Python TutorialGetting Started with PythonPython Basic SyntaxPython DatatypesPython IndentationPython Collection TypesPython Basic Input and OutputPython Built in Modules and FunctionsPython FunctionsChemPy - python packageCreating Python packagesFunctional Programming in PythonIncompatibilities moving from Python 2 to Python 3IoT Programming with Python and Raspberry PIKivy - Cross-platform Python Framework for NUI DevelopmentMutable vs Immutable (and Hashable) in PythonPyInstaller - Distributing Python CodePython *args and **kwargsPython 2to3 toolPython Abstract Base Classes (abc)Python Abstract syntax treePython Alternatives to switch statement from other languagesPython and ExcelPython Anti-PatternsPython ArcPyPython ArraysPython Asyncio ModulePython Attribute AccessPython AudioPython Binary DataPython Bitwise OperatorsPython Boolean OperatorsPython Checking Path Existence and PermissionsPython ClassesPython CLI subcommands with precise help outputPython Code blocks, execution frames, and namespacesPython Collections modulePython Comments and DocumentationPython Common PitfallsPython Commonwealth ExceptionsPython ComparisonsPython Complex mathPython concurrencyPython ConditionalsPython configparserPython Context Managers (with Statement)Python Copying dataPython CountingPython ctypesPython Data SerializationPython Data TypesPython Database AccessPython Date and TimePython Date FormattingPython DebuggingPython DecoratorsPython Defining functions with list argumentsPython DeploymentPython Deque ModulePython DescriptorPython Design PatternsPython DictionaryPython Difference between Module and PackagePython DistributionPython DjangoPython Dynamic code execution with `exec` and `eval`Python EnumPython ExceptionsPython ExponentiationPython Files & Folders I/OPython FilterPython FlaskPython Functools ModulePython Garbage CollectionPython GeneratorsPython getting start with GZipPython graph-toolPython groupby()Python hashlibPython HeapqPython Hidden FeaturesPython HTML ParsingPython HTTP ServerPython IdiomsPython ijsonPython Immutable datatypes(int, float, str, tuple and frozensets)Python Importing modulesPython Indexing and SlicingPython Input, Subset and Output External Data Files using PandasPython Introduction to RabbitMQ using AMQPStorm

Python Parallel computation

From WikiOD

Remarks[edit | edit source]

Due to the GIL (Global interpreter lock) only one instance of the python interpreter executes in a single process. So in general, using multi-threading only improves IO bound computations, not CPU-bound ones. The multiprocessing module is recommended if you wish to parallelise CPU-bound tasks.

GIL applies to CPython, the most popular implementation of Python, as well as PyPy. Other implementations such as Jython and IronPython have no GIL.

Using the multiprocessing module to parallelise tasks[edit | edit source]

import multiprocessing

def fib(n):
    """computing the Fibonacci in an inefficient way
    was chosen to slow down the CPU."""
    if n <= 2:
        return 1
        return fib(n-1)+fib(n-2) 
p = multiprocessing.Pool() 

# Out: [39088169, 24157817, 14930352, 9227465, 5702887, 3524578]

As the execution of each call to fib happens in parallel, the time of execution of the full example is 1.8× faster than if done in a sequential way on a dual processor.

Python 2.2+

Using a C-extension to parallelize tasks[edit | edit source]

The idea here is to move the computationally intensive jobs to C (using special macros), independent of Python, and have the C code release the GIL while it's working.

#include "Python.h"
PyObject *pyfunc(PyObject *self, PyObject *args) {
    // Threaded C code

Using Parent and Children scripts to execute code in parallel[edit | edit source]

import time

def main():
    print "starting work"
    print "work work work work work"
    print "done working"

if __name__ == '__main__':

import os

def main():
    for i in range(5):
        os.system("python &")

if __name__ == '__main__':

This is useful for parallel, independent HTTP request/response tasks or Database select/inserts. Command line arguments can be given to the script as well. Synchronization between scripts can be achieved by all scripts regularly checking a separate server (like a Redis instance).

Using PyPar module to parallelize[edit | edit source]

PyPar is a library that uses the message passing interface (MPI) to provide parallelism in Python. A simple example in PyPar (as seen at looks like this:

import pypar as pp

ncpus = pp.size()
rank = pp.rank()
node = pp.get_processor_name()

print 'I am rank %d of %d on node %s' % (rank, ncpus, node)

if rank == 0:
  msh = 'P0'
  pp.send(msg, destination=1)
  msg = pp.receive(source=rank-1)
  print 'Processor 0 received message "%s" from rank %d' % (msg, rank-1)
  source = rank-1
  destination = (rank+1) % ncpus
  msg = pp.receive(source)
  msg = msg + 'P' + str(rank)
  pypar.send(msg, destination)