[MPICH] Configure Options for Alpha Linux cluster
Rajeev Thakur
thakur at mcs.anl.gov
Thu Aug 18 11:38:08 CDT 2005
You can use the p4 device and pass the option -comm=shared to configure as
described in
http://www-unix.mcs.anl.gov/mpi/mpich/docs/mpichman-chp4/node14.htm under
SMP Clusters. That will use shared memory within a node and TCP across
nodes.
Rajeev
> -----Original Message-----
> From: owner-mpich-discuss at mcs.anl.gov
> [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of David LeBard
> Sent: Wednesday, August 17, 2005 7:04 PM
> To: MPICH Mailing List
> Subject: [MPICH] Configure Options for Alpha Linux cluster
>
> Hi All,
>
> I am trying to properly compile MPICH 1.2.7 for Compaq Alphas running
> RedHat Linux 7.2. Although these are alpha computers and do
> have shared
> memory between 2 processors (ie. 2 procs/node), there is not shared
> memory between all cpus. Using the examples seen in configure's help
> and google, I set my device to p4 and my architecture to alpha (or
> LINUX_ALPHA, both with the same results) and when I run a test job it
> fails with the following error:
>
> 0 - MPI_INIT : MPIRUN chose the wrong device ch_p4; program
> needs device
> ch_shmem
>
> When I choose the device it is asking for (ch_shmem), all mpi
> tests will
> work fine but all mpi processes will sit on the same node
> regardless of
> what the nodefile states. I tried many times playing with
> the nodefile
> and mpirun's options (eg. -nolocal,
> -machinefile=File_With_Only_Non_Local_Nodes), but no matter
> what all the
> processes stay on the same node.
>
> If anyone has experience running mpich with this kind of
> system (cluster
> of dual processors alpha linux), please share with me your configure
> options.
>
> Thanks in advance,
>
> David LeBard
> Dept of Chemistry and Biochemistry
> Arizona State University
>
>
>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 2442 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20050818/b67a9442/attachment.bin>
More information about the mpich-discuss
mailing list