def make_vertices(self, model, n_steps):
"""Make vertices for the filter."""
# Get the complete matrix to be applied by the filter
out_signals = model.get_signals_from_object(self)
# Get the filter and filter routing regions
filter_region, filter_routing_region = make_filter_regions(
model.get_signals_to_object(self)[InputPort.standard],
model.dt,
True,
model.keyspaces.filter_routing_tag,
width=self.size_in)
self._routing_region = filter_routing_region
# Generate the vertices
vertices = flatinsertionlist()
for group in self.groups:
vertices.append(
group.make_vertices(out_signals, model.machine_timestep,
filter_region, filter_routing_region))
# Return the netlist specification
return netlistspec(vertices=vertices,
load_function=self.load_to_machine)
def make_vertices(self, model, n_steps):
"""Make vertices for the filter."""
# Get the complete matrix to be applied by the filter
out_signals = model.get_signals_from_object(self)
# Get the filter and filter routing regions
filter_region, filter_routing_region = make_filter_regions(
model.get_signals_to_object(self)[InputPort.standard],
model.dt, True,
model.keyspaces.filter_routing_tag,
width=self.size_in
)
self._routing_region = filter_routing_region
# Generate the vertices
vertices = flatinsertionlist()
for group in self.groups:
vertices.append(
group.make_vertices(out_signals,
model.machine_timestep,
filter_region,
filter_routing_region)
)
# Return the netlist specification
return netlistspec(vertices=vertices,
load_function=self.load_to_machine)
def test_forced_filter_width(self):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two connections with equivalent
# synapses.
# Create two keyspaces, two signals and two reception parameters with
# different synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3)
rp_b = ReceptionParameters(None, 5)
# Create the type of dictionary that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(specs, 0.001,
width=1)
# Check that the filter region is as expected
for f in filter_region.filters:
assert (f == LowpassFilter(1, False, 0.01) or
f == NoneFilter(1, False)) # noqa: E711
def make_vertices(self, model, *args, **kwargs):
"""Create vertices that will simulate the SDPTransmitter."""
# Build the system region
self._sys_region = SystemRegion(model.machine_timestep,
self.size_in, 1)
# Build the filter regions
in_sigs = model.get_signals_connections_to_object(self)
self._filter_region, self._routing_region = make_filter_regions(
in_sigs[InputPort.standard], model.dt, True,
model.keyspaces.filter_routing_tag
)
# Get the resources
resources = {
Cores: 1,
SDRAM: region_utils.sizeof_regions(
[self._sys_region, self._filter_region, self._routing_region],
None
)
}
# Create the vertex
self._vertex = Vertex(get_application("tx"), resources)
# Return the netlist specification
return netlistspec(self._vertex,
load_function=self.load_to_machine)
def make_vertices(self, model, *args, **kwargs):
"""Create vertices that will simulate the SDPTransmitter."""
# Build the system region
self._sys_region = SystemRegion(model.machine_timestep, self.size_in,
1)
# Build the filter regions
in_sigs = model.get_signals_to_object(self)[InputPort.standard]
self._filter_region, self._routing_region = make_filter_regions(
in_sigs, model.dt, True, model.keyspaces.filter_routing_tag)
# Get the resources
resources = {
Cores:
1,
SDRAM:
region_utils.sizeof_regions(
[self._sys_region, self._filter_region, self._routing_region],
None)
}
# Create the vertex
self._vertex = Vertex(self._label, get_application("tx"), resources)
# Return the netlist specification
return netlistspec(
(self._vertex, ), # Tuple is required
load_function=self.load_to_machine)
def test_forced_filter_width(self):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two connections with equivalent
# synapses.
# Create two keyspaces, two signals and two reception parameters with
# different synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3, None)
rp_b = ReceptionParameters(None, 5, None)
# Create the type of dictionary that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(specs,
0.001,
width=1)
# Check that the filter region is as expected
for f in filter_region.filters:
assert (f == LowpassFilter(1, False, 0.01)
or f == NoneFilter(1, False)) # noqa: E711
def make_vertices(self, model, n_steps): # TODO remove n_steps
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Extract all the filters from the incoming connections to build the
# filter regions.
signals_conns = model.get_signals_to_object(self)[InputPort.standard]
self.filter_region, self.filter_routing_region = make_filter_regions(
signals_conns, model.dt, True, model.keyspaces.filter_routing_tag)
# Use a matrix region to record into (slightly unpleasant)
self.recording_region = regions.MatrixRegion(
np.zeros((self.size_in, n_steps), dtype=np.uint32)
)
# This isn't partitioned, so we just compute the SDRAM requirement and
# return a new vertex.
self.system_region = SystemRegion(model.machine_timestep, self.size_in)
self.regions = [None] * 15
self.regions[0] = self.system_region
self.regions[1] = self.filter_region
self.regions[2] = self.filter_routing_region
self.regions[14] = self.recording_region # **YUCK**
resources = {
Cores: 1,
SDRAM: regions.utils.sizeof_regions(self.regions, None)
}
self.vertex = Vertex(get_application("value_sink"), resources)
# Return the spec
return netlistspec(self.vertex, self.load_to_machine,
after_simulation_function=self.after_simulation)
def test_forced_filter_width(self):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two connections with equivalent
# synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = mock.Mock(name="Signal[A]")
signal_a.keyspace, signal_a.latching = ks_a, False
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = mock.Mock(name="Signal[B]")
signal_b.keyspace, signal_b.latching = ks_b, False
conn_a = mock.Mock(name="Connection[A]")
conn_a.post_obj.size_in = 3
conn_a.synapse = nengo.Lowpass(0.01)
conn_b = mock.Mock(name="Connection[B]")
conn_b.post_obj.size_in = 5
conn_b.synapse = None
# Create the type of dictionary that is expected as input
signals_connections = {
signal_a: [conn_a],
signal_b: [conn_b],
}
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
signals_connections, 0.001, width=1
)
# Check that the filter region is as expected
for f in filter_region.filters:
assert (f == LowpassFilter(1, False, 0.01) or
f == NoneFilter(1, False)) # noqa: E711
def make_vertices(self, model, n_steps): # TODO remove n_steps
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Extract all the filters from the incoming connections to build the
# filter regions.
signals_conns = model.get_signals_to_object(self)[InputPort.standard]
self.filter_region, self.filter_routing_region = make_filter_regions(
signals_conns, model.dt, True, model.keyspaces.filter_routing_tag)
# Use a matrix region to record into (slightly unpleasant)
self.recording_region = regions.MatrixRegion(
np.zeros((self.size_in, n_steps), dtype=np.uint32))
# This isn't partitioned, so we just compute the SDRAM requirement and
# return a new vertex.
self.system_region = SystemRegion(model.machine_timestep, self.size_in)
self.regions = [None] * 15
self.regions[0] = self.system_region
self.regions[1] = self.filter_region
self.regions[2] = self.filter_routing_region
self.regions[14] = self.recording_region # **YUCK**
resources = {
Cores: 1,
SDRAM: regions.utils.sizeof_regions(self.regions, None)
}
self.vertex = Vertex(get_application("value_sink"), resources)
# Return the spec
return netlistspec(self.vertex,
self.load_to_machine,
after_simulation_function=self.after_simulation)
def make_vertices(self, model, n_steps): # TODO remove n_steps
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Extract all the filters from the incoming connections to build the
# filter regions.
signals_conns = model.get_signals_to_object(self)[InputPort.standard]
filter_region, filter_routing_region = make_filter_regions(
signals_conns, model.dt, True, model.keyspaces.filter_routing_tag)
self._routing_region = filter_routing_region
# Make sufficient vertices to ensure that each has a size_in of less
# than max_width.
n_vertices = (
(self.size_in // self.max_width) +
(1 if self.size_in % self.max_width else 0)
)
self.vertices = tuple(
ValueSinkVertex(model.machine_timestep, n_steps, sl, filter_region,
filter_routing_region) for sl in
divide_slice(slice(0, self.size_in), n_vertices)
)
# Return the spec
return netlistspec(self.vertices, self.load_to_machine,
after_simulation_function=self.after_simulation)
def test_equivalent_filters(self, minimise):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two connections with equivalent
# synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = mock.Mock(name="Signal[A]")
signal_a.keyspace, signal_a.latching = ks_a, False
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = mock.Mock(name="Signal[B]")
signal_b.keyspace, signal_b.latching = ks_b, False
conn_a = mock.Mock(name="Connection[A]")
conn_a.post_obj.size_in = 3
conn_a.synapse = nengo.Lowpass(0.01)
conn_b = mock.Mock(name="Connection[B]")
conn_b.post_obj.size_in = 3
conn_b.synapse = nengo.Lowpass(0.01)
# Create the type of dictionary that is expected as input
signals_connections = {
signal_a: [conn_a],
signal_b: [conn_b],
}
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
signals_connections, 0.001,
minimise=minimise,
filter_routing_tag="spam",
index_field="eggs"
)
# Check that the filter region is as expected
assert filter_region.dt == 0.001
if minimise:
assert len(filter_region.filters) == 1
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
else:
assert len(filter_region.filters) == 2
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
assert filter_region.filters[1] == LowpassFilter(3, False, 0.01)
# Check that the routing region is as expected
assert routing_region.filter_routing_tag == "spam"
assert routing_region.index_field == "eggs"
if minimise:
assert (ks_a, 0) in routing_region.keyspace_routes
assert (ks_b, 0) in routing_region.keyspace_routes
else:
if (ks_a, 0) in routing_region.keyspace_routes:
assert (ks_b, 1) in routing_region.keyspace_routes
else:
assert (ks_b, 0) in routing_region.keyspace_routes
def test_equivalent_filters(self, minimise):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signal parameters and two reception
