What is the standard type of operations in MATLAB?

Question

Urs Waldmann 2016-7-5

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/293755-what-is-the-standard-type-of-operations-in-matlab

编辑： Stephen23 2016-7-5

I was wondering what the standard type in MATLAB operations is, i.e. single or double. I guess it is double. But how can I change the operations like e.g. eig() into single precision operations? Is it enough to specify the input argument as a single presision input argument or is there something elese. I was wondering because e.g. for the function zeros() or ones() I have to specify 'single' or 'double'.

Or is there a way to tell MATLAB to do EVERYTHING in single precision?

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

Stephen23 2016-7-5

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/293755-what-is-the-standard-type-of-operations-in-matlab#answer_227691

编辑：Stephen23 2016-7-5

From the numeric data types documentation: "By default, MATLAB stores all numeric values as double-precision floating point". It is not possible to change the default data type globally.

However operations are a different matter entirely... The MATLAB documentation does not say what number format is used for intermediate values in functions like eig. It is likely to be double, or even higher precision through the use of guard digits. Note that many numeric operations, such as eig, are provided by the libraries upon which MATLAB is based, such as LAPACK (earlier used EISPACK and LINPACK), and so the algorithm used in those libraries defines the result that MATLAB returns.

You should definitely read this paper: What Every Computer Scientist Should Know About Floating-Point Arithmetic, by David Goldberg.