Embedded coder - how to control types of temporary variables within embedded Matlab function blocks

I'm generating code from simulink using embedded coder. In my simulink model I have an embedded matlab function containing the matlab command eig. When I try to compile the generated c-code in microCforPic32 I get errors on out of range numbers. In matlab double precision is standard, but my target processor is 32 bit, so I guess I would like to use single precision but dont see how I can set this for the temporary variables generated from matlabs own functions such as eig(). Below is part of the generated code:

/* Function for MATLAB Function: '<S17>/Eig' */ static void Calib_eml_matlab_ztgevc(const creal_T A[4], creal_T V[4]) {...

Heres one of the lines within this function that cause troubles y = anorm; if (2.2250738585072014E-308 > anorm) { y = 2.2250738585072014E-308; }

The tpe of y is defined in top of the function

real_T y;

Where real_T stands for double 64-bit, i'm pretty sure.

I've tried to set the largest atomic size to float but does not change the value 2.2250738585072014E-308. I could of course change it mannualy in the c-code but I would prefer to solve it in matlab.

[Walter Roberson]

I wonder if you should be using the fixed point toolbox ?

[Rebecka]

I had a look, but when running the fixed point adviser, I get that the block containing the eig function does not support fixed point data types