On Aug 17, 2005, at 10:39 AM, Jeff Squyres wrote:
A similar situation is true for MPI_REDUCE_SCATTER -- PMB is passing in
an array of recvcounts that are all 0.
I'll do 2 things for this:
1. Mail the PMB guys and tell them about this problem
--
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/
Even more on this...
Pallas was absorbed by Intel. So the latest version is not PMB 2.2,
it's the IMB 2.3 (Intel MPI Benchmark, which you can download from
http://www.intel.com/cd/software/products/asmo-na/eng/cluster/mpi/
219848.htm).
IMB 2.3 still does these 0 byte reduces, so I'll mail them about it.
But there are bugs in PMB 2.2 that cause a no-op 0-byte Reduce_scatter
to invoke a seg fault in the PMB code base. These bugs are fixed in
IMB 2.3.
--
{+} Jeff Squyres
{+} jsquy...@lam-mpi.org
{+} http://www.lam-mpi.org/
