Unit delay block just have a state variables which stores the previous value of input.
Similarly in Stateflow, you can use a local variable to store the previous value of data. You can use one parallel state with priority-1 which handles all your normal operation and another parallel state with priority-2 which saves the data to previous local variable.
See the Pseudocode below (unfortunately I don't have access to any image hosting website. Thats why I am unable to post any image to explain implementation in stateflow)
/* Parallel state-1 for Operation using previous value */
Out = data + data_prev;
/* Parallel state-2 for saving previous value */
data_prev = data;
But, I can not understand how a unit delay block is affecting your code optimization.
