How to use TMVA with several cores/servers


I am quite ignorant of the details of TMVA, so this might be asking very simple questions. I couldn’t, however, get clear answers by looking into the TMVA page and in the forum.

I wonder if TMVA algorithms are using all the cores in a machine to do the proccessing. If the answer is negative, what is preventing it from doing so, design, lack of implementation, …? If the answer is positive, then, do I have to specify something on my side (for example the number of cores/threads to use)? And, even further, is there a (hopefully easy) way to use cores in more than one machine?