[mpich-discuss] mpich2-1.2.1p1 and Intel 11 Compiler

chan at mcs.anl.gov chan at mcs.anl.gov
Tue Mar 16 11:21:39 CDT 2010


Can you build mpich2-1.2.1p1 without --with-device=ch3:shm
and see if it works with your app ?  If the segfaults still
occur, could you send us either 1) a simple program that reproduces
the problem or 2) a backtrace that shows where segfault occurs.

A.Chan

----- "Jose F Rodriguez" <jfrodrig1971 at gmail.com> wrote:

> Dear all,
> 
> I am experiencing problems using mpich2 complied with intel 11 and
> running a fortran program. In my previous installation, we used gcc
> 4.2 with intel 10 and mpich 1.07 in a quad core AMD with Ubuntu and
> everything goes fine. Even in our cluster with mvapich2-1.2p1, gcc
> 4.1.1 and intel 10.1 there are no problems. 
> 
> At this moment we are trying to install mpich2-1.2.1p1 with intel 11
> and gcc 4.4 (the latest version of all compilers) in a quad core AMD
> as well as in a quad core Intel I5. The result in both cases is the
> same, a segmentation fault occurs in a call to a mpi_sendrecv routine
> in fortran when attempting to send a double_precision array. We have
> try compiling the mpich with all possible alternatives for ch3 but
> nothing. Our configuration looks like this (This configuration is the
> same we have used for our working  mpich2)
> 
> CC=icc FC=ifort CXX=icpc CFLAGS="-fomit-frame-pointer -ip"
> FFLAGS="-fomit-frame-pointer -ip" ./configure --enable-fast
> --enable-f77 --with-device=ch3:shm 2>&1 | tee c.txt
> make 2>&1 | tee m.txt
> 
> We have also tried compiling mpich 1.07 with the new intel and gcc
> compilers and the result is still the same. In addition, if you
> compile mpich 1.07 with intel 10 and the gcc 4.4 compiler you get a
> configuration error in mpich "timer not found". I guess this last
> error is due to the fact that gcc 4.4 is newer than intel 10 causing
> some incompatibilities.
> 
> However, can somebody help in pointing where the problem with the last
> version of mpich2, intel compiler and gcc could be?. It is important
> to keep in mind that we are using them with a fortran program.
> 
> Thank you in advance for all you help
> 
> jose
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss


More information about the mpich-discuss mailing list