# parameters with equivalent synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3, None)
rp_b = ReceptionParameters(nengo.Lowpass(0.01), 3, None)
# Create the data structure that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
specs,
0.001,
minimise=minimise,
filter_routing_tag="spam",
index_field="eggs")
# Check that the filter region is as expected
assert filter_region.dt == 0.001
if minimise:
assert len(filter_region.filters) == 1
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
else:
assert len(filter_region.filters) == 2
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
assert filter_region.filters[1] == LowpassFilter(3, False, 0.01)
# Check that the routing region is as expected
assert routing_region.filter_routing_tag == "spam"
assert routing_region.index_field == "eggs"
if minimise:
assert (signal_a, 0) in routing_region.signal_routes
assert (signal_b, 0) in routing_region.signal_routes
else:
if (signal_a, 0) in routing_region.signal_routes:
assert (signal_b, 1) in routing_region.signal_routes
else:
assert (signal_b, 0) in routing_region.signal_routes
def test_equivalent_filters(self, minimise):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signal parameters and two reception
# parameters with equivalent synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3)
rp_b = ReceptionParameters(nengo.Lowpass(0.01), 3)
# Create the data structure that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
specs, 0.001, minimise=minimise,
filter_routing_tag="spam",
index_field="eggs"
)
# Check that the filter region is as expected
assert filter_region.dt == 0.001
if minimise:
assert len(filter_region.filters) == 1
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
else:
assert len(filter_region.filters) == 2
assert filter_region.filters[0] == LowpassFilter(3, False, 0.01)
assert filter_region.filters[1] == LowpassFilter(3, False, 0.01)
# Check that the routing region is as expected
assert routing_region.filter_routing_tag == "spam"
assert routing_region.index_field == "eggs"
if minimise:
assert (ks_a, 0) in routing_region.keyspace_routes
assert (ks_b, 0) in routing_region.keyspace_routes
else:
if (ks_a, 0) in routing_region.keyspace_routes:
assert (ks_b, 1) in routing_region.keyspace_routes
else:
assert (ks_b, 0) in routing_region.keyspace_routes
def test_different_filters(self):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two reception parameters with
# different synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3)
rp_b = ReceptionParameters(None, 3)
# Create the type of dictionary that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
specs, 0.001,
minimise=True, # Shouldn't achieve anything
filter_routing_tag="spam",
index_field="eggs"
)
# Check that the filter region is as expected
assert filter_region.dt == 0.001
assert len(filter_region.filters) == 2
for f in filter_region.filters:
assert (f == LowpassFilter(3, False, 0.01) or
f == NoneFilter(3, False)) # noqa: E711
# Check that the routing region is as expected
assert routing_region.filter_routing_tag == "spam"
assert routing_region.index_field == "eggs"
if (ks_a, 0) in routing_region.keyspace_routes:
assert (ks_b, 1) in routing_region.keyspace_routes
else:
assert (ks_b, 0) in routing_region.keyspace_routes
def test_different_filters(self):
"""Test construction of filter regions from signals and keyspaces."""
# Create two keyspaces, two signals and two reception parameters with
# different synapses.
ks_a = mock.Mock(name="Keyspace[A]")
signal_a = SignalParameters(keyspace=ks_a, latching=False)
ks_b = mock.Mock(name="Keyspace[B]")
signal_b = SignalParameters(keyspace=ks_b, latching=False)
rp_a = ReceptionParameters(nengo.Lowpass(0.01), 3, None)
rp_b = ReceptionParameters(None, 3, None)
# Create the type of dictionary that is expected as input
specs = [
ReceptionSpec(signal_a, rp_a),
ReceptionSpec(signal_b, rp_b),
]
# Create the regions, with minimisation
filter_region, routing_region = make_filter_regions(
specs,
0.001,
minimise=True, # Shouldn't achieve anything
filter_routing_tag="spam",
index_field="eggs")
# Check that the filter region is as expected
assert filter_region.dt == 0.001
assert len(filter_region.filters) == 2
for f in filter_region.filters:
assert (f == LowpassFilter(3, False, 0.01)
or f == NoneFilter(3, False)) # noqa: E711
# Check that the routing region is as expected
assert routing_region.filter_routing_tag == "spam"
assert routing_region.index_field == "eggs"
if (signal_a, 0) in routing_region.signal_routes:
assert (signal_b, 1) in routing_region.signal_routes
else:
assert (signal_b, 0) in routing_region.signal_routes
def make_vertices(self, model, n_steps): # TODO remove n_steps
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Extract all the filters from the incoming connections to build the
# filter regions.
signals_conns = model.get_signals_to_object(self)[InputPort.standard]
filter_region, filter_routing_region = make_filter_regions(
signals_conns, model.dt, True, model.keyspaces.filter_routing_tag)
self._routing_region = filter_routing_region
# Make sufficient vertices to ensure that each has a size_in of less
# than max_width.
n_vertices = ((self.size_in // self.max_width) +
(1 if self.size_in % self.max_width else 0))
self.vertices = tuple(
ValueSinkVertex(model.machine_timestep, n_steps, sl, filter_region,
filter_routing_region)
for sl in divide_slice(slice(0, self.size_in), n_vertices))
# Return the spec
return netlistspec(self.vertices,
self.load_to_machine,
after_simulation_function=self.after_simulation)
def make_vertices(self, model, n_steps):
"""Make vertices for the filter."""
