how to use parallel programming or multi threading or multicore in matlab?

Question

Jack Ie 2016-4-5

2
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/277216-how-to-use-parallel-programming-or-multi-threading-or-multicore-in-matlab

回答： Walter Roberson 2016-4-5

I have a program in matlab. It run in 1 hour. It is very slow!. how can I use parallel programming or multi threading or multicore in matlab? Is there any other way to improve run time?

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ced 2016-4-5

在 MATLAB Online 中打开

To your actual question, the short answer is: yes, matlab can use multithreading and process things in parallel. Check out Parallel Processing in Matlab

Parallelization is not the first thing I would try though. In many cases, a huge speed up can be achieved by writing the process in an optimized form such as vectorizing operations instead of loops and conditional statements.

You can check which part of your code take up most of the time using the profiler. Just run

profile on

then run your code (or a part of it) and afterwards, call

profile viewer

This will give you a detailed description of which parts of the code are most worth optimizing.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2016-4-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/277216-how-to-use-parallel-programming-or-multi-threading-or-multicore-in-matlab#answer_216475

You would need the Parallel Computing Toolbox for that. It turns out that for a lot of programs, using Parallel Processing is slower rather than faster, because of the communications overhead.

A lot of the time, code can be sped up more by vectorizing and clever rewriting than by switching to Parallel Computing. But it depends on what the code does.