According to the MPI Standard, "only arguments sendbuf, sendcount,
sendtype, root, and comm are significant" on non-root processes in
MPI_Gather. However, in routine "ompi_coll_tuned_gather_intra_binomial"
the value of "rdtype" ("recvtype" of MPI_Gather) is used. This leads to
crash if an u
Looks like a bug to me. I'll file a ticket.
Thanks for the heads up!
On Feb 17, 2010, at 3:59 AM, Michael Hofmann wrote:
> According to the MPI Standard, "only arguments sendbuf, sendcount,
> sendtype, root, and comm are significant" on non-root processes in
> MPI_Gather. However, in routine
Thanks Michael. This should be now fixed by commit r22639. I'll push it into
the next release.
Thanks,
george.
On Feb 17, 2010, at 10:10 , Jeff Squyres wrote:
> Looks like a bug to me. I'll file a ticket.
>
> Thanks for the heads up!
>
> On Feb 17, 2010, at 3:59 AM, Michael Hofmann wro
I'll defer to the Mellanox guys to reply more in detail, but here's a few
thoughts:
- Is MVAPICH using XRC? (I never played with XRC much; it would surprise me if
it caused instability on the order of up to 100 micros -- I ask just to see if
it is an apples-to-apples comparison)
- The nRepeat
On Feb 16, 2010, at 6:01 AM,
wrote:
>I am using PGI compiler version 7.2.4, in my Linux cluster, configured
> with MPICH1. I tried to configure OPENMPI-1.3.3 with this PGI compiler,
> but with no success. While performing #make it gives the following
> error:
Hmm -- odd. I can build Open
I usually prefer the expanded notation:
unsigned char ret;
__asm__ __volatile__ (
"lock; cmpxchgl %3,%4 \n\t"
" sete %0 \n\t"
: "=qm" (ret), "=a" (oldval), "=m" (*addr)
: