Create a codegen implementation of chol function where the size of the matrix is not hard-coded.

I want a C implementation for chol() function. Using MALTAB code gen app I am able to get the C implementation but in this implementation the size of the matrix is hardcoded. How can I get a C implementation in which the size of the matrix is not hard-coded and can be decided during runtime.

The script I am using to call cholesky goes like this:

rows = input('Enter the number of rows for the matrix: ');

cols = input('Enter the number of columns for the matrix: ');

% Create a random matrix of size n

A = rand(rows, cols);

A = A'*A;

%Using the cholesky function

outputMat = cholesky(A);

I have tried this approach. Will this work? Is this even possible? If yes, then how do I go about this?

Answer 1

[Denis Gurchenkov]

Yes this can be done. When you use MATLAB Coder ('codegen' command), you need to decide what are the input arrays that you feed to the entry-point function. Specifically, do you want those arrays to be fixed in size (this is the default behavior), do you want them to be variable-sized but bounded by some maximum size (say, any size up to 20x20), or variable-sized unbounded.

Depending on that decision, the C code would look different, and it would use different types to represent those arrays.

Here is a small example that shows how to generate C code for chol, with all those 3 possible input types.

First, here is the "entry point" function, mychol.m:

%#codegen

function out = mychol(a)

out = chol(a);

end

Here is the doit script that calls codegen command 3 times, with different inputs:

% a is an unbounded variable-sized 2-d matrix of doubles

% b is a variable=sized 2-d matrix bounded by 20x20

% c is a fixed-sized 2-d matrix

a = coder.typeof(1, [Inf Inf], [true true]);

b = coder.typeof(1, [20 20], [true true]);

c = coder.typeof(1, [20 20], [false false]);

codegen mychol -args {a} -d mychol_unbounded -config:lib -report

codegen mychol -args {b} -d mychol_bounded -config:lib -report

codegen mychol -args {c} -d mychol_fixsize -config:lib -report

After your run it, look at ./mychol_bounded/mychol.c, ./mychol_unbounded/mychol.c, ./mychol_fixsize/mychol.c and compare the inputs to main function. If you want to see an example on how to call the main function from your own C code (how to create those emxarray arrays), look at the "examples/main.c" in each of the folders.

Here is a link to some documentation you can use for further details: https://www.mathworks.com/help/coder/ug/what-is-variable-size-data.html