Existence of parallel computing resources (statistics, etc)?

Question

Jakob Sievers 2012-9-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/47460-existence-of-parallel-computing-resources-statistics-etc

Hi there

I am running some long processes within a for-loop and have started wondering if I might be able to speed up the whole thing by using parallel computing and parfor-loops instead. Can anyone recommend me any good resources on this matter? I'm thinking statistics, computer-time vs. number of parallel workers for different problems, etc.

Thanks in advance

Jakob

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Edric Ellis 2012-9-6

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/47460-existence-of-parallel-computing-resources-statistics-etc#answer_58044

There are some simple benchmark results for PARFOR here. Whether your particular problem will speed up using PARFOR depends on a number of factors:

How much of your problem is amenable to parallelisation (Amdahl's law)
Are you going to run only on your local machine, or do you have a cluster
Is your problem already benefiting from MATLAB' built-in multithreading (multithreading generally beats PARFOR on a single machine if it is available)
The data access pattern of your problem (i.e. do you need to transmit large amounts of data compared to the amount of computation you're doing)