# Get the outgoing transforms and keys
sigs = model.get_signals_from_object(self)
if OutputPort.standard in sigs:
outgoing = sigs[OutputPort.standard]
transform, output_keys, sigs_pars_slices = \
get_transforms_and_keys(outgoing)
else:
transform = np.array([[]])
output_keys = list()
sigs_pars_slices = list()
size_out = len(output_keys)
# Calculate how many cores and chips to use.
if self.n_cores_per_chip is None or self.n_chips is None:
# The number of cores is largely a function of the input size, we
# try to ensure that each core is receiving a max of 32 packets per
# timestep.
n_cores_per_chip = int(min(16, np.ceil(self.size_in / 32.0)))
# The number of chips is now determined by the size in (columns in
# the transform matrix), the size out (rows in the transform
# matrix) and the number of cores per chip.
n_chips = self.n_chips or 1
n_cores = n_chips * n_cores_per_chip
while True:
rows_per_core = int(np.ceil(float(size_out) /
(n_cores * n_chips)))
load_per_core = rows_per_core * self.size_in
# The 8,000 limits the number of columns in each row that we
# need to process. This is a heuristic.
if load_per_core <= 8000 or n_chips > 9:
# The load per core is acceptable or we're using way too
# many chips
break
if n_cores < 16:
# Increase the number of cores per chip if we can
n_cores += 1
else:
# Otherwise increase the number of chips
n_chips += 1
# Store the result
self.n_cores_per_chip = n_cores
self.n_chips = n_chips
# Slice the input space into the given number of subspaces, this is
# repeated on each chip.
input_slices = list(divide_slice(slice(0, self.size_in),
self.n_cores_per_chip))
# Slice the output space into the given number of subspaces, this is
# sliced across all of the chips.
output_slices = divide_slice(slice(0, size_out),
self.n_cores_per_chip * self.n_chips)
# Construct the output keys and transform regions; the output keys and
# sliced, and the transform is sliced by rows.
self.output_keys_region = regions.KeyspacesRegion(
output_keys, fields=[regions.KeyField({'cluster': 'cluster'})],
partitioned_by_atom=True
)
self.transform_region = regions.MatrixRegion(
np_to_fix(transform),
sliced_dimension=regions.MatrixPartitioning.rows
)
# Construct the system region
self.system_region = SystemRegion(self.size_in, model.machine_timestep)
# Get the incoming filters
incoming = model.get_signals_to_object(self)
self.filters_region, self.routing_region = make_filter_regions(
incoming[InputPort.standard], model.dt, True,
model.keyspaces.filter_routing_tag, width=self.size_in
)
# Make the vertices and constraints
iter_output_slices = iter(output_slices)
cons = list() # List of constraints
# For each chip that we'll be using
for _ in range(self.n_chips):
chip_vertices = list()
# Each core is given an input slice and an output slice. The same
# set of input slices is used per chip, but we iterate through the
# whole list of output slices.
for in_slice, out_slice in zip(input_slices,
iter_output_slices):
# Determine the amount of SDRAM required (the 24 additional
# bytes are for the application pointer table). We also
# include this cores contribution to a shared SDRAM vector.
sdram = (24 + 4*(in_slice.stop - in_slice.start) +
self.system_region.sizeof() +
self.filters_region.sizeof_padded() +
self.routing_region.sizeof_padded() +
self.output_keys_region.sizeof_padded(out_slice) +
self.transform_region.sizeof_padded(out_slice))
# Create the vertex and include in the list of vertices
v = ParallelFilterSlice(in_slice, out_slice,
{Cores: 1, SDRAM: sdram},
sigs_pars_slices)
chip_vertices.append(v)
self.vertices.append(v)
# Create a constraint which will force all of the vertices to exist
# of the same chip.
cons.append(SameChipConstraint(chip_vertices))
# Return the spec
return netlistspec(self.vertices, self.load_to_machine,
constraints=cons)
def make_vertices(self, model, n_steps):
"""Make vertices for the filter."""
# Get the outgoing transforms and keys
sigs = model.get_signals_from_object(self)
if OutputPort.standard in sigs:
outgoing = sigs[OutputPort.standard]
transform, output_keys, sigs_pars_slices = \
get_transforms_and_keys(outgoing)
else:
transform = np.array([[]])
output_keys = list()
sigs_pars_slices = list()
size_out = len(output_keys)
# Calculate how many cores and chips to use.
if self.n_cores_per_chip is None or self.n_chips is None:
# The number of cores is largely a function of the input size, we
# try to ensure that each core is receiving a max of 32 packets per
# timestep.
n_cores_per_chip = int(min(16, np.ceil(self.size_in / 32.0)))
# The number of chips is now determined by the size in (columns in
# the transform matrix), the size out (rows in the transform
# matrix) and the number of cores per chip.
n_chips = self.n_chips or 1
n_cores = n_chips * n_cores_per_chip
while True:
rows_per_core = int(
np.ceil(float(size_out) / (n_cores * n_chips)))
load_per_core = rows_per_core * self.size_in
# The 8,000 limits the number of columns in each row that we
# need to process. This is a heuristic.
if load_per_core <= 8000 or n_chips > 9:
# The load per core is acceptable or we're using way too
# many chips
break
if n_cores < 16:
# Increase the number of cores per chip if we can
n_cores += 1
else:
# Otherwise increase the number of chips
n_chips += 1
# Store the result
self.n_cores_per_chip = n_cores
self.n_chips = n_chips
# Slice the input space into the given number of subspaces, this is
# repeated on each chip.
input_slices = list(
divide_slice(slice(0, self.size_in), self.n_cores_per_chip))
# Slice the output space into the given number of subspaces, this is
# sliced across all of the chips.
output_slices = divide_slice(slice(0, size_out),
self.n_cores_per_chip * self.n_chips)
