Fixedpoint conversion, abstracting datatypes away from the model-file

Hi

There exist a nice video/document about best-practice when converting M-code to fixed-point Fixedpoint conversion of M-code

Here it is nicely described how you can abstract the actual datatype away from your algorithm-code leaving it pure and target-independent (i.e. its in double). However I can not find any experience on doing the same thing (which is desireable for me and our design-platform) in Simulink block-diagrams. The fixedpointtool, videos and guidelines simply "hardcode" the datatypes into the individual blocks.

Does anybody know some best-practice on this ?

kr Erik