Advanpix Multi precision toolbox

Question

Kamuran 2012-9-27

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/49244-advanpix-multi-precision-toolbox

I am seeking an efficient way of doing my numerical calculation in quad or more precision. Using a symbolic toolbox makes the code so slow that I will need years to finish a calculation. I found Advanpix selling a multi-precision toolbox for MATLAB but I am a little skeptical because if someone outside MATLAB can do this why not MATLAB doing it?

Anybody has any experience with Advanpix?

Thank you

Erdem

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

Pavel Holoborodko 2012-9-28

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/49244-advanpix-multi-precision-toolbox#answer_60231

"...if someone outside MATLAB can do this why not MATLAB doing it?"

There are many important things which MATLAB is not doing (for whatever reason). For example, see the list of third-party toolboxes for MATLAB here:

http://www.walkingrandomly.com/?p=4452

It is easy to test if Advanpix mp-toolbox suits your needs - full functional trial version is available for download from the website.

P.S. I'm author of Advanpix Multiprecision Computing Toolbox.

2 个评论
显示无隐藏无

Kamuran 2012-9-28

Let me ask you one more thing. Will Multiprecision Computing Toolbox work with GPU computing such as Matlab's own GPUarray way or using Jacket (I am looking into that from the link you sent me). As you might understood, I am trying to take all the juice Matlab can offer me because I am planing to get into high end CFD calculations where I need all the improvement I can get.

Thank you

Pavel Holoborodko 2012-9-28

编辑：Pavel Holoborodko 2013-7-18

Acceleration by GPU is not supported in toolbox (yet). And yes, huge matrices processing will be painfully slow.

We are working on employing quadruple precision arithmetic (instead of arbitrary precision) for big matrices. Hopefully this will give us significant boost in performance. New version will be ready in a few months...

Update (21 Feb 2013): As was advertised, now Advanpix toolbox implements highly optimized computations with quadruple precision (34 decimal digits). For example, now SVD/EIG are by x70-80 times faster comparing to symbolic toolbox.

Now toolbox makes it even possible to handle large matrices using quadruple precision.

Some examples can be found on the page: Fast Quadruple Precision Computations

请先登录，再进行评论。