# Construct the output keys and transform regions; the output keys and
# sliced, and the transform is sliced by rows.
self.output_keys_region = regions.KeyspacesRegion(
output_keys,
fields=[regions.KeyField({'cluster': 'cluster'})],
partitioned_by_atom=True)
self.transform_region = regions.MatrixRegion(
np_to_fix(transform),
sliced_dimension=regions.MatrixPartitioning.rows)
# Construct the system region
self.system_region = SystemRegion(self.size_in, model.machine_timestep)
# Get the incoming filters
incoming = model.get_signals_to_object(self)
self.filters_region, self.routing_region = make_filter_regions(
incoming[InputPort.standard],
model.dt,
True,
model.keyspaces.filter_routing_tag,
width=self.size_in)
# Make the vertices and constraints
iter_output_slices = iter(output_slices)
cons = list() # List of constraints
# For each chip that we'll be using
for _ in range(self.n_chips):
chip_vertices = list()
# Each core is given an input slice and an output slice. The same
# set of input slices is used per chip, but we iterate through the
# whole list of output slices.
for in_slice, out_slice in zip(input_slices, iter_output_slices):
# Determine the amount of SDRAM required (the 24 additional
# bytes are for the application pointer table). We also
# include this cores contribution to a shared SDRAM vector.
sdram = (24 + 4 * (in_slice.stop - in_slice.start) +
self.system_region.sizeof() +
self.filters_region.sizeof_padded() +
self.routing_region.sizeof_padded() +
self.output_keys_region.sizeof_padded(out_slice) +
self.transform_region.sizeof_padded(out_slice))
# Create the vertex and include in the list of vertices
v = ParallelFilterSlice(in_slice, out_slice, {
Cores: 1,
SDRAM: sdram
}, sigs_pars_slices)
chip_vertices.append(v)
self.vertices.append(v)
# Create a constraint which will force all of the vertices to exist
# of the same chip.
cons.append(SameChipConstraint(chip_vertices))
# Return the spec
return netlistspec(self.vertices,
self.load_to_machine,
constraints=cons)
def make_vertices(self, model, n_steps):
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Build encoders, gain and bias regions
params = model.params[self.ensemble]
ens_regions = dict()
# Convert the encoders combined with the gain to S1615 before creating
# the region.
encoders_with_gain = params.scaled_encoders
ens_regions[EnsembleRegions.encoders] = regions.MatrixRegion(
tp.np_to_fix(encoders_with_gain),
sliced_dimension=regions.MatrixPartitioning.rows)
# Combine the direct input with the bias before converting to S1615 and
# creating the region.
bias_with_di = params.bias + np.dot(encoders_with_gain,
self.direct_input)
assert bias_with_di.ndim == 1
ens_regions[EnsembleRegions.bias] = regions.MatrixRegion(
tp.np_to_fix(bias_with_di),
sliced_dimension=regions.MatrixPartitioning.rows)
# Convert the gains to S1615 before creating the region
ens_regions[EnsembleRegions.gain] = regions.MatrixRegion(
tp.np_to_fix(params.gain),
sliced_dimension=regions.MatrixPartitioning.rows)
# Extract all the filters from the incoming connections
incoming = model.get_signals_to_object(self)
(ens_regions[EnsembleRegions.input_filters],
ens_regions[EnsembleRegions.input_routing]) = make_filter_regions(
incoming[InputPort.standard],
model.dt,
True,
model.keyspaces.filter_routing_tag,
width=self.ensemble.size_in)
(ens_regions[EnsembleRegions.inhibition_filters],
ens_regions[EnsembleRegions.inhibition_routing]) = \
make_filter_regions(
incoming[EnsembleInputPort.global_inhibition], model.dt, True,
model.keyspaces.filter_routing_tag, width=1
)
# Extract all the decoders for the outgoing connections and build the
# regions for the decoders and the regions for the output keys.
outgoing = model.get_signals_from_object(self)
if OutputPort.standard in outgoing:
decoders, output_keys = \
get_decoders_and_keys(outgoing[OutputPort.standard], True)
else:
decoders = np.array([])
output_keys = list()
size_out = decoders.shape[0]
ens_regions[EnsembleRegions.decoders] = regions.MatrixRegion(
tp.np_to_fix(decoders / model.dt),
sliced_dimension=regions.MatrixPartitioning.rows)
ens_regions[EnsembleRegions.keys] = regions.KeyspacesRegion(
output_keys,
fields=[regions.KeyField({'cluster': 'cluster'})],
partitioned_by_atom=True)
