if ts.fields_metadata[field]['type'] == 'vector':
available_coord = ['x', 'y', 'z']
# Register each coordinate of the vector
for coord in available_coord:
fieldName = field + '/' + coord
standardName = self.GiveStandardNameForOpenPMDQuantity(
fieldName)
self.AddDomainField(
FolderField("openPMD",
fieldName,
standardName,
self._dataLocation,
ts.iterations,
hasNonISUnits=True))
# Scalar field
if ts.fields_metadata[field]['type'] == 'scalar':
fieldName = field
standardName = self.GiveStandardNameForOpenPMDQuantity(
field)
self.AddDomainField(
FolderField("openPMD",
fieldName,
standardName,
self._dataLocation,
ts.iterations,
hasNonISUnits=True))
# Register the available species
if ts.avail_species is not None:
for species in ts.avail_species:
self.AddSpecies(Species(species))
for species_quantity in ts.avail_record_components[species]:
if species_quantity == "id":
self.AddRawDataTagsToSpecies(
species,
RawDataTags("openPMD", species_quantity,
self._dataLocation, ts.iterations,
species, species_quantity))
else:
self.AddRawDataToSpecies(
species,
FolderRawDataSet(
"openPMD",
species_quantity,
self.GiveStandardNameForOpenPMDQuantity(
species_quantity),
self._dataLocation,
ts.iterations,
species,
species_quantity,
hasNonISUnits=True))
np.set_printoptions(precision=2, linewidth=80)
pr = signac.get_project(root="../signac", search=False)
# We see there are a few values of a0 in the project. We now want the job id of the job with a certain a0 value.
job_id_set = pr.find_job_ids({'a0': 3})
job_id = next(iter(job_id_set))
# get the job handler
job = pr.open_job(id=job_id)
# get path to job's hdf5 files
h5_path = os.path.join(job.ws, "diags", "hdf5")
# open the full time series and see iteration numbers
time_series = OpenPMDTimeSeries(h5_path, check_all_files=True)
energy = hv.Dimension('energy', label='E', unit='MeV')
count = hv.Dimension('frequency', label='dQ/dE', unit='pC/MeV')
integral_streams = [
streams.Stream.define(
'Iteration',
iteration=param.Integer(default=4600,
doc='Time step in the simulation'))(),
streams.PointerX(rename={'x': 'limit_b'}),
streams.Tap(rename={'x': 'limit_a'})
]
e_max = 25 # MeV
def run_cpu_gpu_deposition(show=False, particle_shape='cubic'):
# Skip this test if cuda is not installed
if not cuda_installed:
return
# Perform deposition for a few timesteps, with both the CPU and GPU
for hardware in ['cpu', 'gpu']:
if hardware == 'cpu':
use_cuda = False
elif hardware == 'gpu':
use_cuda = True
# Initialize the simulation object
sim = Simulation(Nz,
zmax,
Nr,
rmax,
Nm,
dt,
zmin=zmin,
use_cuda=use_cuda,
particle_shape=particle_shape)
sim.ptcl = []
# Add an electron bunch (set the random seed first)
np.random.seed(0)
add_elec_bunch_gaussian(sim, sig_r, sig_z, n_emit, gamma0, sig_gamma,
Q, N)
# Add a field diagnostic
sim.diags = [
FieldDiagnostic(diag_period,
sim.fld,
fieldtypes=['rho', 'J'],
comm=sim.comm,
write_dir=os.path.join('tests', hardware))
]
### Run the simulation
sim.step(N_step)
# Check that the results are identical
ts_cpu = OpenPMDTimeSeries('tests/cpu/hdf5')
ts_gpu = OpenPMDTimeSeries('tests/gpu/hdf5')
for iteration in ts_cpu.iterations:
for field, coord in [('rho', ''), ('J', 'x'), ('J', 'z')]:
# Jy is not tested because it is zero
print('Testing %s at iteration %d' % (field + coord, iteration))
F_cpu, info = ts_cpu.get_field(field, coord, iteration=iteration)
F_gpu, info = ts_gpu.get_field(field, coord, iteration=iteration)
tolerance = 1.e-13 * (abs(F_cpu).max() + abs(F_gpu).max())
if not show:
assert np.allclose(F_cpu, F_gpu, atol=tolerance)
else:
if not np.allclose(F_cpu, F_gpu, atol=tolerance):
plot_difference(field, coord, iteration, F_cpu, F_gpu,
info)
# Remove the files used
shutil.rmtree('tests/cpu')
