Actually, what akvadrako described is just called multiprocessing - not SMP.

SMP refers specifically to systems with uniform latency to shared memory. So any distributed memory topology like ccNUMA w/ interconnect would not fit that definition.