Dear Urbain,
Like any other program, Gama models do not have "a way" to speed up execution. The first cause of slow execution is due to the choices in the implementation. Hence, the answer to your question is model specific. As a consequence, you'll have to find yourself which part of your code take the most execution time; to do that, use the machine_time global variable to track execution time, from start to end of procedure : store machine_time at the beginning of a procedure (actions or reflexes) and at the end, then subtracts the latest with the first to have execution time in millisecond. Once you have done benchmarking, you can get back here with specific code part to optimize.
Best,
Kevin