Classes and Methods

PyGAS is composed of a small set of core components that implement the PGAS extension to Python, a modest set of convenience functions for higher-level functionality, and a collection of tools commonly used when writing programs using PyGAS.

Core Components

class pygas.Proxy(obj)

This is the fundamental PyGAS object. It wraps an existing object and mimics its behavior, even when the original object is on a remote thread. UPC programmers can think of Proxies as shared pointers because they store a (thread_id, local_addr) tuple internally.

owner

The id of the thread where the original object resides.

__init__(obj)

Initialize a Proxy to the given object.

Parameters:obj (object) – the object to be wrapped

resolve()

Return a copy of the original object wrapped by this Proxy.

Return type:object

__getattr__(name)

Get a copy of a remote attribute.

Parameters:name (string) – the attribute name

__setattr__(name, value)

Set a remote attribute to a copy of a local object.

Parameters:

• name (string) – the attribute name
• value (object) – the new attribute value

Return type:

None

__call__(*args, **kwargs)

Invoke the __call__ method on the original object and return a copy of the result.

Parameters:

• args (dictionary) – positional arguments
• kwargs – keyword arguments

Return type:

object

pygas.share(obj, from_thread=0)

Create and broadcast a Proxy to the given object. This is a collective operation. It is effectively syntactic sugar for:

broadcast(Proxy(obj), from_thread=...)

Parameters:

• obj (object) – the object to be shared.
• from_thread (int) – the thread wishing to share obj.

Return type:

Proxy

Extended Components

pygas.THREADS = 1

The number of threads in this parallel job.

pygas.MYTHREAD = 0

The thread number of the current thread.

pygas.broadcast(obj, from_thread=0)

Broadcast copies of the given object. This is a collective operation.

Parameters:

• obj (object) – the object to be copied.
• from_thread (int) – the thread wishing to copy obj.

Returns:

A copy of the given object.

Miscellaneous Tools

class pygas.SplitTimer(name='timer')

Implements a context manager that simplifies timing sections of code. Use like:

with SplitTimer("computation") as timer:
    compute()
print timer.report()

average()

Average time of all splits.

Return type:float

report()

Return report of performance.

Return type:string