Different costs on different runs of quadprog

Hello, I am solving an optimization problem using quadprog in Matlab. I observe that the value of the optimization function is different for different runs (without changing any paramters ofcourse), which results into different controls being generated. Although the difference is not huge, the difference in costs is aorund 0.2 %

What would be the reason behind this and is this normal?

I could add the code if required.

[Ramnarayan Krishnamurthy]

Can you please provide the data and the code to reproduce this behavior? There is a chance that a part of the code is not deterministic. Either way, it would be useful to reproduce this behavior.