There was very little response to these notes, so I'm moving forward as per
the initial mailings. Here is what was concluded - holler if you have a
On 12/4/07 8:09 AM, "Ralph H Castain" <rhc_at_[hidden]> wrote:
> Yo all
> As (I hope) many of you know, we are in a final phase of revamping ORTE to
> simplify the code, enhance scalability, and improve reliability. In working
> on this effort, we recently uncovered four issues that merit broader
> discussion (apologies in advance for verbosity). Although these somewhat
> relate, I realize that people may care about some and not others. Hence, to
> provide the chance to only comment on those you -do- care about, and to at
> least somewhat constrain the length of the emails, I will be sending out a
> series of four emails in this area.
> The issues will include:
> I. Support for non-MPI jobs
We will maintain the support to launch (both at mpirun time and dynamically)
non-MPI jobs. We will not require a command-line switch warning us that
"this is not an MPI job" - this maintains the existing capability.
We will add an MPI_Info key to indicate that the comm_spawned job will not
be calling MPI_Init so we don't attempt to modex with it.
> II. Interaction between the ROUTED and GRPCOMM frameworks
We will leave this "as-is" - so if you select a pairing that clashes (e.g.,
the default grpcomm and a ring routing algo), your startup (including OpenIB
wireup) performance will stink. User beware.
> III. Collective communications across daemons
Nobody has time to work on this one. I will insert hooks where these
collectives can go, and may play with at least one implementation.
> IV. RTE/MPI relative modex responsibilities
We are still wrestling with this one - hope to resolve it sometime soon.
> Please feel free to contact me and/or comment on any of these issues. As a
> reminder, if you do comment back to the Devel mailing list, please use
> "reply all" so I will also receive a copy of the message.