[mvapich-discuss] Broadcast using infiniband's hardware multicast

Dhabaleswar Panda panda at cse.ohio-state.edu
Wed Mar 14 10:25:30 EDT 2012


Hi,

Thanks for your note. Several years back. the hardware multicast-based
approach was not scalable for collectives, compared to multi-core-based
approach. That's why this design from MVAPICH was not carried out to
MVAPICH2. Instead, multi-core-based approach was stressed in the MVAPICH2
series.

Hope this helps.

Thanks,

DK

On Wed, 14 Mar 2012, NARDI Luigi wrote:

> Hello,
>
> I'm interested in using the MPI-level broadcast using infiniband's
> hardware multicast functionality with MVAPICH2.
>
> I have seen nice results in some papers of years 2005 and 2006.
>
> May I know what is the implementation state of the art on this work?
>
> I have seen that MVAPICH 0.9.9 User and Tuning Guide
> (https://computing.llnl.gov/tutorials/linux_clusters/mvapich_user_guide_
> 099.pdf
> <https://computing.llnl.gov/tutorials/linux_clusters/mvapich_user_guide_
> 099.pdf> ) in 2007 talks about this functionality
>
> but nothing is said about that on MVAPICH2 1.8 User Guide
> (http://mvapich.cse.ohio-state.edu/support/mvapich2-1.8_alpha2_user_guid
> e.pdf
> <http://mvapich.cse.ohio-state.edu/support/mvapich2-1.8_alpha2_user_guid
> e.pdf> ) in 2012.
>
> Thanks in advance
>
> Luigi Nardi
>
> *******************************
> This e-mail contains information for the intended recipient only. It may contain proprietary material or confidential information. If you are not the intended recipient you are not authorised to distribute, copy or use this e-mail or any attachment to it. Murex cannot guarantee that it is virus free and accepts no responsibility for any loss or damage arising from its use. If you have received this e-mail in error please notify immediately the sender and delete the original email received, any attachments and all copies from your system.
>



More information about the mvapich-discuss mailing list