# The population length region stores information about groups of
# co-operating cores.
ens_regions[EnsembleRegions.population_length] = \
regions.ListRegion("I")
# The ensemble region contains basic information about the ensemble
ens_regions[EnsembleRegions.ensemble] = EnsembleRegion(
model.machine_timestep, self.ensemble.size_in)
# The neuron region contains information specific to the neuron type
ens_regions[EnsembleRegions.neuron] = LIFRegion(
model.dt, self.ensemble.neuron_type.tau_rc,
self.ensemble.neuron_type.tau_ref)
# Manage profiling
n_profiler_samples = 0
self.profiled = getconfig(model.config, self.ensemble, "profile",
False)
if self.profiled:
# Try and get number of samples from config
n_profiler_samples = getconfig(model.config, self.ensemble,
"profile_num_samples")
# If it's not specified, calculate sensible default
if n_profiler_samples is None:
n_profiler_samples = (len(EnsembleSlice.profiler_tag_names) *
n_steps * 2)
# Create profiler region
ens_regions[EnsembleRegions.profiler] = regions.Profiler(
n_profiler_samples)
ens_regions[EnsembleRegions.ensemble].n_profiler_samples = \
n_profiler_samples
# Manage probes
for probe in self.local_probes:
if probe.attr in ("output", "spikes"):
self.record_spikes = True
elif probe.attr == "voltage":
self.record_voltages = True
else:
raise NotImplementedError(
"Cannot probe {} on Ensembles".format(probe.attr))
# Set the flags
ens_regions[EnsembleRegions.ensemble].record_spikes = \
self.record_spikes
ens_regions[EnsembleRegions.ensemble].record_voltages = \
self.record_voltages
# Create the probe recording regions
ens_regions[EnsembleRegions.spikes] = regions.SpikeRecordingRegion(
n_steps if self.record_spikes else 0)
ens_regions[EnsembleRegions.voltages] = regions.VoltageRecordingRegion(
n_steps if self.record_voltages else 0)
# Create constraints against which to partition, initially assume that
# we can devote 16 cores to every problem.
sdram_constraint = partition.Constraint(128 * 2**20,
0.9) # 90% of 128MiB
dtcm_constraint = partition.Constraint(16 * 64 * 2**10,
0.9) # 90% of 16 cores DTCM
# The number of cycles available is 200MHz * the machine timestep; or
# 200 * the machine timestep in microseconds.
cycles = 200 * model.machine_timestep
cpu_constraint = partition.Constraint(cycles * 16,
0.8) # 80% of 16 cores compute
# Form the constraints dictionary
def _make_constraint(f, size_in, size_out, **kwargs):
"""Wrap a usage computation method to work with the partitioner."""
def f_(vertex_slice):
# Calculate the number of neurons
n_neurons = vertex_slice.stop - vertex_slice.start
# Call the original method
return f(size_in, size_out, n_neurons, **kwargs)
return f_
partition_constraints = {
sdram_constraint:
_make_constraint(_lif_sdram_usage, self.ensemble.size_in,
size_out),
dtcm_constraint:
_make_constraint(_lif_dtcm_usage, self.ensemble.size_in, size_out),
cpu_constraint:
_make_constraint(_lif_cpu_usage, self.ensemble.size_in, size_out),
}
# Partition the ensemble to create clusters of co-operating cores
self.clusters = list()
vertices = list()
constraints = list()
for sl in partition.partition(slice(0, self.ensemble.n_neurons),
partition_constraints):
# For each slice we create a cluster of co-operating cores. We
# instantiate the cluster and then ask it to produce vertices which
# will be added to the netlist.
cluster = EnsembleCluster(sl, self.ensemble.size_in, size_out,
ens_regions)
self.clusters.append(cluster)
# Get the vertices for the cluster
cluster_vertices = cluster.make_vertices(cycles)
vertices.extend(cluster_vertices)
# Create a constraint which forces these vertices to be present on
# the same chip
constraints.append(SameChipConstraint(cluster_vertices))
# Return the vertices and callback methods
return netlistspec(vertices,
self.load_to_machine,
after_simulation_function=self.after_simulation,
constraints=constraints)
def make_vertices(self, model, n_steps):
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Build encoders, gain and bias regions
params = model.params[self.ensemble]
ens_regions = dict()
# Convert the encoders combined with the gain to S1615 before creating
# the region.
encoders_with_gain = params.scaled_encoders
ens_regions[EnsembleRegions.encoders] = regions.MatrixRegion(
tp.np_to_fix(encoders_with_gain),
sliced_dimension=regions.MatrixPartitioning.rows)
# Combine the direct input with the bias before converting to S1615 and
# creating the region.
bias_with_di = params.bias + np.dot(encoders_with_gain,
self.direct_input)
assert bias_with_di.ndim == 1
ens_regions[EnsembleRegions.bias] = regions.MatrixRegion(
tp.np_to_fix(bias_with_di),
sliced_dimension=regions.MatrixPartitioning.rows)
# Convert the gains to S1615 before creating the region
ens_regions[EnsembleRegions.gain] = regions.MatrixRegion(
tp.np_to_fix(params.gain),
sliced_dimension=regions.MatrixPartitioning.rows)
# Extract all the filters from the incoming connections
incoming = model.get_signals_to_object(self)
(ens_regions[EnsembleRegions.input_filters],
ens_regions[EnsembleRegions.input_routing]) = make_filter_regions(
incoming[InputPort.standard], model.dt, True,
model.keyspaces.filter_routing_tag,
width=self.ensemble.size_in
)
(ens_regions[EnsembleRegions.inhibition_filters],
ens_regions[EnsembleRegions.inhibition_routing]) = \
make_filter_regions(
incoming[EnsembleInputPort.global_inhibition], model.dt, True,
model.keyspaces.filter_routing_tag, width=1
)
# Extract all the decoders for the outgoing connections and build the
# regions for the decoders and the regions for the output keys.
outgoing = model.get_signals_from_object(self)
if OutputPort.standard in outgoing:
decoders, output_keys = \
get_decoders_and_keys(outgoing[OutputPort.standard], True)
else:
decoders = np.array([])
output_keys = list()
size_out = decoders.shape[0]
ens_regions[EnsembleRegions.decoders] = regions.MatrixRegion(
tp.np_to_fix(decoders / model.dt),
sliced_dimension=regions.MatrixPartitioning.rows)
ens_regions[EnsembleRegions.keys] = regions.KeyspacesRegion(
output_keys,
fields=[regions.KeyField({'cluster': 'cluster'})],
partitioned_by_atom=True
)
# The population length region stores information about groups of
# co-operating cores.
ens_regions[EnsembleRegions.population_length] = \
regions.ListRegion("I")
# The ensemble region contains basic information about the ensemble
ens_regions[EnsembleRegions.ensemble] = EnsembleRegion(
model.machine_timestep, self.ensemble.size_in)
# The neuron region contains information specific to the neuron type
ens_regions[EnsembleRegions.neuron] = LIFRegion(
model.dt, self.ensemble.neuron_type.tau_rc,
self.ensemble.neuron_type.tau_ref
)
# Manage profiling
n_profiler_samples = 0
self.profiled = getconfig(model.config, self.ensemble, "profile",
False)
if self.profiled:
# Try and get number of samples from config
n_profiler_samples = getconfig(model.config, self.ensemble,
"profile_num_samples")
# If it's not specified, calculate sensible default
if n_profiler_samples is None:
n_profiler_samples = (len(EnsembleSlice.profiler_tag_names) *
n_steps * 2)
# Create profiler region
ens_regions[EnsembleRegions.profiler] = regions.Profiler(
n_profiler_samples)
ens_regions[EnsembleRegions.ensemble].n_profiler_samples = \
n_profiler_samples
# Manage probes
for probe in self.local_probes:
if probe.attr in ("output", "spikes"):
self.record_spikes = True
elif probe.attr == "voltage":
self.record_voltages = True
else:
raise NotImplementedError(
"Cannot probe {} on Ensembles".format(probe.attr)
)
# Set the flags
ens_regions[EnsembleRegions.ensemble].record_spikes = \
self.record_spikes
ens_regions[EnsembleRegions.ensemble].record_voltages = \
self.record_voltages
# Create the probe recording regions
ens_regions[EnsembleRegions.spikes] = regions.SpikeRecordingRegion(
n_steps if self.record_spikes else 0)
ens_regions[EnsembleRegions.voltages] = regions.VoltageRecordingRegion(
n_steps if self.record_voltages else 0)
# Create constraints against which to partition, initially assume that
# we can devote 16 cores to every problem.
sdram_constraint = partition.Constraint(128 * 2**20,
0.9) # 90% of 128MiB
dtcm_constraint = partition.Constraint(16 * 64 * 2**10,
0.9) # 90% of 16 cores DTCM
# The number of cycles available is 200MHz * the machine timestep; or
# 200 * the machine timestep in microseconds.
cycles = 200 * model.machine_timestep
cpu_constraint = partition.Constraint(cycles * 16,
0.8) # 80% of 16 cores compute
# Form the constraints dictionary
def _make_constraint(f, size_in, size_out, **kwargs):
"""Wrap a usage computation method to work with the partitioner."""
def f_(vertex_slice):
# Calculate the number of neurons
n_neurons = vertex_slice.stop - vertex_slice.start
# Call the original method
return f(size_in, size_out, n_neurons, **kwargs)
return f_
partition_constraints = {
sdram_constraint: _make_constraint(_lif_sdram_usage,
self.ensemble.size_in,
size_out),
dtcm_constraint: _make_constraint(_lif_dtcm_usage,
self.ensemble.size_in, size_out),
cpu_constraint: _make_constraint(_lif_cpu_usage,
self.ensemble.size_in, size_out),
}
# Partition the ensemble to create clusters of co-operating cores
self.clusters = list()
vertices = list()
constraints = list()
for sl in partition.partition(slice(0, self.ensemble.n_neurons),
partition_constraints):
# For each slice we create a cluster of co-operating cores. We
# instantiate the cluster and then ask it to produce vertices which
# will be added to the netlist.
cluster = EnsembleCluster(sl, self.ensemble.size_in, size_out,
ens_regions)
self.clusters.append(cluster)
# Get the vertices for the cluster
cluster_vertices = cluster.make_vertices(cycles)
vertices.extend(cluster_vertices)
# Create a constraint which forces these vertices to be present on
# the same chip
constraints.append(SameChipConstraint(cluster_vertices))
# Return the vertices and callback methods
return netlistspec(vertices, self.load_to_machine,
after_simulation_function=self.after_simulation,
constraints=constraints)
def make_vertices(self, model, n_steps): # TODO remove n_steps
"""Construct the data which can be loaded into the memory of a
SpiNNaker machine.
