How can I achieve Matrix-Vector element-wise operations in Simulink Matlab Function Block?

TL;dr matlab seems to support elementwise operations when one of the elements is a vector of the appropriate length (e.g. [4x3] .^ [4x1]). Simulink seems to require that they are both matrices of identical sizes (e.g. [4x3] .^ [4x3]). Can I achieve the matlab functionality in a simulink matlab function block? If not, why?

Here is the MWE; a matlab function block that spits the results into a display block:

Here is the code for the matlab function block. It's a one-line function, and I have three different versions.

This one doesn't work.

function ans = fcn()
ans = ones(4,3) .^ ones(4,1);

It produces an error about matrix dimensions needing to agree (see below or the full error). It suggests specifying the output size, which I tried (4x3) and that did not resolve the issue. Additionally, this line of code runs perfectly fine in matlab:

ones(4,3) .^ ones(4,1)

ans =

     1     1     1
     1     1     1
     1     1     1
     1     1     1

This one does work because the dimensions match exactly:

function ans = fcn()
ans = ones(4,3) .^ ones(4,3);

This is my workaround to use the original dimensions:

function ans = fcn()
ans = ones(4,3) .^ repmat(ones(4,1),1,3);

I know it's only one extra function call, but I would love to avoid excessive repmat() calls throughout my entire project. Is there a way to achieve the elementwise operation with a matrix and a vector in a simulink matlab function? If not, is there some rational why simulink doesn't support this functionality?

Thank you!

Here's the full error message:

Matrix dimensions must agree. Function 'MATLAB Function1' (#104.49.71), line 4, column 7: "ones(4,3) .^ ones(4,1)" Launch diagnostic report.

Component:MATLAB Function | Category:Coder error

Errors occurred during parsing of MATLAB function 'MWE/MATLAB Function1'

Component:MATLAB Function | Category:Coder error

Simulink cannot determine sizes and/or types of the outputs for block 'MWE/MATLAB Function1' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

Component:MATLAB Function | Category:Coder error

Simulink cannot determine sizes and/or types of the outputs for block 'MWE/MATLAB Function1' due to errors in the block body, or limitations of the underlying analysis. The errors might be inaccurate. Fix the indicated errors, or explicitly specify sizes and/or types for all block outputs.

Answer 1

[Jon]

I'm not sure why the MATLAB Function block can not handle the implicit scalar expansion, but I just tried using an Interpreted MATLAB function block which calls a function which executes ones(4,3).^ones(4,1) and it works without any errors. So if your application allows you to use Interpreted MATLAB function blocks that may be a work around. I'm not sure what limitations/performance tradeoffs there are for those two kinds of blocks which seem to do somewhat similar jobs.

[Landen]

Thank you, but I'm strictly interested in using a matlab function for performance reasons. My workaround using matrep() is going to be sufficient to continue using the matlab function, but I would really like to utilize implicit scalr expansion if possible. This function is a bottleneck for a larger simulation, so it needs to be fast. Here's a link about the tradeoffs between a matlab function and an interpreted matlab function:

https://www.mathworks.com/help/simulink/ug/comparison-of-custom-block-functionality.html

[Jon]

Using bsxfun seems to work inside of a MATLAB function block. I don't know if the performance hit for repmat is better or worse than using bsxfun, but it is at least another possibility

y = bsxfun(@power,ones(4,3),ones(4,1))

[Landen]

Thanks! This is also a great workaround, and a bit more readable. I still don't really understand what under-the-hood limitations prevent simulink from applying implicit expansion (and if there is a way to directly circumvent those limitations), but I guess I may never know. Either way, my code will work at least.

[Jon]

Glad this at least helps a little. You could report this as a bug to MATLAB. One good thing about reporting the bug, is that if it is eventually fixed they send you a notification.

https://www.mathworks.com/support/contact_us.html

[Landen]

Thanks, I'll do that!