Overview
- When a variable is created in thread local storage every thread will have its own copy which it can update.
- Thread local storage is not automatic variables.
- They are valid for the lifetime of the thread.
- They need not be locked for synchronization unless shared among threads.
- In Python, the class threading.local incorporates the concept of Thread local storage.
- A thread local variable can be added by instantiating a threading.local object and adding attributes to it. Each attribute added to the threading.local object is a thread local variable.
Example
# Example Python program that uses thread local storage
# to calculate rolling sums in different threads
# and adding them to a global sum
import threading
import time
# Final sum
g_sum = 0
# Calculates rolling sum and update the global sum
def counterThread(numRange, tLock, tIndex):
# Create thread local storage
tls = threading.local()
# Add attribute to the thread local storage
# Per thread Sum
tls.tSum = 0
for i in range(0, numRange):
# No locking needed
tls.tSum += i
# Posting final sum to the global
tLock.acquire()
global g_sum
g_sum += tls.tSum
tLock.release()
# Create locks
tLock = threading.Lock()
# Create threads that compute individual rolling sums
t1 = threading.Thread(target = counterThread(5, tLock, 1))
t2 = threading.Thread(target = counterThread(2, tLock, 2))
t3 = threading.Thread(target = counterThread(7, tLock, 3))
t1.start()
t2.start()
t3.start()
t1.join()
t2.join()
t3.join()
print("Final sum after adding up numbers from each thread:%d"%g_sum)Output
Final sum after adding up numbers from each thread:32