LAM/MPI creates completely new sockets after the restart rather than
expecting BLCR to handle this.
It is my understanding that it simply (re)invokes the same code that is
used in MPI_Init() to build the original sockets.
Please note that there is no longer any active development on LAM/MPI and
that the developers who are still active in the MPI community have
(nearly?) all moved to work on the Open MPI implementation instead.
Additionally, the Open MPI implementation on BLCR support uses features
added to BLCR more recently than what was available when LAM/MPI's support
was implemented (making it a better example to learn from).
-Paul
On Thu, Oct 4, 2012 at 2:32 PM, researcher researcher <
researcherfr_at_[hidden]> wrote:
> Hello,
>
> Please i ask how we ckeckpoint parallel applications in lam-mpi? i read
> that we use blcr but blcr does not save the sate of sokets ?
> Could you explain me please how we retart our applications ?
>
>
> Thank you so much.
> BestRegards.
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
--
Paul H. Hargrove PHHargrove_at_[hidden]
Future Technologies Group
Computer and Data Sciences Department Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory Fax: +1-510-486-6900
|
