We introduce a population genetics model in which the operatorsare effectively computable -- computable in polynomial time onProbabilistic Turing Machines. We shall show that in this modela population can encode easily large amount of informationfrom enviroment into genetic code. Then it can process theinformation as a paralel computer. More precisely, we show thatit can simulate polynomial space computations in polynomiallymany steps, even if the recombination rules are very simple.