This explains everything Botpro, in addition to C++ please take a look at something called R you will be enlightened by it, the best part you won't have to reinvent the wheel (in particular with regards to stochastic processes and such).
For debugging it should be great but when would try use it for generating tick data would it work as good as 30 sec/minute bars or will be differences on larger scale if generating higher resolution?
For those who want to go further, I recommend some more complex models like Ornstein-Uhlenbeck(best for mean reverting regimes). Heston model is a very interesting one but be aware that is next to impossible to calibrate Heston if you have only the time series of the given instrument. You need the option prices to calibrate it correctly. Maybe add a jump process too to simulate a spike in the price.