[mvapich-discuss] No asynchronous progress in mvapich2.

[sreeram potluri]

Hi Ziaul,

1) Truly one-sided designs for passive (lock-unlock) semantics are
currently available at the intra-node level only in MVAPICH2. Please look
at the following pointers for information about two modes that offer this:

LiMIC based windows
http://mvapich.cse.ohio-state.edu/support/user_guide_mvapich2-1.8rc1.html#x1-560006.4

Shared-memory based windows
http://mvapich.cse.ohio-state.edu/support/user_guide_mvapich2-1.8rc1.html#x1-570006.5

2) We have better RDMA-based implementations for one-sided communication
using active synchronization semantics (post-wait/start-complete, fence).
These provide better overlap performance. You can explore these.

3) Future release will have support for truly one-sided designs for passive
(lock/unlock) at the inter-node level

Hope this information is useful. Let me know if you need any further help.

Sreeram Potluri

On Sat, Apr 14, 2012 at 7:36 PM, Ziaul Haque Olive <mzh.olive at gmail.com>wrote:

>
> Hello,
>
> I ran a mpi (one-sided communication) program with 2 processes. process 1
> is in infinite loop and process 0 is trying to put some values on process
> 1's window, using lock/unlock mechanism. finally process-0 calls MPI_Abort.
> But the program does not stop running. Following is my code snippet,
>
>  29        if( id == 0 ){
>  30
>  31                 MPI_Win_lock( MPI_LOCK_EXCLUSIVE, 1, 0, win );
>  32                 MPI_Put( send_buf, 10, MPI_INT, 1, 0, 10, MPI_INT, win
> );
>  33                 MPI_Win_unlock( 1, win );
>  34
>  35                 MPI_Abort(MPI_COMM_WORLD, 1);
>  36         }
>  37
>  38         else if( id == 1 ){
>  39
>  40                 while(1){
>  41                         i = 1;
>  42                 }
>  43         }
>
> It seems process-0 is stuck in lock/unlock section of the code.
>
>
> Thanks,
> Ziaul
>
>
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20120414/5754b03e/attachment.html