Can't use mtimesx, "A C/C++ compiler has not been selected"

r_mat = [1,0,0; 0,0,0; 0,0,0];

dog = im2double(imread('download.jpg'));
r_dog = mtimesx(dog, r_mat);

imshow(r_dog);

I wasn't sure how to set up the mtimesx function so I just copied all the files from the download into the directory where my script is located.

I try to run the above code but I get the error "A C/C++ compiler has not been selected with mex -setup". I installed the MinGW add-on and checked that it is the selected compiler.

How do I get the function mtimesx to work?

[Anton Loubman]

I checked and selected the compiler using

>> mex -setup

The result is "MEX configured to use 'MinGW64 Compiler (C++)' for C++ language compilation."

The same is shown for C.

Answer 1

[Matt J]

Run mex- -setup from the Matlab prompt,

>> mex -setup

and select one of the available compilers.

[Anton Loubman]

"I installed the MinGW add-on and checked that it is the selected compiler. "

I already did, it is the selected compiler.

[Matt J]

Execute this, and show us the full copy/pasted output

>> mtimesx(rand(3,3,2),rand(3,3))

[Anton Loubman]

>> main
 
... Build routine for mtimesx
... Checking for PC
... Finding path of mtimesx C source code files
... Found file mtimesx.c in C:\Users\anton\OneDrive\Documents\MATLAB\mtimesx.c
... Found file mtimesx_RealTimesReal.c in C:\Users\anton\OneDrive\Documents\MATLAB\mtimesx_RealTimesReal.c
Error using mtimesx_build (line 169)
A C/C++ compiler has not been selected with mex -setup

Error in mtimesx (line 271)
mtimesx_build;

Error in main (line 1)
mtimesx(rand(3,3,2),rand(3,3))

This was the output for your command. Then I ran

mex -setup

and this was the output

>> mex -setup
MEX configured to use 'MinGW64 Compiler (C)' for C language compilation.

To choose a different C compiler, select one from the following:
MinGW64 Compiler (C)  mex -setup:C:\Users\anton\AppData\Roaming\MathWorks\MATLAB\R2020a\mex_C_win64.xml C
Microsoft Visual C++ 2019 (C)  mex -setup:'C:\Program Files\MATLAB\R2020a\bin\win64\mexopts\msvc2019.xml' C

To choose a different language, select one from the following:
 mex -setup C++ 
 mex -setup FORTRAN

[Matt J]

Ah well. No idea. But I've attached a compiled MEX for mtimesx for Win64 that you can use.

[Anton Loubman]

Well it worked, sort of. For some reason it still refuses to multiply a 225x225x3 matrix by a 3x3 matrix, because the "Inner matrix dimensions must agree". If I remove the im2double call then it gives the error "Unrecognized function or variable 'mtimesx_sparse'."

Should I give up on multiplying matrices of dimension greater than 2? Is it that unnecessary?

Edit: I added the rest of the functions from the original mtimesx zip folder and it finds mtimesx_sparse now. However it still won't work unless I remove im2double, but if I do that it outputs

Error using  * 
MTIMES (*) is not fully supported for integer classes. At least one argument must be scalar.

Error in mtimesx_sparse (line 47)
        result = a * b;

So I think I'll just give up on trying to multiplying matrices that have more than two dimensions. Can't imagine it's this hard for no reason.

[Matt J]

Well it worked, sort of. For some reason it still refuses to multiply a 225x225x3 matrix by a 3x3 matrix

I can assure you that im2double has nothing to do with the problem. A 225x225x3 matrix is definitely not compatible for slice-wise multiplication with a 3x3 matrix, and I'm not sure why you think it is. It would be equivalent to trying to do this:

A=rand(225,225,3); B=rand(3,3);

for i=1:3, C(:,:,i)=A(:,:,i)*B; end  %what C=mtimesx(A,B) is trying to do

If instead, you are trying to take every A(i,j,:), make it into a row vector, and pre-multiply the row vector by, then you don't need mtimesx for that. You would instead do,

C=reshape( reshape(A,[],3)*B , [225,225,3]);

[Anton Loubman]

You're right, in my head the dimensions made sense because the left matrix ended with a 3 and the right one starts with a 3, my bad.

At any rate thanks a lot for the help.

Correction: the matrices are compatible, as confirmed by numpy in python and a for loop implementation of matrix multiplication in MATLAB. I still don't know why mtimesx didn't multiply them so I'll just use workarounds.