Hi All,
For those who were not able to attend the Forum meeting in Chicago last
week, this is an update to inform you about the status of the
nonblocking collective I/O ticket.
We decided to add nonblocking version of the collective I/O routines to
access data with individual file pointers and explicit offsets (and
deprecate their split collective counterpart) but not for shared file
pointers. The semantics for nonblocking collective shared file pointers
were just too weird to have a majority of the forum agreeing upon. Since
we don't really have a use case for them and in order stay in the MPI
3.0 bracket we decided to leave them out of this ticket.
The current pdf attached to the ticket is the draft that was read at the
last meeting and passed its formal reading. We will vote on this ticket
at the next meeting in Japan.
Thanks,
Mohamad
