def grid_function(grid,operators):
index_position = numpy.zeros(0,dtype=DT.IVECTOR) # The IJ location of the particle
while True:
(block, metadata) = (yield) # an array of particle datatype
blen = len(block)
particle_position = block['loc']
if index_position.shape != particle_position.shape:
index_position = numpy.zeros(particle_position.shape,numpy.int32)
else:
index_position[:]=0
grid.indexof(particle_position, out=index_position)
for target in operators:
# send: Particles, index_position, weights and metadata for each cell
target.send((block,index_position,DT.PRECISION(1.0), metadata))
gd.allocate('hit', numpy.bool8)
gd.allocate('time_hit', numpy.int32)
gd.allocate('oil_volume', DT.SPRECISION)
# make the aggregator for the results
aggregator = aggregate_coroutine(gd)
# make the gridder
gridder = grid_coroutine(grid, target=aggregator)
# make the data streamer
nblocks = 100
streamer = stream_coroutine(target=gridder)
scale = DT.PRECISION(60.0)
offset = DT.PRECISION(-10.0)
for i in xrange(nSims):
if i == 2:
continue
fname = fbase.format(i + 1)
logger.info('Reading from file: "%s"' % fname)
file_reader = OilModelDirectAccessReader(fname)
streamer.send(file_reader)
x, y = grid.meshgrid_c()
matplotlib.rcParams['xtick.direction'] = 'out'
matplotlib.rcParams['ytick.direction'] = 'out'
def grid_function(grid, operators):
# Dummy allocations - will be reallocated as needed.
index_position = numpy.zeros(
0, dtype=DT.IVECTOR) # The IJ location of the particle
prev_index_position = numpy.zeros(
0, dtype=DT.IVECTOR
) # The IJ location of the particle at the previous timestep
index_diff = numpy.zeros(
0, dtype=DT.IVECTOR) # The IJ location change between timesteps
index_sum = numpy.zeros(
0, dtype=DT.IVECTOR) # The sum of the IJ location change
nsamples_per_grid = 3
# Temporary hack:
block_number = 0
while True:
(block, metadata) = (yield) # an array of particle datatype
block_number += 1
blen = len(block)
particle_position = block['loc']
prev_particle_position = block['prev_loc']
# the number of dimensions in a position vector
ndims = particle_position.shape[1]
if index_position.shape != particle_position.shape:
index_position = numpy.zeros(particle_position.shape, numpy.int32)
else:
index_position[:] = 0
if prev_index_position.shape != prev_particle_position.shape:
prev_index_position = numpy.zeros(prev_particle_position.shape,
numpy.int32)
else:
prev_index_position[:] = 0
if index_diff.shape != prev_particle_position.shape:
index_diff = numpy.zeros(prev_particle_position.shape, numpy.int32)
else:
index_diff[:] = 0
if index_sum.shape != blen:
index_sum = numpy.zeros(blen, numpy.int32)
else:
index_sum[:] = 0
# Calculate the IJ index of each particle now and previously
grid.indexof(particle_position, out=index_position)
grid.indexof(prev_particle_position, out=prev_index_position)
#evaluate the absolute value of the index space difference
index_diff[...] = numpy.abs(index_position - prev_index_position)
index_sum[...] = numpy.sum(index_diff, axis=1)
for i in xrange(blen):
if index_sum[i] > 0:
delta_pos = particle_position[i, :] - prev_particle_position[
i, :]
samples = index_sum[i] * nsamples_per_grid
interpolated_positions = delta_pos.reshape(
1, ndims) * numpy.arange(samples).reshape(
samples, 1) / DT.PRECISION(samples)
interpolated_positions += prev_particle_position[i, :].reshape(
1, ndims)
interpolated_index_positions = grid.indexof(
interpolated_positions)
# Make the weight array the correct size
weight = DT.PRECISION(1.0 / samples) * numpy.ones((samples, ))
for target in operators:
target.send((block[i], interpolated_index_positions,
weight, metadata))
else:
# Make the weight array the correct size
weight = numpy.ones((1, ))
for target in operators:
target.send(
(block[i], prev_index_position[i, :].reshape(1, ndims),
weight, metadata))