shutil.rmtree('tests/gpu')
def test_lpa_sim_twoproc_restart():
"Test the checkpoint/restart mechanism with two proc"
temporary_dir = './tests/tmp_test_dir'
# Create a temporary directory for the simulation
# and copy the example script into this directory
if os.path.exists( temporary_dir ):
shutil.rmtree( temporary_dir )
os.mkdir( temporary_dir )
shutil.copy('./docs/source/example_input/lwfa_script.py', temporary_dir )
# Shortcut for the script file, which is repeatedly changed
script_filename = os.path.join( temporary_dir,'lwfa_script.py' )
# Read the script and check that the targeted lines are present
with open( script_filename ) as f:
script = f.read()
# Check that the targeted lines are present
if script.find('save_checkpoints = False') == -1 \
or script.find('use_restart = False') == -1 \
or script.find('track_electrons = False') == -1 \
or script.find('n_order = -1') == -1 \
or script.find('N_step = 200') == -1:
raise RuntimeError('Did not find expected lines in lwfa_script.py')
# Modify the script so as to enable checkpoints and particle tracking
script = script.replace('save_checkpoints = False',
'save_checkpoints = True')
script = script.replace('track_electrons = False',
'track_electrons = True')
script = script.replace('N_step = 200', 'N_step = 101')
script = script.replace('n_order = -1',
'n_order = 16')
with open( script_filename, 'w' ) as f:
f.write(script)
# Launch the modified script from the OS, with 2 proc
response = os.system(
'cd %s; mpirun -np 2 python lwfa_script.py' %temporary_dir )
assert response==0
# Modify the script so as to enable restarts
script = script.replace('use_restart = False',
'use_restart = True')
script = script.replace('save_checkpoints = True',
'save_checkpoints = False')
with open( script_filename, 'w' ) as f:
f.write(script)
# Launch the modified script from the OS, with 2 proc
response = os.system(
'cd %s; mpirun -np 2 python lwfa_script.py' %temporary_dir )
assert response==0
# Check that the particle ids are unique at each iterations
ts = OpenPMDTimeSeries( os.path.join( temporary_dir, 'diags/hdf5') )
print('Checking particle ids...')
start_time = time.time()
for iteration in ts.iterations:
pid, = ts.get_particle(["id"], iteration=iteration)
assert len(np.unique(pid)) == len(pid)
end_time = time.time()
print( "%.2f seconds" %(end_time-start_time))
# Suppress the temporary directory
shutil.rmtree( temporary_dir )
def restart_from_checkpoint( sim, iteration=None ):
"""
Fills the Simulation object `sim` with data saved in a checkpoint.
More precisely, the following data from `sim` is overwritten:
- Current time and iteration number of the simulation
- Position of the boundaries of the simulation box
- Values of the field arrays
- Size and values of the particle arrays
Any other information (e.g. diagnostics of the simulation, presence of a
moving window, presence of a laser antenna, etc.) need to be set by hand.
For this reason, a successful restart will often require to modify the
original input script that produced the checkpoint, rather than to start
a new input script from scratch.
NB: This function should always be called *before* the initialization
of the moving window, since the moving window infers the position of
particle injection from the existing particle data.
Parameters
----------
sim: a Simulation object
The Simulation object into which the checkpoint should be loaded
iteration: integer (optional)
The iteration number of the checkpoint from which to restart
If None, the latest checkpoint available will be used.
"""
# Import openPMD-viewer
try:
from opmd_viewer import OpenPMDTimeSeries
except ImportError:
raise ImportError(
'The package `opmd_viewer` is required to restart from checkpoints.'
