Re: amber-developers: fixup to max14

From: Volodymyr Babin <>
Date: Wed, 12 Nov 2008 17:24:22 -0500 (EST)

It hangs because the 'newbalance' (see below) does not get
communicated to other nodes. Does it make sense to add
mpi_allgather here or rather not? The relevant snapshot from
nonbond_list.f is:

      else if (balance) then

         ! --- If this was not a rebalance step, then check
         ! whether the list has gotten more than a tolerance
         ! away from the last balance. If so, trigger
         ! a new balance by setting trial and newbalance
         ! newbalance will need to be communicated to
         ! all processors before the next list build.

         if(listdiff > listdiffmax)then
         end if
      end if ! (trial)
Received on Fri Dec 05 2008 - 14:35:50 PST
Custom Search