Dear Dirk, many thanks for your reply,
>On 17 October 2008 at 12:42, Simone Giannerini wrote: >| Dear all, >| >| I managed to install successfully Rmpi 0.5-5 on a quad opteron machine (8 >| cores overall) running on OpenSUSE 11.0 and Open MPI 1.5.2. >| >| this is what I get >| >| > library(Rmpi) >| [gauss:24207] mca: base: component_find: unable to open osc pt2pt: file not >| found (ignored) >| libibverbs: Fatal: couldn't read uverbs ABI version. >| -------------------------------------------------------------------------- >| [0,0,0]: OpenIB on host gauss was unable to find any HCAs. >| Another transport will be used instead, although this may result in >| lower performance. >| -------------------------------------------------------------------------- > >I am surprised that your googling did lead to you stumbling on dozens of >posts on this telling you that the config file > > /etc/openmpi/openmpi-mca-params.conf (location for Debian etc) > >can be changed to explicitly setting btl to 'no openib' as in > > # Disable the use of InfiniBand > # btl = ^openib > btl = ^openib > thanks, I had also read some of your previous posts but for sure I should have tried harder. >which will suppress the warning by suppressing the load of IB. Better still, >newer Open MPI release do this by default. > I managed to upgrade to version 1.2.6 that has that option enabled by default. Unfortunately though, it only suppresses the following warning libibverbs: Fatal: couldn't read uverbs ABI version. while the others remain. In any case as I pointed out in my previous post the issue is that only two threads are spawn per session. Maybe I should try to check whether Ompi is working properly before getting to Rmpi. I will try to work it out, of course any suggestions would be heartily welcome. ciao Simone >| I have searched the archives and found that the following suggestion was >| given for a similar problem: >| >| > Open MPI has Infiniband module compiled but there is no IB device found >| > on your host. Try to add "--mca btl ^openib" string to your command >| > line. > >That's one way of suppressing it, but not the only one. > >| Since I am not calling mpi directly but through Rmpi I do not know where to >| put that flag, I might contact the Rmpi mantainer, in any case, I would be >| grateful if you had further suggestions. > >There is nothing Rmpi can do there so contacting Dr Yu, while generally a >good idea with actual Rmpi issues, is not really advised here. > >Cheers, Dirk -- ______________________________________________________ Simone Giannerini Dipartimento di Scienze Statistiche "Paolo Fortunati" Universita' di Bologna Via delle belle arti 41 - 40126 Bologna, ITALY Tel: +39 051 2098262 Fax: +39 051 232153 http://www2.stat.unibo.it/giannerini/ ______________________________________________________
