[mvapich-discuss] MVAPICH2 and OFED 1.5.2 without libibcommon

[Jyh-Shyong]

Hi,

You are right,  the program I tried to compile using mvapich2/ofed-1.5 
includes a fftw3 library
which was compiled with the old version of mvapich2/ofed-1.4.  After 
recompiling the fftw3
with mvapich2/ofed-1.5, the program is compiled successfully.   Thank you.

Jyh-Shyong Ho

於 2011/2/25 下午 11:43, Jonathan Perkins 提到:
> Hello, I've looked into this issue a bit and I don't see where we
> would be depending on ibcommon.  I think this is an artifact from a
> previous installation.  When building the mvapich2 library there is a
> sample mpi program `cpi' that is built in the examples directory.
> This should have failed to build if there was a dependency on the
> non-existent library.
>
> Are the applications simply using mpicc in order to build?  Do you
> have multiple MPIs installed?  Can you check to make sure that you're
> using the correct mpicc and its linking against the new mpicc.  Try
> using `which mpicc' if you're not using the full path to mpicc
> explicitly.  Are you able to build the osu_benchmarks (available in
> the osu_benchmarks sub-directory and at
> http://mvapich.cse.ohio-state.edu/benchmarks/osu-micro-benchmarks-3.3.tar.gz)?
>
> On Fri, Feb 25, 2011 at 7:28 AM, Jyh-Shyong<c00jsh00 at nchc.org.tw>  wrote:
>> Hi,
>>
>> I installed OFED-1.5.2 and the mvapich2 included in the package on our
>> cluster,
>> and I found that when I try to compile some application programs using
>> mvapich2
>> MPI libraries, I always got an error message during the link stage:
>>
>> cannont find  -libcommon
>>
>> I realized that libibcommon has been removed from OFED-1.5.2, but I don't
>> inderstand
>> why MPI packages included in OFED-1.5.2 still have the dependency on
>> libibcommon?
>> Or is there a way to avoid this problem?
>>
>> I als tried to build mvapich2 with the source downloaded directly from its
>> official page
>> and got the same error,  the MPI　environment still need libibcommon.so or
>> libibcommon.a.
>>
>> Any suggestion is appreciated.
>>
>> Regards
>>
>> Jyh-Shyong Ho, Ph.D.
>> Research Scientist
>> National Center for HIgh Performance Computing
>> Hsinchu, Taiwan, ROC
>>
>>
>> _______________________________________________
>> mvapich-discuss mailing list
>> mvapich-discuss at cse.ohio-state.edu
>> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>>
>>
>
>