execution speed changes by sequence of running the code

Question

hamed amini 2017-4-4

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/333641-execution-speed-changes-by-sequence-of-running-the-code

回答： Steven Lord 2017-4-4

the question updated: the sequence of running different codes changes the run time.

--------------

initially it appeared to be due to difference in file name: any similar experience? changing the function name is changing the execution time (picture attached).

2 个评论
显示无隐藏无

Steven Lord 2017-4-4

What else was running on your machine when you ran HAPP_dry_KX3 that was not running when HAPP_dry_KX1 was running? Include Windows "phoning home" to look for updates, virus scanners scanning your machine, etc.

Had you run HAPP_dry_KX1 prior to capturing that picture, then run HAPP_dry_KX3 for the first time while generating that picture? If so, in a new MATLAB session try running HAPP_dry_KX3 multiple times then run HAPP_dry_KX1 for the first time.

hamed amini 2017-4-4

- Tnx. Closed the Matlab session and opened a new one. The timing became the same indeed (both took the longer time).

- computer is not busy with other applications

- These (HAPP_dry_KX1 & HAPP_dry_KX3) are part of optimization algorithms with rather large parameter space (7-D). I also have another version of the optimization algorithm (HAPP_dry_KX2 & HAPP_dry_KX4) with less number of variables (shorter run time). After running HAPP_dry_KX1 and HAPP_dry_KX3 several time, I ran HAPP_dry_KX2, and again HAPP_dry_KX1 and HAPP_dry_KX3; this time both (1 & 3) have shorter run time.

- It seems the sequence of running of these codes makes a difference; however could not find the logic behind this behavior.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Steven Lord 2017-4-4

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/333641-execution-speed-changes-by-sequence-of-running-the-code#answer_261764

There is some first run overhead (see the "Tips for Measuring Performance" section of that page.) So if you compare the Nth (N > 1) run of one function and the 1st run of another function that contains the same code in a different file, in the latter case you'll measure the overhead in addition to the execution of the code.