Who's going to SC'10 this year? (/me raises hand) If you're there, be sure to
drop by and say hello (I'll be in and around the Cisco Booth).
Dr. George Bosilca (U. Tennessee Knoxville) and I will host our normal "Open
MPI State of the Union" BOF on Wednesday at 12:15pm. It's a great place to
It would be best if an IB vendor replies (hint hint!), but it is likely that
you have some kind of hardware issue on that node (e.g., a bad / flakey HCA,
etc.). You should probably run a full set of layer-0 diagnostics on your
fabric to make sure it's clean.
I say this because back when Cisco
Ouch; bad FAQ maintainer!
I've fixed it -- thanks for letting us know, Josh!
On Nov 5, 2010, at 3:52 PM, Joshua Bernstein wrote:
> Thanks Samuel,
>
> I should have checked ompi_info myself.
>
> The FAQ on the website should probably be updated to reflect this function
> change.
>
> -Joshua
What kind of network did you do these tests on?
Were both processes on the same machine or different machines?
On Nov 10, 2010, at 12:29 AM, Yiannis Papadopoulos wrote:
> Hello,
>
> I am using OpenMPI 1.4.1. I have a small test case that calls
> MPI_Test() too many times. I see one or two rand
On Nov 10, 2010, at 12:31 AM, tmish...@jcity.maeda.co.jp wrote:
> Recently I installed OpenMPI 1.4.3 on my cluster. I found that
> SYSTEM CPU is higher than older versions of 1.4.X, when I ran
> our FEM program. Further more, OMPI_1.4.3 is a little bit slower
> than 1.4.2. What is the differnece b
On Nov 10, 2010, at 4:37 PM, Jim Kusznir wrote:
> (working command):
> rpmbuild -bb --define 'install_in_opt 1' --define 'install_modulefile
> 1' --define 'modules_rpm_name environment-modules' --define
> 'build_all_in_one_rpm 0' --define 'configure_options
> --with-tm=/opt/torque --enable-contrib
Hi,
section 4.11
http://www.mpi-forum.org/docs/mpi-20-html/node54.htm
describes how memory allocation can be done in a Fortran program:
REAL A
POINTER (P, A(100,100)) ! no memory is allocated
CALL MPI_ALLOC_MEM(4*100*100, MPI_INFO_NULL, P, IERR)
! memory is allocated
...
A(3,5) = 2.71;
I think I have a solution which solve both issues:
- deal with the overlapping regions
- maintain the level of today's performance
Please try r24047.
george.
On Nov 11, 2010, at 06:46 , Jed Brown wrote:
> On Thu, Nov 11, 2010 at 12:36, Number Cruncher
> wrote:
> However as commented here
It is a quad core machine (Intel Core2 Quad CPU Q9550 - 64bit).
OpenMPI is the one that comes with Fedora 13.
I tested it on a cluster of Intel Xeon (32bit) with Gigabit Ethernet
with OpenMPI 1.3.3, the results are:
// with no timeout
Total time: 1.00309 secs
Estimated calls: 1116537
MPI_Test cal
> Assumedly you're running your tests on otherwise-idle machines. Is the
difference repeatable?
Yes, I ran the test on other-nodes. The difference is noticeable and almost
same.
I'm planing to do more simple test to make the problem easier. Then, I'll
send the results.
Does this mailing list acc
10 matches
Mail list logo
