Import constant array in s-function - Avoid repetitive import of array in each step

In a level-2 c s-function, at each sampling time I would like to access part of a rather large but constant matrix. To my understanding, there are two approaches I could use, where both are calling this matrix in the mdlOutputs function.

1) Import it as an input. I think that in this case I will increase the computational time because I have to import it repeatedly in every sampling step.

2) Import it as a parameter. In this case I still have to read it in the mdlOutputs, so I do not think that I will achieve anything better compared to the previous approach (I might be wrong though).

Is there any way to avoid first loading the whole matrix and then picking the part that I need to process? And in general, which would be the most efficient way to implement it?

[Walter Roberson]

I seem to recall there is a model initialization phase. You could possibly import the matrix during that phase, and store it into a static array.

[jpro89]

Indeed there is one. But to do so, the approach that I have in mind is to copy the matrix itself in the s-function script, which is not really possible due to the big size of the matrix.

[Walter Roberson]

engGetVariable during the initialization phase?

If you find it necessary to import it as a "signal" then the signal could be set to the full value at time 0 and could be set empty (or trivial) at all other times and you would ignore the value except during initialization.