Still having problems loading libraries with 2015a and windows 10

Further to my previous question, I tried following the example in the loadlibrary() help menu:

>> matlabroot

ans =

C:\Program Files\MATLAB\R2015a

>> addpath(fullfile(matlabroot,'extern','examples','shrlib'))
>> loadlibrary('shrlibsample')

Index exceeds matrix dimensions.

Error in loadlibrary>getLoadlibraryCompilerConfiguration (line 527)

Error in loadlibrary (line 263)

Can someone please explain why I can't even load the dll example that is in the matlab help? Is this a windows 10 compatibility problem?

Answer 1

[Shruti Sapre]

Hi Tim,

I understand that you are trying to load the dll from example on the Mathworks help page. The error you mentioned could be due to absence of a supported compiler on your machine.

Please refer to the link below for limitations of “loadlibrary”:

http://www.mathworks.com/help/matlab/ref/loadlibrary.html#zmw57dd0e389433

The below command could help confirm the presence of a supported compiler:

>> mex –setup

"mex" has to be setup to point to a supported compiler.

If there is no compiler installed, please install a supported compiler and try again. Here is a link that lists the supported compilers:

http://www.mathworks.com/support/compilers/R2015a/index.html

Hope this helps!

-Shruti

[Tim G.]

Shruti,

I am confused as to why i need to have a compiler installed on my system in order to open and access a pre-existing DLL.

Regardless, I have the sdk 7.1 and MS Visual Studio 2013 Professional installed. I also have perl installed.

When i type mex -setup i get an error

 >>  mex -setup
Error using mex
No supported compiler or SDK was found. For options, visit 
http://www.mathworks.com/support/compilers/R2015a/win64.html.

Answer 2

[Walter Roberson]

I have not yet seen any evidence that SDK 7.1 can be installed on Windows 10; there is a conflict with .NET versions that I have not seen resolved yet.

loadlibrary() that is not given a prototype file must be supplied with a .h file that lists the available routines and the datatypes of the output and of all of the inputs. It then requires a compiler (or at least pre-processor perhaps) to build from that a MATLAB-readable description that can be used to convert inputs and outputs in order to call the routine successfully.

When you succesfully use loadlibrary() with a header file, there is an option to output from that a prototype file which MATLAB can use instead without needing a compiler. That prototype file can be used on systems without compilers, or can be deployed with an application compiled with MATLAB Compiler. Therefor if you still have access to a Windows system with a compiler and with MATLAB, you could generate the prototype file on there and copy it to your Windows 10 system.

[Philip Borghesani]

Walter is correct for 32 bit systems. On 64 bit systems a thunk file (dll) is also required to load the library. This file is automatically built from generated c code by the loadlibrary command using the configured compiler but can be copied and used on other systems that do not have a compiler.