[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
|
Re: [ptp-dev] SCI contribution
|
Jay,
I'd say it provides a subset of MRnet functionality, but is written in C, so avoids messy C++ dependencies. It's also a subset of STCI, but is available, works, and is supported.
There is an effort underway to develop a common API for this type of infrastructure, and SCI could provide the basis for one implementation. It seems to me that having multiple implementations is probably a good thing. Also, if it's part of PTP then we have more control over ensuring that it meets our requirements, as well as being able to distribute it as part of a PTP installation (which we couldn't do for MRnet or STCI).
Cheers,
Greg
On Jan 21, 2010, at 4:48 PM, Jay Alameda wrote:
> Greg,
>
> This is intriguing, however, would it be possible to understand how this
> relates/compares/etc to other existing efforts such as MRnet
> (http://www.paradyn.org/mrnet/) or STCI (http://stci.wikidot.com/)? I'm
> not an expert in any of them, but I can definitely see the need for a good
> communication infrastructure...
>
> Jay
>
>
> -----Original Message-----
> From: ptp-dev-bounces@xxxxxxxxxxx [mailto:ptp-dev-bounces@xxxxxxxxxxx] On
> Behalf Of Greg Watson
> Sent: Thursday, January 21, 2010 3:16 PM
> To: Parallel Tools Platform general developers
> Subject: [ptp-dev] SCI contribution
>
> Hi all,
>
> IBM is interested in contributing some communication infrastructure
> software (called "SCI") that could be used by PTP for scalable
> communication on large clusters of distributed processors. One component
> of PTP that would immediately benefit from this is the parallel debugger,
> which could use the infrastructure to scale to much larger systems than it
> does currently, and is something that has been in need of improvement for
> some time (i.e. on my "wish list"). SCI is pure C code, and would replace
> a significant portion of the SDM, which is also written in C. There may be
> other components of PTP that could also benefit from this code, such as
> for performance analysis or other kinds of data collection.
>
> I'm interested in feedback from the PTP community as to whether this
> contribution would be useful and appropriate for the PTP project.
>
> Thanks,
> Greg_______________________________________________
> ptp-dev mailing list
> ptp-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/ptp-dev
>
> _______________________________________________
> ptp-dev mailing list
> ptp-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/ptp-dev
