Intel Fortran 8.0, gcc 3.2 (Redhat 8.0)

Message

cvl · #1 Unread post by **cvl** » Sat May 15, 2004 9:01 pm

For unknown reasons, the ROMS 2.1 release Makefile.*ifort failed to work on my systems with the Intel Fortran compiler, Vers. 8.

1) PIIIM laptop, Redhat 8.0
2) Dual Athlon, Redhat 8.0
3) 16-node beowulf of dual PIII's, Redhat, version unknown.

Eliminating the libocean.a archive and returning to the old long list of *.o files worked just fine, however. I'd be mildly curious to know the logic behind the use of ar and why it failed on my system. The obscure error referred to MAIN__ and for_main.o.

Now, back to dueling it out with mpich version 1.2.5.2 on the Beowulf cluster. Anyone with experience in configuring these, I'd love to know more.

arango · #2 Unread post by **arango** » Mon May 17, 2004 7:38 pm

I am runing ifc version 7.x and ifort version 8.0 on my desktop and laptop under Mandrake Linux. I haven't seen such problems. The new structure of ROMS/TOMS requires a library so we can couple to other systems. Notice that the new driver of ROMS/TOMS follows the Earth System Modeling Framework (ESMF) conventions. Libraries are also required for the tangent linear and adjoint models. I think that the problem that you are having is associated with your Linux environment. The Unix ar tool is used to create, modify, and extract object files from libraries. Various Unix environments require different commands and/or flags. Perhaps, the flags in the Makefile that you are using are not appropriate for Redhat Linux. Try man ar and check if this is the case with Redhat Linux.

Rob Hetland told me last week that he was able to run version 2.1 on his PC cluster. He is now using LAM/MPI for his distributed-memory applications. This new MPI library is more efficient than MPICH. I highly recommend to install this efficient library. See http://www.lam-mpi.org for more details.

Good luck, H

klzhuang · #3 Unread post by **klzhuang** » Fri May 28, 2004 2:06 am

I have encountered the same problem of main_. Have you solved that? Would you please tell me your suggestion?

Thanks

arango · #4 Unread post by **arango** » Fri May 28, 2004 3:33 am

Please review the following post:

viewtopic.php?t=70

and correct minor bugs with ROMS/TOMS version 2.1. The problem is likely associated with nonstandard usage of CPP. Seems that the Redhat Linux cpp program is very strict about its constructs.

Good luck

dbergondo · #5 Unread post by **dbergondo** » Fri Jun 11, 2004 1:32 pm

I am using ifort with Red Hat Linux 7.1 and I am having the same problem with MAIN and for_main.o. Below is the information from Intel about the problem, but I still haven't been able to resolve the problem. Any suggestions?

Intel® Fortran Compiler for Linux*
Compiler Reports "undefined reference to 'main'"
This is a known issue and occurs when the compiler is invoked without passing it any source files. It is a harmless message and occurs due to the presence of the - Xlinker option in the compiler .cfg file located in the bin directory. This is done to add the path of the Intel® Compiler shared libraries, so it is not necessary to modify LD_LIBRARY_PATH to run executables produced by the Intel Compilers.