Question

所以我正在使用timeit模块查看变量赋值的速度，我发现了一些非常有趣的结果。我发现多行分配通常比单行分配更快。这是为什么？

另外，值得注意的是，当将time.sleep(1)置于下面的循环中时，它似乎会影响速度结果显着。这是为什么？

我的代码和结果如下：

没有 time.sleep ：

dis results for myfunc1:
  5           0 LOAD_CONST               2 ((10, 10))
              3 UNPACK_SEQUENCE          2
              6 STORE_FAST               0 (x)
              9 STORE_FAST               1 (y)
             12 LOAD_CONST               0 (None)
             15 RETURN_VALUE        
dis results for myfunc2:
  7           0 LOAD_CONST               1 (10)
              3 STORE_FAST               0 (x)

  8           6 LOAD_CONST               1 (10)
              9 STORE_FAST               1 (y)
             12 LOAD_CONST               0 (None)
             15 RETURN_VALUE        


Speed results for single line assignment:
4.61830006532e-06
4.10515561362e-06
4.10515561362e-06
4.61830006532e-06
4.61830006532e-06
4.10515561362e-06
4.10515561362e-06
4.10515561362e-06
4.10515561362e-06
4.10515561362e-06
4.10515561362e-06


Speed Results for multi-line assignment:
4.10515561362e-06
3.59201116192e-06
4.10515561362e-06
3.59201116192e-06
4.10515561362e-06
3.59201116192e-06
3.59201116192e-06
4.10515561362e-06
4.10515561362e-06
3.59201116192e-06
4.10515561362e-06

time.sleep ：

dis results for myfunc1:
  5           0 LOAD_CONST               2 ((10, 10))
              3 UNPACK_SEQUENCE          2
              6 STORE_FAST               0 (x)
              9 STORE_FAST               1 (y)
             12 LOAD_CONST               0 (None)
             15 RETURN_VALUE        
dis results for myfunc2:
  7           0 LOAD_CONST               1 (10)
              3 STORE_FAST               0 (x)

  8           6 LOAD_CONST               1 (10)
              9 STORE_FAST               1 (y)
             12 LOAD_CONST               0 (None)
             15 RETURN_VALUE        


Speed results for single line assignment:
4.61830006532e-06
5.13144451708e-06
9.03134234993e-05
0.000157022202221
8.77477012411e-05
4.61830006504e-06
4.7722434009e-05
1.59074780033e-05
0.000105707757051
0.000324307293475
1.43680446492e-05


Speed Results for multi-line assignment:
1.33417557446e-05
0.000202692058421
1.33417557446e-05
0.000163693080093
1.3854900196e-05
1.3854900196e-05
0.000279663726175
0.000142141013121
0.000202692058423
0.000200639480614
0.000103655179245

代码：

import dis,timeit,time

#define the different methods of assignment
def myfunc1():
    x,y=10,10
def myfunc2():
    x=10
    y=10

#used simply for viewing how they are assigned
print "dis results for myfunc1:"
dis.dis(myfunc1)
print "dis results for myfunc2:"
dis.dis(myfunc2)

print "\n"

print "Speed results for single line assignment:"
#repeat 10 times, with looping 100 times per repeat to show speed and consistency
for x in range(11):
    print timeit.timeit(stmt="x,y=10,10",number=100)
    #time.sleep(1)
print "\n"

print "Speed Results for multi-line assignment:"
#repeat with the same method for previous test
for x in range(11):
    print timeit.timeit(stmt="x=10;y=10;",number=100)
    #time.sleep(1)

基准：Python 2.7.3 (default, Apr 10 2012, 23:31:26) [MSC v.1500 32 bit (Intel)] on win32

Answer 1

我发现多行分配通常比单行分配更快。这是为什么？

因为UNPACK_SEQUENCE比LOAD_CONST慢。

另外，值得注意的是，当将time.sleep（1）放在下面的循环中时，它似乎会显着影响速度结果。这是为什么？

因为100次迭代对于快速操作而言太低，并且受（低）时钟精度的影响很大。使用number=10000000重新测试，无论睡眠如何，您都会得到大致相同的结果。

多线变量分配速度和通量

1 个答案: