Simulator User Interface¶
- pypdevs.simulator.loadCheckpoint(name)[source]
Load a previously created simulation from a saved checkpoint.
- Parameters
name – the name of the model to provide some distinction between different simulation models
- Returns
either None if no recovery was possible, or the Simulator object after simulation
- class pypdevs.simulator.Simulator(model)[source]
Associates a hierarchical DEVS model with the simulation engine.
- __init__(model)[source]
Constructor of the simulator.
- Parameters
model – a valid model (created with the provided functions)
- simulate()[source]
Start simulation with the previously set options. Can be reran afterwards to continue the simulation of the model (or reinit it first and restart simulation), possibly after altering some aspects of the model with the provided methods.
All configuration of the simulator is forwarded to the Simulator Configuration class. These methods can simply be invoked on a Simulator object.
- class pypdevs.simconfig.SimulatorConfiguration(sim)[source]
All necessary simulator configuration options are provided. The necessary checks will be made and the simulator will be adapted accordingly.
- __init__(sim)[source]
Constructor
- Parameters
sim – the simulator to alter with actions on this configurator
- registerState(variable, model)[source]
Registers the state of a certain model to an attribute of the simulator AFTER simulation has finished
- Parameters
variable – name of the attribute to assign to
model – the AtomicDEVS model or its model id to fetch the state from
- setActivityRelocatorBasicBoundary(swappiness)[source]
Sets the use of the activity relocator called ‘Basic Boundary’.
- Parameters
swappiness – how big the deviation from the average should be before scheduling relocations
- setActivityRelocatorCustom(filename, classname, *args)[source]
Sets the use of a custom relocator
- Parameters
filename – filename containing the relocator
classname – classname of the relocator
args – all other args are passed to the constructor
- setActivityTrackingVisualisation(visualize, x=0, y=0)[source]
Set the simulation to visualize the results from activity tracking. An x and y parameter can be given to visualize it in a cell style.
- Parameters
visualize – whether or not to visualize it
x – the horizontal size of the grid (optional)
y – the vertical size of the grid (optional)
- setAllowLocalReinit(allowed=True)[source]
Allow a model to be reinitialized in local simulation. This is not the case by default, as it would be required to save a copy of the model in memory during setup. Generating such a copy can be time consuming and the additional memory consumption could be unacceptable. Distributed simulation is unaffected, since this always requires the creation of a copy. If this is False and reinitialisation is done in a local simulation, an exception will be thrown.
Warning
The state that is accessible after the simulation will NOT be updated if this configuration parameter is used. If you want to have fully up to date states, you should also set the setFetchAllAfterSimulation() configuration parameter.
- Parameters
allowed – whether or not to allow reinitialization
- setAutoAllocator()[source]
Sets the use of an initial allocator that simply distributes the root models. This is a static allocator, meaning that no event activity will be generated.
- setCell(x_size=None, y_size=None, cell_file='celltrace', multifile=False)[source]
Sets the cell tracing flag of the simulation
- Parameters
cell – whether or not verbose output should be generated
x_size – the horizontal length of the grid
y_size – the vertical length of the grid
cell_file – the file to save the generated trace to
multifile – if True, each timestep will be save to a seperate file (nice for visualisations)
- setCheckpointing(name, checkpoint_interval)[source]
Warning
name parameter will be used as a filename, so avoid special characters
Sets the interval between 2 checkpoints in terms of GVT calculations. This option generates PDC files starting with ‘name’. This is only possible when using MPI.
- Parameters
name – name to prepend to each checkpoint file
checkpoint_interval – number of GVT runs that are required to trigger a checkpoint. For example 3 means that a checkpoint will be created after each third GVT calculation
- setClassicDEVS(classicDEVS=True)[source]
Use Classic DEVS instead of Parallel DEVS. This option does not affect the use of Dynamic Structure DEVS or realtime simulation. Not usable with distributed simulation.
- Parameters
classicDEVS – whether or not to use Classic DEVS
- setCustomTracer(tracerfile, tracerclass, args)[source]
Sets the use of a custom tracer, loaded at run time.
Calling this function multiple times will register a tracer for each of them (thus output to multiple files is possible, though more inefficient than simply (manually) copying the file at the end).
- Parameters
tracerfile – the file containing the tracerclass
tracerclass – the class to instantiate
args – arguments to be passed to the tracerclass’s constructor
- setDSDEVS(dsdevs=True)[source]
Whether or not to enable Dynamic Structure DEVS simulation. If this is set to True, the modelTransition method will be called on all transitioned models. If this is False, the modelTransition method will not be called, even if one is defined! Enabling this incurs a (slight) slowdown in the simulation, due to the additional function calls and checks that have to be made. Currently only available in local simulation.
- Parameters
dsdevs – enable or not
- setDrawModel(draw_model=True, output_file='model.dot', hide_edge_labels=False)[source]
Whether or not to draw the model and its distribution before simulation starts.
