python multiprocess fails to start

Question

Here is my code for a simple multiprocessing task in python

from multiprocessing import Process

def myfunc(num):

    tmp = num * num
    print 'squared O/P will be ', tmp
    return(tmp)


a = [ i**3 for i in range(5)] ## just defining a list

task = [Process(target = myfunc, args = (i,)) for i in a]  ## creating processes

for each in task : each.start()  # starting processes <------ problem line

for each in task : each.join()   # waiting all to finish up

When I run this code, it hangs at certain point, so to identify it I ran it line by line in python shell and found that when I call 'each.start()' The shell pops out a dialogue box as:

" The program is still running , do you want to kill it? '

and I select 'yes' the shell closes.

When I replace Process with 'threading.Thread' the same code runs but with this nonsense output:

Squared Squared Squared Squared Squared  0    1491625
36496481

Is there any help in this regard ? thank in advance

To run my python codes I use Idlex IDE and I start it from terminal.

I have Intel Xeon Processor with 4 cores / 8 Threads, and 8GB RAM

Answer 1

Save the code above into a .py file and then run it in a gnome-terminal with

python myfile.py

Where "myfile.py" is the filename you saved to.

I would assume that the IDE you are using is confused somehow by Process()

Answer 2

With a little thought I finally found the problem.

This is happening because in Python, the float and int objects are not 'thread-safe', meaning the memory allocated to calculate any function's value by one thread/process can be overwritten by another and hence they show absurd values. This is called a race condition.

To solve this problem, use deque() from the collections module or, even better, use the 'Lock' facility. deque() works with arrays but it's meant for arrays of the same kind (much like MATLAB arrays) and is thread/process safe. 'Lock' avoids race conditions.

So the edit would be :

def myfunc(num):

    lock.acquire()

    .......some code .....
    .......some code......

    lock.release()

That's all.

But one problem still persists and that is with the multiprocessing module. Even after calling 'lock', the problem mentioned in the question remains.