Working with Persistent Variables

This example shows how to generate a MEX function from a MATLAB function, compute_average, that uses persistent variables. It illustrates that you must clear the state of persistent variables before using the function to compute the average of a new set of values.


There are no prerequisites for this example.

About the compute_average Function

The compute_average.m function uses two persistent variables, the accumulated sum and the number of values added so far, so that you can call the function with one value at a time.

type compute_average
% y = compute_average(x)
% This function takes an input scalar value 'x' and returns the average
% value so far.
function y = compute_average(x) %#codegen
assert(isa(x,'double')); % Input is scalar double

% Declare two persistent variables 'sum' and 'cnt'.
persistent sum cnt;

% Upon the first call we need to initialize the variables.
if isempty(sum)
    sum = 0;
    cnt = 0;

% Compute the accumulated sum and the number of values so far.
sum = sum + x;
cnt = cnt + 1;

% Return the current average.
y = sum / cnt;

The %#codegen directive indicates that the MATLAB code is intended for code generation.

Generate the MEX Function

First, generate a MEX function using the command codegen followed by the name of the MATLAB file to compile.

codegen compute_average

By default, codegen generates a MEX function named hello_world_mex in the current folder. This allows you to test the MATLAB code and MEX function and compare the results.

Run the MEX Function

(10 + 20 + 100) / 3 = 43.3333

ans = 10
ans = 15
ans = 43.3333

Clear the Internal State of Persistent Variables

Clear the persistent variables by using the clear mex command.

clear mex

Run the MEX Function Again to Calculate the Average of a Different Set of Values

(10 + 20 + 30 + 40) / 4 = 25

ans = 10
ans = 15
ans = 20
ans = 25