"Kirk, Benjamin (JSC-EG311)" writes:
> Jed, does the PETSc team have any experience with the NBC algorithm
> deadlocking with OpenMPI-1.7?
>
> I know they don't officially advertise MPI_Ibarrier(), but it is there...
I've had problems with other things in the past, but I don't recall this
speci
Jed, does the PETSc team have any experience with the NBC algorithm deadlocking
with OpenMPI-1.7?
I know they don't officially advertise MPI_Ibarrier(), but it is there...
See https://github.com/benkirk/mpi_playground.git - I've implemented the
nonblocking consensus & it seems to work reliably
On Sep 28, 2013, at 7:59 PM, Jed Brown
wrote:
> "Kirk, Benjamin (JSC-EG311)" writes:
>
>> MPI_Win_create (&recv_buf[0], recv_buf.size()*sizeof(unsigned int),
>> sizeof(unsigned int),
>>MPI_INFO_NULL, MPI_COMM_WORLD,
>>&win);
>
> In practice, this is implemen
"Kirk, Benjamin (JSC-EG311)" writes:
> MPI_Win_create (&recv_buf[0], recv_buf.size()*sizeof(unsigned int),
> sizeof(unsigned int),
> MPI_INFO_NULL, MPI_COMM_WORLD,
> &win);
In practice, this is implemented using something like an MPI_Allreduce
with 3*int*comm_s
So I finally started to look into these… Pretty powerful for the unknown
communication partner problem.
Check out the attached. Note that the MPI_Put() is inside a loop over all
processors, but it need not be.
There are a lot of places we *know* we don't have any data to send to another
proce
