Python multithreading and inconsistent output

Question

import threading

x = 0;

class Thread1(threading.Thread):
    def run(self):
        global x
        for i in range(1,100000):
            x = x + 1

class Thread2(threading.Thread):
    def run(self):
        global x
        for i in range(1,100000):
            x = x - 1

#create two threads
t1 = Thread1()
t2 = Thread2()

#start the threads
t1.start()
t2.start()

#wait for the threads to finish
t1.join()
t2.join()

print x;

Running this multiple times produces different output, some in the negative and some in the positive. Is it because the two threads are using the same global x? I don't understand why: shouldn't the net effect (output) be the same after all the dust settles?

Your operations are not atomic. You should use locks for correct results. — akaRem
– akaRem, Commented Dec 1, 2012 at 22:46

senderle · Accepted Answer · 2012-12-02 22:20:28Z

3

Not necessarily. Imagine the following sequence of events. We'll begin at a precise moment after the program has been running for a bit; both Thread1 and Thread2 are inside their for loops, and x = 0

Thread1 has control. It accesses x to determine the value of x + 1; x is 0, so the result is 1. However...
Before Thread1 completes the assignment, control passes to Thread2. x is still 0.
Thread2 now accesses x. It calculates x - 1, which is -1, because x is still 0. Because of the unpredictability of thread timing, it manages to complete the operation, assigning -1 to x.
Control now returns to Thread1. It has already calculated the value of x + 1 to be 1. It assigns 1 to x.

Both threads have completed an iteration, and the value of x should be 0, but its actual value is 1.

edited Dec 2, 2012 at 22:20

answered Dec 1, 2012 at 22:31

senderle

152k36 gold badges218 silver badges244 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

DaveP · Accepted Answer · 2012-12-01 22:33:14Z

2

This is a classic concurrency problem in multithreaded computing. Each thread must read the current value of x from memory, modify it, then write it back. If thread 1 reads the value, then thread 2 updates it, when thread 1 writes the value back it will cancel thread 2's update. This is why you should always use proper synchronisation constructs such as semaphores etc.

answered Dec 1, 2012 at 22:33

DaveP

7,1222 gold badges27 silver badges37 bronze badges

Comments

Yossi · Accepted Answer · 2012-12-01 22:29:37Z

1

It is a classical race condition. Wiki has a good explanation of this situation.

answered Dec 1, 2012 at 22:29

Yossi

12.1k2 gold badges56 silver badges67 bronze badges

Collectives™ on Stack Overflow

Python multithreading and inconsistent output

3 Answers 3

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related