[CFD_D_BOY]

Hi, recently i run the case using mpich mode (it sometimes leaves some "semaphores" or "shared memory segments" hanging on some nodes), and i need to clean up them each time.

what's the advantage of each parallel calculation mode in CFX?

Although the CFX tutorial suggetion HP MPI for on Linux and Windows if you want to utilize high speed interconnects (such as Myrinet, Infiniband, and so on) on those platforms.