use of win_ptr->fence_cnt in MPIDI_Win_fence()
|Reported by:||David Gingold <david.gingold@…>||Owned by:|
MPICH2 folks -- I'm trying to make sense of code that I cribbed from the ch3 implementation, and I think this bit in ch3u_rma_sync.c is likely wrong: if ((win_ptr->fence_cnt == 0) && ((assert & MPI_MODE_NOSUCCEED) ! = 1)) MPI_MODE_NOSUCCEED is 16384 so assert & MPI_MODE_NOSUCCEED is never 1. But even with that fixed to !(assert & MPI_MODE_NOSUCCEED), I'm puzzled by the logic around this. Asserting MPI_MODE_NOSUCCEED should only make the code path shorter, yet it seems to have the opposite effect here. What am I missing? -dg -- David Gingold Principal Software Engineer SiCortex Three Clock Tower Place, Suite 210 Maynard MA 01754 (978) 897-0214 x224
Change History (5)
Note: See TracTickets for help on using tickets.