I noticed that my old scheme of process placement seems to be have a different outcome with the current master. I used to start a single process per node on my cluster using a hostfile together with “map-by node”, with a command line that looked like this “mpirun —hostfile arc —map-by node -np 2 hostname”.
Assuming my host file contains arc00 slots=20 arc01 slots=21 The above command in 1.10 gives the expected: arc00 arc01 However on master it gives: arc00 arc00 It seems that “—map-by node” was completely ignored, and to get back the same behavior I have to force replace it with -npernode 1 ? Also my processes, despite the fact that I asked for 1 per node, are not bound to the first core. Shouldn’t we release the process binding when we know there is a single process per node (as in the above case) ? George.
