On 06/24/2013 10:01 AM, Jeff Squyres (jsquyres) wrote: > Put differently: A's FINALIZE is collective with B/first FINALIZE (by > definition), but in practice it may need to synchronize with B/second > FINALIZE. Good point. Would requiring MPI to not synchronize when it's not finalizing fix this? -- Pavan Balaji http://www.mcs.anl.gov/~balaji