'\nPlease install it from https://github.com/openPMD/openPMD-viewer')
# Verify that the restart is valid (only for the first processor)
# (Use the global MPI communicator instead of the `BoundaryCommunicator`,
# so that this also works for `use_all_ranks=False`)
if comm.rank == 0:
check_restart( sim, iteration )
comm.barrier()
# Choose the name of the directory from which to restart:
# one directory per processor
checkpoint_dir = 'checkpoints/proc%d/hdf5' %comm.rank
ts = OpenPMDTimeSeries( checkpoint_dir )
# Select the iteration, and its index
if iteration is None:
iteration = ts.iterations[-1]
# Find the index of the closest iteration
i_iteration = np.argmin( abs(np.array(ts.iterations) - iteration) )
# Modify parameters of the simulation
sim.iteration = iteration
sim.time = ts.t[ i_iteration ]
# Export available species as a list
avail_species = ts.avail_species
if avail_species is None:
avail_species = []
# Load the particles
# Loop through the different species
if len(avail_species) == len(sim.ptcl):
for i in range(len(sim.ptcl)):
name = 'species %d' %i
load_species( sim.ptcl[i], name, ts, iteration, sim.comm )
else:
raise RuntimeError( \
"""Species numbers in checkpoint and simulation should be same, but
got {:d} and {:d}. Use add_new_species method to add species to
simulation or sim.ptcl = [] to remove them""".format(len(avail_species),
len(sim.ptcl)) )
# Record position of grid before restart
zmin_old = sim.fld.interp[0].zmin
# Load the fields
# Loop through the different modes
for m in range( sim.fld.Nm ):
# Load the fields E and B
for fieldtype in ['E', 'B']:
for coord in ['r', 't', 'z']:
load_fields( sim.fld.interp[m], fieldtype,
coord, ts, iteration )
# Record position after restart (`zmin` is modified by `load_fields`)
# and shift the global domain position in the BoundaryCommunicator
zmin_new = sim.fld.interp[0].zmin
sim.comm.shift_global_domain_positions( zmin_new - zmin_old )
def add_elec_bunch_openPMD(sim,
ts_path,
z_off=0.,
species=None,
select=None,
iteration=None,
boost=None,
filter_currents=True):
"""
Introduce a relativistic electron bunch in the simulation,
along with its space charge field, loading particles from an openPMD
timeseries.
Parameters
----------
sim : a Simulation object
The structure that contains the simulation.
ts_path : string
The path to the directory where the openPMD files are.
For the moment, only HDF5 files are supported. There should be
one file per iteration, and the name of the files should end
with the iteration number, followed by '.h5' (e.g. data0005000.h5)
z_off: float (in meters)
Shift the particle positions in z by z_off. By default the initialized
phasespace is centered at z=0.
species: string
A string indicating the name of the species
This is optional if there is only one species
select: dict, optional
Either None or a dictionary of rules
to select the particles, of the form
'x' : [-4., 10.] (Particles having x between -4 and 10 microns)
'ux' : [-0.1, 0.1] (Particles having ux between -0.1 and 0.1 mc)
'uz' : [5., None] (Particles with uz above 5 mc)
iteration: integer (optional)
The iteration number of the openPMD file from which to extract the
particles.
boost : a BoostConverter object, optional
A BoostConverter object defining the Lorentz boost of
the simulation.
filter_currents : bool, optional
Whether to filter the currents in k space (True by default)
"""
# Import openPMD viewer
try:
from opmd_viewer import OpenPMDTimeSeries
except ImportError:
raise ImportError(
'The package `opmd_viewer` is required to restart from checkpoints.'
'\nPlease install it from https://github.com/openPMD/openPMD-viewer'
)
ts = OpenPMDTimeSeries(ts_path)
# Extract phasespace and particle weights
x, y, z, ux, uy, uz, w = ts.get_particle(
['x', 'y', 'z', 'ux', 'uy', 'uz', 'w'],
iteration=iteration,
species=species,
select=select)
# Convert the positions from microns to meters
x *= 1.e-6
y *= 1.e-6
z *= 1.e-6
# Shift the center of the phasespace to z_off
z = z - (np.amax(z) + np.amin(z)) / 2 + z_off
# Add the electrons to the simulation, and calculate the space charge
add_elec_bunch_from_arrays(sim,
x,
y,
z,
ux,
uy,
uz,
w,
boost=boost,
filter_currents=filter_currents)
def run_sim(script_name, n_MPI, checked_fields, test_checkpoint_dir=False):
"""
Runs the script `script_name` from the folder docs/source/example_input,
with `n_MPI` MPI processes. The simulation is then restarted with
the same number of processes ; the code checks that the restarted results
are identical.