- Parameters
draw_model – whether or not to draw the model
output_file – file to output to
hide_edge_labels – whether or not to hide the labels of the connections, this speeds up the model drawing and allows for reasonably sized diagrams
- setFetchAllAfterSimulation(fetch=True)[source]
Update the complete model by fetching all states from all remote locations. This is different from ‘registerState’, as it will fetch everything and it will modify the original model instead of adding an attribute to the Simulator object
- Parameters
fetch – whether or not to fetch all states from all models
- setGVTInterval(gvt_int)[source]
Sets the interval in seconds between 2 GVT calculations. This is the time between the ending of the previous run and the start of the next run, to prevent overlapping calculations.
Note
Parameter should be at least 1 to prevent an overload of GVT calculations
- Parameters
gvt_int – interval in seconds (float or integer)
- setGreedyAllocator()[source]
Sets the use of the greedy allocator that is contained in the standard PyPDEVS distribution.
- setInitialAllocator(allocator)[source]
Sets the use of an initial allocator instead of the manual allocation. Can be set to None to use manual allocation (default).
- Parameters
allocator – the allocator to use for assigning the initial locations
- setListenPorts(port, function)[source]
Sets a listener on a DEVS port. When an event arrives at that output port, the provided function will be called with the bag (as if that function were the extTransition!).
Listening to ports is only allowed in realtime simulation! Remember to return from the calling function as soon as possible to minimize delays. Only a single listener is supported per port.
- Parameters
port – the port to listen to, can be anything.
function – the function to call when the event arrives. It should take a single parameter (the event bag).
- setLocationCellMap(locationmap, x=0, y=0)[source]
Set the simulation to produce a nice Cell DEVS like output file of the current location. This file will be regenerated as soon as some relocations are processed.
- Parameters
locationmap – whether or not to generate this file
x – the horizontal size of the grid
y – the vertical size of the grid
- setLogging(destination, level)[source]
Sets the logging destination for the syslog server.
- Parameters
destination – A tuple/list containing an address, port pair defining the location of the syslog server. Set to None to prevent modification
level – the level at which logging should happen. This can either be a logging level from the logging module, or it can be a string specifying this level. Accepted strings are: ‘debug’, ‘info’, ‘warning’, ‘warn’, ‘error’, ‘critical’
- setManualRelocator()[source]
Sets the use of the manual relocator (the default). This mode allows the user to add manual relocation directives.
- setMemoization(memo=True)[source]
Use memoization to prevent repeated int/ext/confTransition calls when revertion was performed.
- Parameters
memo – enable or not
- setMessageCopy(copy_method)[source]
Sets the type of message copying to use, this will have an impact on performance. It is made customizable as some more general techniques will be much slower.
- Parameters
copy_method – Either an ID of the option, or (recommended) a string specifying the method, see options below
- setModelAttribute(model, attr, value)[source]
Reinitialize an attribute of the model itself
Calling this method will cause a recomputation of the timeAdvance for this model. Its results will be used relative to the time of the last transition.
- Parameters
model – model whose attribute to set
attr – string representation of the attribute to change
value – value to assign
- setModelState(model, new_state)[source]
Reinitialize the state of a certain model
Calling this method will cause a recomputation of the timeAdvance for this model. Its results will be used relative to the time of the last transition.
- Parameters
model – model whose state to change
new_state – state to assign to the model
- setModelStateAttr(model, attr, value)[source]
Reinitialize an attribute of the state of a certain model
Calling this method will cause a recomputation of the timeAdvance for this model. Its results will be used relative to the time of the last transition.
- Parameters
model – model whose state to change
attr – string representation of the attribute to change
value – value to assign
- setRealTime(realtime=True, scale=1.0)[source]
Sets the use of realtime instead of ‘as fast as possible’.
- Parameters
realtime – whether or not to use realtime simulation
scale – the scale for scaled real time, every delay will be multiplied with this number
- setRealTimeInputFile(generator_file)[source]
Sets the realtime input file to use. If realtime is not yet set, this will auto-enable it.
- Parameters
generator_file – the file to use, should be a string, NOT a file handle. None is acceptable if no file should be used.
- setRealTimePlatformGameLoop()[source]
Sets the realtime platform to Game Loop. If realtime is not yet set, this will auto-enable it.
- Parameters
fps – the number of times the game loop call should be made per second
- setRealTimePlatformThreads()[source]
Sets the realtime platform to Python Threads. If realtime is not yet set, this will auto-enable it.
- setRealTimePlatformTk(tk)[source]
Note
this clearly requires Tk to be present.
Sets the realtime platform to Tk events. If realtime is not yet set, this will auto-enable it.
- setRealTimePorts(ports)[source]
Sets the dictionary of ports that can be used to put input on. If realtime is not yet set, this will auto-enable it.
- Parameters
ports – dictionary with strings as keys, ports as values
- setRelocationDirective(time, model, destination)[source]
Creates a relocation directive, stating that a relocation of a certain model should happen at or after the specified time (depending on when the GVT progresses over this time).
If multiple directives exist for the same model, the one with the highest time will be executed.
