Sparse matrix construction in simulink

Hi,

Background: In a simulink model I want to construct and solve a linear system. My matrix is sparse and I want to exploit this property. Hence I want to construct a sparse matrix and solve it. I want to do this without calling an extrinsic matlab function, because this slows down the simulation and, also, I need to use the solution of the linear system back in my simulation. My approach is to create a c/c++ function that does the job and then mex it. This way I hope to use it in my simulink model like any other library block.

Current situation: I already have two m functions that construct the matrix and solves it respectively. I just need to recreate this function in c/c++. In those functions, however, I use cell arrays and sparse matrix operations. Therefore I cannot use Matlab coder to auto-generate code for me. I guess I need to it the hard way by making everything in c/c++!

My problem: I am not very familiar with c/c++. I have also never created any "mexable" c/c++ code. Right now I am trying to build small snippets of my code in c and then compile it. I want to create the following c-version of this function:

[p] = test(Fx,c)

p = cell(size(Fx,1),1);
for i=1:size(Fx,1);
  p{i} = Fx{i}*c;
end

NOTE: Fx is a sparse matrix!

How does one do this? How should the c code look like so that it can readily be converted into a mex function?

- Nithin

Answer 1

[Kaustubha Govind]

You'll probably need to write this directly as a MEX-function because there's no C equivalent to a cell-array. See Passing Structures and Cell Arrays to find out how to use cell-arrays in MEX-functions.

In particular, the examples phonebook.c and explore.c in $matlabroot/extern/examples/mex are a good place to start learning about using cell-arrays.