More precisely:
- The first simulation is run for N_step, then the random seed is reset
(for reproducibility) and the code runs for N_step more steps.
- Then a second simulation is launched, which reruns the last N_step.
"""
temporary_dir = './tests/tmp_test_dir'
# Create a temporary directory for the simulation
# and copy the example script into this directory
if os.path.exists(temporary_dir):
shutil.rmtree(temporary_dir)
os.mkdir(temporary_dir)
shutil.copy('./docs/source/example_input/%s' % script_name, temporary_dir)
# Shortcut for the script file, which is repeatedly changed
script_filename = os.path.join(temporary_dir, script_name)
# Read the script and check
with open(script_filename) as f:
script = f.read()
# Change default N_step, diag_period and checkpoint_period
script = replace_string(script, 'N_step = int(T_interact/sim.dt)',
'N_step = 200')
script = replace_string(script, 'diag_period = 50', 'diag_period = 10')
script = replace_string(script, 'checkpoint_period = 100',
'checkpoint_period = 50')
# For MPI simulations: modify the script to use finite-order
if n_MPI > 1:
script = replace_string(script, 'n_order = -1', 'n_order = 16')
# Modify the script so as to enable checkpoints
script = replace_string(script, 'save_checkpoints = False',
'save_checkpoints = True')
if test_checkpoint_dir:
# Try to change the name of the checkpoint directory
checkpoint_dir = './test_chkpt'
script = replace_string(
script, 'set_periodic_checkpoint( sim, checkpoint_period )',
'set_periodic_checkpoint( sim, checkpoint_period, checkpoint_dir="%s" )'
% checkpoint_dir)
script = replace_string(
script, 'restart_from_checkpoint( sim )',
'restart_from_checkpoint( sim, checkpoint_dir="%s" )' %
checkpoint_dir)
else:
checkpoint_dir = './checkpoints'
script = replace_string(script, 'track_electrons = False',
'track_electrons = True')
# Modify the script to perform N_step, enforce the random seed
# (should be the same when restarting, for exact comparison),
# and perform again N_step.
script = replace_string(
script, 'sim.step( N_step )',
'sim.step( N_step ); np.random.seed(0); sim.step( N_step )')
with open(script_filename, 'w') as f:
f.write(script)
# Launch the script from the OS
command_line = 'cd %s' % temporary_dir
if n_MPI == 1:
command_line += '; python %s' % script_name
else:
# Use only one thread for multiple MPI
command_line += '; NUMBA_NUM_THREADS=1 MKL_NUM_THREADS=1 '
command_line += 'mpirun -np %d python %s' % (n_MPI, script_name)
response = os.system(command_line)
assert response == 0
# Move diagnostics (for later comparison with the restarted simulation)
shutil.move(os.path.join(temporary_dir, 'diags'),
os.path.join(temporary_dir, 'original_diags'))
# Keep only the checkpoints from the first N_step
N_step = int(get_string('N_step = (\d+)', script))
period = int(get_string('checkpoint_period = (\d+)', script))
for i_MPI in range(n_MPI):
for step in range(N_step + period, 2 * N_step + period, period):
os.remove(
os.path.join(
temporary_dir, '%s/proc%d/hdf5/data%08d.h5' %
(checkpoint_dir, i_MPI, step)))
# Modify the script so as to enable restarts
script = replace_string(script, 'use_restart = False',
'use_restart = True')
# Redo only the last N_step
script = replace_string(
script,
'sim.step( N_step ); np.random.seed(0); sim.step( N_step )',
'np.random.seed(0); sim.step( N_step )',
)
with open(script_filename, 'w') as f:
f.write(script)
# Launch the modified script from the OS, with 2 proc
response = os.system(command_line)
assert response == 0
# Check that restarted simulation gives the same results
# as the original simulation
print('Checking restarted simulation...')