- Parameters
time – time after which the relocation should happen
model – the model to relocate at the specified time. Can either be its ID, or an AtomicDEVS or CoupledDEVS model. Note that providing a CoupledDEVS model is simply a shortcut for relocating the COMPLETE subtree elsewhere, as this does not stop at kernel boundaries.
destination – the location to where the model should be moved
- setRelocationDirectives(directives)[source]
Sets multiple relocation directives simultaneously, easier for batch processing. Behaviour is equal to running setRelocationDirective on every element of the iterable.
- Parameters
directives – an iterable containing all directives, in the form [time, model, destination]
- setRemoveTracers()[source]
Removes all currently registered tracers, might be useful in reinitialised simulation.
- setSchedulerActivityHeap(locations=None)[source]
Use the basic activity heap scheduler, this is the default.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerCustom(filename, scheduler_name, locations=None)[source]
Use a custom scheduler
- Parameters
filename – filename of the file containing the scheduler class
scheduler_name – class name of the scheduler contained in the file
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerDirtyHeap(locations=None)[source]
Use the basic activity heap scheduler, but without periodic cleanup. The same scheduler as the one used in VLE.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerDiscreteTime(locations=None)[source]
Use a basic ‘discrete time’ style scheduler. If the model is scheduled, it has to be at the same time as all other scheduled models. It isn’t really discrete time in the sense that it allows variable step sizes, only should ALL models agree on it.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
Warning
Only use in local simulation!
- setSchedulerHeapSet(locations=None)[source]
Use a scheduler containing 3 different datastructures. It is still experimental, though can provide noticeable performance boosts.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerMinimalList(locations=None)[source]
Use a simple scheduler that keeps a list of all models and traverses it each time in search of the first one. Slight variation of the sorted list scheduler.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerNoAge(locations=None)[source]
Warning
do not use this scheduler if the time advance can be equal to 0. This scheduler strips of the age from every scheduled model, which means that ages do not influence the scheduling.
Use a stripped scheduler that doesn’t care about the age in a simulation. It is equivalent in design to the HeapSet scheduler, but uses basic floats instead of tuples.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerPolymorphic(locations=None)[source]
Use a polymorphic scheduler, which chooses at run time between the HeapSet scheduler or the Minimal List scheduler. Slight overhead due to indirection and statistics gathering.
Warning
Still unstable, don’t use!
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setSchedulerSortedList(locations=None)[source]
Use an extremely simple scheduler that simply sorts the list of all models. Useful if lots of invalidations happen and nearly all models are active.
- Parameters
locations – if it is an iterable, the scheduler will only be applied to these locations. If it is None, all nodes will be affected.
- setShowProgress(progress=True)[source]
Shows progress in ASCII in case a termination_time is given
- Parameters
progress – whether or not to show progress
- setStateSaving(state_saving)[source]
Sets the type of state saving to use, this will have a high impact on performance. It is made customizable as some more general techniques will be much slower, though necessary in certain models.
- Parameters
state_saving – Either an ID of the option, or (recommended) a string specifying the method, see options below.
- deepcopy¶
use the deepcopy module
- pickle0¶
use the (c)pickle module with pickling protocol 0
- pickleH¶
use the (c)pickle module with the highest available protocol
- pickle¶
use the (c)pickle module
- copy¶
use the copy module (only safe for flat states)
- assign¶
simply assign states (only safe for primitive states)
- none¶
equivalent to assign (only safe for primitive states)
- custom¶
define a custom ‘copy’ function in every state and use this
- setTerminationCondition(condition)[source]
Sets the termination condition for the simulation. Setting this will remove a previous termination time and condition. This condition will be executed on the controller
- Parameters
condition – a function to call that returns a boolean whether or not to halt simulation
- setTerminationModel(model)[source]
Marks a specific AtomicDEVS model as being used in a termination condition. This is never needed in case no termination_condition is used. It will _force_ the model to run at the controller, ignoring the location that was provided in the model itself. Furthermore, it will prevent the model from migrating elsewhere.
- Parameters
model – an AtomicDEVS model that needs to run on the controller and shouldn’t be allowed to migrate
- setTerminationTime(time)[source]
Sets the termination time for the simulation. Setting this will remove a previous termination time and condition.
- Parameters
time – time at which simulation should be halted
- setVCD(filename)[source]
Sets the use of a VCD tracer.
Calling this function multiple times will register a tracer for each of them (thus output to multiple files is possible, though more inefficient than simply (manually) copying the file at the end).
- Parameters
filename – string representing the filename to write the trace to
- setVerbose(filename=None)[source]
Sets the use of a verbose tracer.
Calling this function multiple times will register a tracer for each of them (thus output to multiple files is possible, though more inefficient than simply (manually) copying the file at the end).
- Parameters
filename – string representing the filename to write the trace to, None means stdout
- setXML(filename)[source]
Sets the use of a XML tracer.
Calling this function multiple times will register a tracer for each of them (thus output to multiple files is possible, though more inefficient than simply (manually) copying the file at the end).
- Parameters
filename – string representing the filename to write the trace to