"""
# Build encoders, gain and bias regions
params = model.params[self.ensemble]
# Convert the encoders combined with the gain to S1615 before creating
# the region.
encoders_with_gain = params.scaled_encoders
self.encoders_region = regions.MatrixRegion(
tp.np_to_fix(encoders_with_gain),
sliced_dimension=regions.MatrixPartitioning.rows
)
# Combine the direct input with the bias before converting to S1615 and
# creating the region.
bias_with_di = params.bias + np.dot(encoders_with_gain,
self.direct_input)
assert bias_with_di.ndim == 1
self.bias_region = regions.MatrixRegion(
tp.np_to_fix(bias_with_di),
sliced_dimension=regions.MatrixPartitioning.rows
)
# Convert the gains to S1615 before creating the region
self.gain_region = regions.MatrixRegion(
tp.np_to_fix(params.gain),
sliced_dimension=regions.MatrixPartitioning.rows
)
# Extract all the filters from the incoming connections
incoming = model.get_signals_connections_to_object(self)
self.input_filters, self.input_filter_routing = make_filter_regions(
incoming[InputPort.standard], model.dt, True,
model.keyspaces.filter_routing_tag, width=self.ensemble.size_in
)
self.inhib_filters, self.inhib_filter_routing = make_filter_regions(
incoming[EnsembleInputPort.global_inhibition], model.dt, True,
model.keyspaces.filter_routing_tag, width=1
)
self.mod_filters, self.mod_filter_routing = make_filter_regions(
{}, model.dt, True, model.keyspaces.filter_routing_tag
)
# Extract all the decoders for the outgoing connections and build the
# regions for the decoders and the regions for the output keys.
outgoing = model.get_signals_connections_from_object(self)
decoders, output_keys = \
get_decoders_and_keys(model, outgoing[OutputPort.standard], True)
size_out = decoders.shape[1]
# TODO: Include learnt decoders
self.pes_region = PESRegion()
self.decoders_region = regions.MatrixRegion(
tp.np_to_fix(decoders / model.dt),
sliced_dimension=regions.MatrixPartitioning.rows
)
self.output_keys_region = regions.KeyspacesRegion(
output_keys, fields=[regions.KeyField({'cluster': 'cluster'})]
)
# Create the recording regions for locally situated probes
self.spike_region = None
self.probe_spikes = False
self.voltage_region = None
self.probe_voltages = False
for probe in self.local_probes:
# For each probe determine which regions and flags should be set
if probe.attr in ("output", "spikes"):
# If spikes are being probed then ensure that the flag is set
# and a region exists.
if not self.probe_spikes:
self.spike_region = SpikeRegion(n_steps)
self.probe_spikes = True
elif probe.attr in ("voltage"):
# If voltages are being probed then ensure that the flag is set
# and a region exists.
if not self.probe_voltages:
self.voltage_region = VoltageRegion(n_steps)
self.probe_voltages = True
# If profiling is enabled
num_profiler_samples = 0
if getconfig(model.config, self.ensemble, "profile", False):
# Try and get number of samples from config
num_profiler_samples = getconfig(model.config, self.ensemble,
"profile_num_samples")
# If it's not specified, calculate sensible default
if num_profiler_samples is None:
num_profiler_samples =\
len(EnsembleLIF.profiler_tag_names) * n_steps * 2
# Create profiler region
self.profiler_region = regions.Profiler(num_profiler_samples)
# Create the regions list
self.regions = [
SystemRegion(self.ensemble.size_in,
size_out,
model.machine_timestep,
self.ensemble.neuron_type.tau_ref,
self.ensemble.neuron_type.tau_rc,
model.dt,
self.probe_spikes,
self.probe_voltages,
num_profiler_samples
),
self.bias_region,
self.encoders_region,
self.decoders_region,
self.output_keys_region,
self.input_filters,
self.input_filter_routing,
self.inhib_filters,
self.inhib_filter_routing,
self.gain_region,
self.mod_filters,
self.mod_filter_routing,
self.pes_region,
self.profiler_region,
self.spike_region,
self.voltage_region,
]
# Partition the ensemble and get a list of vertices to load to the
# machine. We can expect to be DTCM or CPU bound, so the SDRAM bound
# can be quite lax to allow for lots of data probing.
# TODO: Include other DTCM usage
def cpu_usage(sl):
"""Calculate the CPU usage (in cycles) based on the number of
neurons and the size_in and size_out of the ensemble.
The equation and coefficients are taken from: "An Efficient
SpiNNaker Implementation of the NEF", Mundy, Knight, Stewart and
Furber [IJCNN 2015]
"""
n_neurons = (sl.stop - sl.start)
return (245 + 43*self.ensemble.size_in + 100 + 702*size_out +
188 + 69*n_neurons + 13*n_neurons*self.ensemble.size_in)
self.vertices = list()
sdram_constraint = partition.Constraint(8*2**20) # Max 8MiB
dtcm_constraint = partition.Constraint(64*2**10, .75) # 75% of 64KiB
cpu_constraint = partition.Constraint(200000, .8) # 80% of 200k cycles
constraints = {
sdram_constraint: lambda s: regions.utils.sizeof_regions(
self.regions, s),
# **HACK** don't include last three regions in DTCM estimate
# (profiler and spike recording)
dtcm_constraint: lambda s: regions.utils.sizeof_regions(
self.regions[:-3], s) + 5*(s.stop - s.start),
cpu_constraint: cpu_usage,
}
app_name = (
"ensemble_profiled" if num_profiler_samples > 0
else "ensemble"
)
for sl in partition.partition(slice(0, self.ensemble.n_neurons),
constraints):
resources = {
Cores: 1,
SDRAM: regions.utils.sizeof_regions(self.regions, sl),
}
vsl = VertexSlice(sl, get_application(app_name), resources)
self.vertices.append(vsl)
# Return the vertices and callback methods
return netlistspec(self.vertices, self.load_to_machine,
after_simulation_function=self.after_simulation)
