[mvapich-discuss] 0.9.8 build error
Steve Jones
stevejones at stanford.edu
Wed Nov 1 14:42:21 EST 2006
Thanks DK. I'm trying to test with all possible configurations, one is using
the latest topspin driver with 0.9.8. Fortunately we're able to build and
test with it now.
Thanks for the response.
Steve
Quoting Dhabaleswar Panda <panda at cse.ohio-state.edu>:
> Steve,
>
> It looks like our e-mails crossed.
>
> Glad to know that you got help from Cisco and the problem has been
> resolved. Thanks for sharing the fix with the list.
>
> I would also like to mention that if you use the Gen2 interface of
> MVAPICH, you will get more features, better performance and better
> scalability.
>
> Thanks,
>
> DK
>
>
> > hi.
> >
> > here's the issue and fix provided by cisco.
> >
> > the topspin vapi.h in /usr/include/topspin/vapi/vapi.h defines
> > VAPI_VERSION_CODE as 0x40000000 (for vapi 4.0.0). The mvapich 098
> > mpid/vapi/viainit.c file that shows these missing definition errors
> > includes them under
> > #if VAPI_VERSION_CODE >= VAPI_VERSION(4,1,0)
> >
> > changing this in viainit.c to
> > #if VAPI_VERSION_CODE >= VAPI_VERSION(4<<12,1,0)
> >
> >
> > the build now completes.
> >
> > thanks.
> >
> > steve
> >
> >
> > Quoting Abhinav Vishnu <vishnu at cse.ohio-state.edu>:
> >
> > > Hi Steve,
> > >
> > > Thanks for trying out MVAPICH-0.9.8 and letting us know of the error
> > > which you are facing with the topspin stack.
> > > >
> > > > i'm encountering a build error for 0.9.8 using gnu/intel compilers
> > > (haven't
> > > > tested others) in the "make mpir in directory mpid/vapi" stage of
> the
> > > > build. i've included the configure script and error message below,
> > > along
> > > > with system information.
> > > >
> > > > host: em64t
> > > > kernel: 2.6.9-42.0.2.ELsmp - topspin
> > > > topspin: topspin-ib-mod-rhel4-2.6.9-42.0.2.ELsmp-3.2.0-110
> > > > firmware/adapter: v3.4.000 build 3.2.0.110, with label
> > > 'HCA.CougarCub.A1'
> > > >
> > > > making mpir in directory mpid/vapi
> > > >
> > > > for file in queue.c sbcnst2.c tr2.c sbcnst2.h tr2.h ; do \
> > > > if [ ! -s $file ] ; then \
> > > > ln -s ../util/$file; \
> > > > fi; \
> > > > done
> > > > gcc -DHAVE_CONFIG_H -I. -I/root/admin/mvapich-0.9.8/mpid/vapi
> > > > -I/root/admin/mvapich-0.9.8/include
> -I/root/admin/mvapich-0.9.8/include
> > > > -I/root/admin/mvapich-0.9.8/mpid/vapi
> > > -I/root/admin/mvapich-0.9.8/mpid/util
> > > > -DMPID_DEVICE_CODE -DHAVE_UNAME=1 -DHAVE_NETDB_H=1
> > > -DHAVE_GETHOSTBYNAME=1
> > > > -D_EM64T_ -DUSE_INLINE -DEARLY_SEND_COMPLETION -DRDMA_FAST_PATH
> > > > -DVIADEV_RPUT_SUPPORT -DLAZY_MEM_UNREGISTER -D_SMP_
> > > -fPIC
> > > > -fno-strict-aliasing -D_SMP_RNDV_ -D_PCI_X_ -D_SDR_
> > > > -I/usr/local/topspin/include
> -I/usr/local/topspin/include/vapi
> > > > -DHAVE_MPICHCONF_H -I/root/admin/mvapich-0.9.8
> > > > -I/root/admin/mvapich-0.9.8/mpid/vapi -I. -Wall -c viainit.c
> > > > viainit.c: In function `async_event_handler':
> > > > viainit.c:170: error: `VAPI_CLIENT_REREGISTER' undeclared (first
> use in
> > > this
> > > > function)
> > > > viainit.c:170: error: (Each undeclared identifier is reported only
> once
> > > > viainit.c:170: error: for each function it appears in.)