start_time = time.time()
ts1 = OpenPMDTimeSeries(os.path.join(temporary_dir, 'diags/hdf5'))
ts2 = OpenPMDTimeSeries(os.path.join(temporary_dir, 'original_diags/hdf5'))
compare_simulations(ts1, ts2, checked_fields)
end_time = time.time()
print("%.2f seconds" % (end_time - start_time))
# Check that the particle IDs are unique
print('Checking particle ids...')
start_time = time.time()
for iteration in ts1.iterations:
pid, = ts1.get_particle(["id"],
iteration=iteration,
species="electrons")
assert len(np.unique(pid)) == len(pid)
end_time = time.time()
print("%.2f seconds" % (end_time - start_time))
# Suppress the temporary directory
shutil.rmtree(temporary_dir)
def run_simulation(gamma_boost):
"""
Run a simulation with a laser pulse going through a gas jet of ionizable
N5+ atoms, and check the fraction of atoms that are in the N5+ state.
Parameters
----------
gamma_boost: float
The Lorentz factor of the frame in which the simulation is carried out.
"""
# The simulation box
zmax_lab = 20.e-6 # Length of the box along z (meters)
zmin_lab = 0.e-6
Nr = 3 # Number of gridpoints along r
rmax = 10.e-6 # Length of the box along r (meters)
Nm = 2 # Number of modes used
# The particles of the plasma
p_zmin = 5.e-6 # Position of the beginning of the plasma (meters)
p_zmax = 15.e-6
p_rmin = 0. # Minimal radial position of the plasma (meters)
p_rmax = 100.e-6 # Maximal radial position of the plasma (meters)
n_e = 1. # The plasma density is chosen very low,
# to avoid collective effects
p_nz = 2 # Number of particles per cell along z
p_nr = 1 # Number of particles per cell along r
p_nt = 4 # Number of particles per cell along theta
# Boosted frame
boost = BoostConverter(gamma_boost)
# Boost the different quantities
beta_boost = np.sqrt(1. - 1. / gamma_boost**2)
zmin, zmax = boost.static_length([zmin_lab, zmax_lab])
p_zmin, p_zmax = boost.static_length([p_zmin, p_zmax])
n_e, = boost.static_density([n_e])
# Increase the number of particles per cell in order to keep sufficient
# statistics for the evaluation of the ionization fraction
if gamma_boost > 1:
p_nz = int(2 * gamma_boost * (1 + beta_boost) * p_nz)
# The laser
a0 = 1.8 # Laser amplitude
lambda0_lab = 0.8e-6 # Laser wavelength
# Boost the laser wavelength before calculating the laser amplitude
lambda0, = boost.copropag_length([lambda0_lab], beta_object=1.)
# Duration and initial position of the laser
ctau = 10. * lambda0
z0 = -2 * ctau
# Calculate laser amplitude
omega = 2 * np.pi * c / lambda0
E0 = a0 * m_e * c * omega / e
B0 = E0 / c
def laser_func(F, x, y, z, t, amplitude, length_scale):
"""
Function that describes a Gaussian laser with infinite waist
"""
return( F + amplitude * math.cos( 2*np.pi*(z-c*t)/lambda0 ) * \
math.exp( - (z - c*t - z0)**2/ctau**2 ) )
# Resolution and number of timesteps
dz = lambda0 / 16.
dt = dz / c
Nz = int((zmax - zmin) / dz) + 1
N_step = int(
(2. * 40. * lambda0 + zmax - zmin) / (dz * (1 + beta_boost))) + 1
# Get the speed of the plasma
uz_m, = boost.longitudinal_momentum([0.])
v_plasma, = boost.velocity([0.])
