1. From my reading of the text is seems clear that the comm passed to
MPI_Win_init() must be an intra-communicator but we don't seem to
make this explicit.
2. On page 9 line 16.
achieved by such checks has to be waited against the added
should read
achieved by such checks has to be weighed against the added
Now for some questions concerning the semantics/validity of certain call
sequences.
Consider the following scenario. All calls pertain to a
single window which we assume both processes have open.
Process 0 Process 1
--------- ---------
A. put(1) post(0)
start(1) wait
B. put(1)
complete
In process 0 after the complete it is my understanding that put B MUST
be complete at the origin but that this is not necessarily the case for
put A. Further the complete must not wait for put A to complete at the
origin before returning otherwise deadlock may occur.
Meanwhile in process 1 after the wait put B must be complete at the
target but this is once more not necessarily the case for put A.
Is this correct?
In the discussion of MPI_WIN_COMPLETE it is stated that
"Each call to MPI_WIN_START(...,win) has to be matched by a unique
following call to MPI_WIN_COMPLETE(win), and vice versa."
I am not quite sure what this means. For example does it make the
following illegal? Once again all calls pertain to a single window which
we assume both processes have open.
Process 0 Process 1 Process 2
--------- --------- ---------
A. start(1) post(0) post(0)
B. start(2) wait wait
C. complete
D. complete
I believe that it should be illegal but supposing it is not then one
would presume that complete C should match start B and complete D match
start A but I see nothing in the text that mandates this.
---nick.
-=-
Nick Nevin nevin@osc.edu
Ohio Supercomputer Center http://www.osc.edu/lam.html