> > > > viainit.c:172: error: `VAPI_ECC_DETECT' undeclared (first use in
> this
> > > > function)
> > > > viainit.c:173: error: `VAPI_PATH_MIG_ARMED' undeclared (first use
> in
> > > this
> > > > function)
> > >
> > > We have tested MVAPICH extensively with VAPI libraries provided by
> > > Mellanox (IBGD, version 1.8.2). During our testing we did not see
> these
> > > errors. Can you try out with IBGD and let us know the outcome of your
> > > experimentation?
> > >
> > > Also, as you are aware, the IB community is moving towards
> Openfabrics
> > > based drivers. I would strongly suggest you to upgrade your stack
> with
> > > openfabrics gen2 drivers. Please let us know the outcome.
> > >
> > > Thanks,
> > >
> > > :- Abhinav
> > >
> > > > make[3]: *** [viainit.o] Error 1
> > > > Exit status from make was 2
> > > > make[2]: *** [mpilib] Error 1
> > > > make[1]: *** [mpi-modules] Error 2
> > > > make: *** [mpi] Error 2
> > > >
> > > >
> > > >
> > > > #!/bin/bash
> > > >
> > > > source ./make.mvapich.def
> > > > arch
> > > >
> > > > # Mandatory variables. All are checked except CXX and F90.
> > > > MTHOME=/usr/local/topspin
> > > > PREFIX=/share/apps/mvapich/gnu
> > > > export CC=gcc
> > > > export CXX=g++
> > > > export F77=g77
> > > > export RSHCOMMAND=ssh
> > > > IO_BUS="_PCI_X_"
> > > > ARCH="_EM64T_"
> > > > LINKS="_SDR_"
> > > >
> > > > export LIBS="-L${MTHOME}/lib64 -lmosal -lvapi -lpthread"
> > > > export FFLAGS="-L${MTHOME}/lib64 -fPIC"
> > > > export CFLAGS="-D${ARCH} -DUSE_INLINE -DEARLY_SEND_COMPLETION
> > > > -DRDMA_FAST_PATH \
> > > > -DVIADEV_RPUT_SUPPORT -DLAZY_MEM_UNREGISTER -D_SMP_
> \
> > > > -fPIC -fno-strict-aliasing -D_SMP_RNDV_ \
> > > > $SUPPRESS -D${IO_BUS} -D${LINKS} \
> > > > -I${MTHOME}/include -I${MTHOME}/include/vapi"
> > > >
> > > > # Prelogue
> > > > make distclean &>/dev/null
> > > >
> > > > # Configure MVAPICH
> > > >
> > > > echo "Configuring MVAPICH..."
> > > >
> > > > ./configure --with-device=vapi --with-arch=LINUX -prefix=${PREFIX}
> \
> > > > --enable-cxx --enable-debug \
> > > > --enable-devdebug \
> > > > --enable-f77 \
> > > > --enable-romio=with-aio --without-mpe \
> > > > -lib="-L${MTHOME}/lib64 -lmosal -lvapi -lpthread"
> > > > _______________________________________________
> > > > mvapich-discuss mailing list
> > > > mvapich-discuss at cse.ohio-state.edu
> > > > http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
> > >
> >
> >
> > -++**==--++**==--++**==--++**==--++**==--++**==--++**==--++**==
> > Be sure to attend the Rocks Clusters BoF, Supercomputing 2006
> > http://sc06.supercomputing.org/schedule/event_detail.php?evid=5169
> > _______________________________________________
> > mvapich-discuss mailing list
> > mvapich-discuss at cse.ohio-state.edu
> > http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
> >
>
>
-++**==--++**==--++**==--++**==--++**==--++**==--++**==--++**==
Be sure to attend the Rocks Clusters BoF, Supercomputing 2006
http://sc06.supercomputing.org/schedule/event_detail.php?evid=5169
More information about the mvapich-discuss
mailing list