I have a wind farm simulation model which computes among others wind speeds at millions of grid nodes in a three-dimensional volume. The model mainly consists of c s-functions.
When increasing the number of nodes in a uniform grid - basically a three-dimensional matrix - further and further, at some point I get this message:
Simulink supports total block I/O buffer size up to 2147483647 number of bytes. The total block I/O size is calculated based on data type size, number of elements, and signal type (real/complex) of all signals in the model. Consider reducing signal sizes before compiling or simulating the model.
How is the "total block I/O buffer size" exactly determined? I'd love to see a more precise definiton of "The total block I/O size is calculated based on data type size, number of elements, and signal type (real/complex) of all signals in the model."
Let's assume there are two c s-functions (mex). Will all these things contribute to the "total block I/O buffer size":
- number of inputs and outputs
- dimensions and size of inputs and outputs
- dialog parameters
- internal DWork
- internal variables (int_T, real_T, real32T, vectors, pointers, ...)
- function step size (I have a multi rate model)
- number of computations/functions per time step
As it says "total block I/O buffer size", I would assume only and exclusively dimensions and size of block inputs/outputs within the simulink model are considered. Applying this logic to a simple Simulink model configuration as shown here:
I would guess the "total block I/O buffer size" is the sum of:
- 1000 (from C1)
- 1000 (to c s-function input)
- 200x50x20 (from C2)
- 200x50x20 (to gain)
- 200x50x20 (from gain)
- 200x50x20 (to c s-function)
- 3 times 700x50x20 (from c s-function)
- 3 times 700x50x20 (to "To Workspace" block)
ist that correct? Or is there some redundancy/duplication in this approach? Or did I miss something?
