there are two method for get timestamp, first is just use system call, second is use system call for the first time, and then start a background thread and let system to schedule a inc function every second. from this test, we can see that the second method is more efficient.

t.py

#!/usr/bin/env python

import threading, time

time1 = 0

def get_time1():
	return time1

def get_time2():
	return int(time.time())#this involve system call

def periodic_inc_time():
	global time1
	time1 += 1

def test1(times):
	time1 = int(time.time())
	threading.Timer(1, periodic_inc_time).start()
	for i in range(times):
		get_time1()

def test2(times):
	for i in range(times):
		get_time2()

if __name__ == '__main__':
	times = 1000000

	start = time.time()
	test2(times)
	end = time.time()
	print "for test2 " + str(end - start)#for test2 0.422958850861

	#because test1 will start a background thread, we test it lastly
	start = time.time()
	test1(times)
	end = time.time()
	print "for test1 " + str(end - start)#for test1 0.156961917877