# The diagnostics
diag_period = N_step - 1 # Period of the diagnostics in number of timesteps
# Initialize the simulation object
sim = Simulation(
Nz,
zmax,
Nr,
rmax,
Nm,
dt,
p_zmax,
p_zmax, # No electrons get created because we pass p_zmin=p_zmax
p_rmin,
p_rmax,
p_nz,
p_nr,
p_nt,
n_e,
zmin=zmin,
initialize_ions=False,
v_comoving=v_plasma,
use_galilean=False,
boundaries='open',
use_cuda=use_cuda)
sim.set_moving_window(v=v_plasma)
# Add the N atoms
p_zmin, p_zmax, Npz = adapt_to_grid(sim.fld.interp[0].z, p_zmin, p_zmax,
p_nz)
p_rmin, p_rmax, Npr = adapt_to_grid(sim.fld.interp[0].r, p_rmin, p_rmax,
p_nr)
sim.ptcl.append(
Particles(q=e,
m=14. * m_p,
n=0.2 * n_e,
Npz=Npz,
zmin=p_zmin,
zmax=p_zmax,
Npr=Npr,
rmin=p_rmin,
rmax=p_rmax,
Nptheta=p_nt,
dt=dt,
use_cuda=use_cuda,
uz_m=uz_m,
grid_shape=sim.fld.interp[0].Ez.shape,
continuous_injection=False))
sim.ptcl[1].make_ionizable(element='N',
level_start=0,
target_species=sim.ptcl[0])
# Add a laser to the fields of the simulation (external fields)
sim.external_fields = [
ExternalField(laser_func, 'Ex', E0, 0.),
ExternalField(laser_func, 'By', B0, 0.)
]
# Add a field diagnostic
sim.diags = [
ParticleDiagnostic(diag_period, {"ions": sim.ptcl[1]},
write_dir='tests/diags',
comm=sim.comm)
]
if gamma_boost > 1:
T_sim_lab = (2. * 40. * lambda0_lab + zmax_lab - zmin_lab) / c
sim.diags.append(
BoostedParticleDiagnostic(zmin_lab,
zmax_lab,
v_lab=0.,
dt_snapshots_lab=T_sim_lab / 2.,
Ntot_snapshots_lab=3,
gamma_boost=gamma_boost,
period=diag_period,
fldobject=sim.fld,
species={"ions": sim.ptcl[1]},
comm=sim.comm,
write_dir='tests/lab_diags'))
# Run the simulation
sim.step(N_step, use_true_rho=True)
# Check the fraction of N5+ ions at the end of the simulation
w = sim.ptcl[1].w
ioniz_level = sim.ptcl[1].ionizer.ionization_level
# Get the total number of N atoms/ions (all ionization levels together)
ntot = w.sum()
# Get the total number of N5+ ions
n_N5 = w[ioniz_level == 5].sum()
# Get the fraction of N5+ ions, and check that it is close to 0.32
N5_fraction = n_N5 / ntot
print('N5+ fraction: %.4f' % N5_fraction)
assert ((N5_fraction > 0.30) and (N5_fraction < 0.34))
# Check consistency in the regular openPMD diagnostics
ts = OpenPMDTimeSeries('./tests/diags/hdf5/')
last_iteration = ts.iterations[-1]
w, q = ts.get_particle(['w', 'charge'],
species="ions",
iteration=last_iteration)
# Check that the openPMD file contains the same number of N5+ ions
n_N5_openpmd = np.sum(w[(4.5 * e < q) & (q < 5.5 * e)])
assert np.isclose(n_N5_openpmd, n_N5)
# Remove openPMD files
shutil.rmtree('./tests/diags/')
# Check consistency of the back-transformed openPMD diagnostics
if gamma_boost > 1.:
ts = OpenPMDTimeSeries('./tests/lab_diags/hdf5/')
last_iteration = ts.iterations[-1]
w, q = ts.get_particle(['w', 'charge'],
species="ions",
iteration=last_iteration)
# Check that the openPMD file contains the same number of N5+ ions
n_N5_openpmd = np.sum(w[(4.5 * e < q) & (q < 5.5 * e)])
assert np.isclose(n_N5_openpmd, n_N5)
# Remove openPMD files
shutil.rmtree('./tests/lab_diags/')