Easier way to build large nonlinear dynamic model and use control design

I am building a large nonlinear dynamic model and would like to use analysis tools that are in Simulink control design. However, creating the model using Simulink blocks is not a practical solution because there are so many calculations involved. Making small changes to the model is tedious and time-consuming when everything is in blocks.

I thought that converting the blocks into an Embedded Matlab function would make things easier, but several major problems were introduced. The main issue is that Embedded Matlab functions give a compiler error when I use any mathematical operator that potentially yields infinite or complex numbers (such as log, sqrt, etc) even though the values are always within realistic limits during simulation. I use a lot of these operators in my model, so I end up having to convert these functions back to Simulink blocks for everything to work.

What is the easiest way to do what I just described? If I stick with Simulink, is there a way to get past the compiler error when using operators such as log and sqrt? If possible, I want to avoid manually generating the linearized state space matrices (i.e. not using Simulink) unless its the only solution to my problem.