[mpich-discuss] Missing Nemesis ib

Halim Amer aamer at anl.gov
Wed Sep 20 11:11:59 CDT 2017


I seems you have mismatch in the OFED stack. Try installing the Mellanox 
OFED stack if you are using the bundled OFED stack right now.

Make sure MXM works before trying MPICH. Use the mxm/bin/mxm_perftest 
from your MXM installation to test that things work properly. If it 
doesn't work, then contact your admin or Mellanox cause it is not an 
MPICH problem.

Halim
www.mcs.anl.gov/~aamer

On 9/19/17 7:14 AM, Jason Collins wrote:
> Thank you very much.
> 
> I have compiled with "CH3:nemesis:mxm". The compilation was successful.
> 
> Now I have a new problem. I tested the test "./icp" and get the 
> following error.
> 
> # mpiexec -f hosts -n 4 ./cpi
> [1505822776.546898] [compute1:16212:0] sys.c:744 MXM WARN Conflicting 
> CPU frequencies detected, using: 3459.84
> [1505822776.546898] [compute1:16213:0] sys.c:744 MXM WARN Conflicting 
> CPU frequencies detected, using: 3459.84
> [1505822776.546951] [compute1:16216:0] sys.c:744 MXM WARN Conflicting 
> CPU frequencies detected, using: 3459.84
> [1505822776.547039] [compute1:16214:0] sys.c:744 MXM WARN Conflicting 
> CPU frequencies detected, using: 3459.84
> [1505822776.561357] [compute1:16214:0] ib_dev.c:533 MXM WARN failed call 
> to ibv_exp_use_priv_env(): Function not implemented
> [1505822776.561371] [compute1:16214:0] ib_dev.c:544 MXM ERROR 
> ibv_query_device() returned 38: Function not implemented
> [1505822776.561386] [compute1:16218:0] ib_dev.c:533 MXM WARN failed call 
> to ibv_exp_use_priv_env(): Function not implemented
> [1505822776.561396] [compute1:16218:0] ib_dev.c:544 MXM ERROR 
> ibv_query_device() returned 38: Function not implemented
> [1505822776.561426] [compute1:16225:0] ib_dev.c:533 MXM WARN failed call 
> to ibv_exp_use_priv_env(): Function not implemented
> [1505822776.561442] [compute1:16225:0] ib_dev.c:544 MXM ERROR 
> ibv_query_device() returned 38: Function not implemented
> Fatal error in MPI_Init: Other MPI error, error stack:
> MPIR_Init_thread(474).........:
> MPID_Init(190)................: channel initialization failed
> MPIDI_CH3_Init(89)............:
> MPID_nem_init(320)............:
> MPID_nem_mxm_init(158)........:
> MPID_nem_mxm_get_ordering(464): mxm_init failed (Input/output error)
> 
> 
> El vie., 15 sept. 2017 a las 16:01, Halim Amer (<aamer at anl.gov 
> <mailto:aamer at anl.gov>>) escribió:
> 
>     The "nemesis:ib" netmod does not exist anymore. Try "ch3:nemesis:mxm"
>     with a dependency on Mellanox's MXM library (can be obtained from the
>     HPCX package at www.mellanox.com/products/hpcx
>     <http://www.mellanox.com/products/hpcx>) or "ch3:nemesis:ofi"
>     with a dependency on libfabric (which would be built to support the IB
>     or MXM providers; see https://ofiwg.github.io/libfabric/).
> 
>     Halim
>     www.mcs.anl.gov/~aamer <http://www.mcs.anl.gov/~aamer>
> 
>     On 9/15/17 4:20 AM, Jason Collins wrote:
>      > Hello everyone.
>      >
>      > Recently, I downloaded Mpich-3.2
>      >
>      > I want to configure with support for InfiniBand. I've put the
>     following
>      > command:
>      >
>      > # ./configure --prefix=/my/path --with-device=ch3:nemesis:ib
>      >
>      > And I get the following error:
>      >
>      > configure: error: Network module ib is unknown
>      > "./src/mpid/ch3/channels/nemesis/netmod/ib"
>      >
>      > When I check the path I confirm that in the folder "netmod" does not
>      > exist the folder "ib". How can this be solved?
>      >
>      > Many thanks.
>      >
>     <https://audio1.spanishdict.com/audio?lang=en&text=when-i-check-the-path-i-confirm-that-within-the-folder-netmod-the-folder-does-not-exist-ib-how-can-this-be-solved-many-thanks>
>      >
>      >
>      > _______________________________________________
>      > discuss mailing list discuss at mpich.org <mailto:discuss at mpich.org>
>      > To manage subscription options or unsubscribe:
>      > https://lists.mpich.org/mailman/listinfo/discuss
>      >
>     _______________________________________________
>     discuss mailing list discuss at mpich.org <mailto:discuss at mpich.org>
>     To manage subscription options or unsubscribe:
>     https://lists.mpich.org/mailman/listinfo/discuss
> 
> 
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
> 
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss


More information about the discuss mailing list