Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
28499 Discussions

Internal compiler error: internal abort

Guowei_H_
Beginner
1,565 Views

Dear Sir / Madam,

Intel compiler gave following error:

catastrophic error: **Internal compiler error: internal abort** Please report this error along with the circumstances in which it occurred in a Software Problem Report.  Note: File and line given may not be explicit cause of this error.
compilation aborted for /homec/jicg41/jicg4146/cesm1_2_2/models/rof/rtm/src/riverroute/RtmMod.F90 (code 1)

 

Compiler:

ifort (IFORT) 15.0.0 20140723

Command:

mpif90  -c -I.  -I/work/jicg41/jicg4146/CESM/test5/exe/intel14/mpich/nodebug/nothreads/include -I/work/jicg41/jicg4146/CESM/test5/exe/intel14/mpich/nodebug/nothreads/MCT/noesmf/a1l1r1i1o1g1w1/csm_share -I/usr/local/software/juropatest/Stage1/software/MPI/intel/2015.0.090/psmpi/5.1.0-1/netCDF-Fortran/4.2/include -I/usr/local/software/juropatest/Stage1/software/Compiler/intel/2015.0.090/psmpi/5.1.0-1/include -I/work/jicg41/jicg4146/CESM/test5/exe/intel14/mpich/nodebug/nothreads/include -I/homec/jicg41/jicg4146/cesm1_2_2/models/csm_share/shr -I. -I/homec/jicg41/jicg4146/cesm1_2_2/scripts/test5/SourceMods/src.rtm -I/homec/jicg41/jicg4146/cesm1_2_2/models/rof/rtm/src/riverroute -I/homec/jicg41/jicg4146/cesm1_2_2/models/rof/rtm/src/cpl_mct -I/homec/jicg41/jicg4146/cesm1_2_2/models/rof/rtm/src/cpl_share -I/work/jicg41/jicg4146/CESM/test5/exe/lib/include  -fp-model source -convert big_endian -assume byterecl -ftz -traceback -assume realloc_lhs  -O2   -DLINUX  -DNDEBUG -DMCT_INTERFACE -DHAVE_MPI -DFORTRANUNDERSCORE -DNO_R16 -DCPRINTEL  -DLINUX -DCPRINTEL14  -DHAVE_SLASHPROC -free   /homec/jicg41/jicg4146/cesm1_2_2/models/rof/rtm/src/riverroute/RtmMod.F90

Attached the file.

Thanks!

 

0 Kudos
1 Solution
Martyn_C_Intel
Employee
1,565 Views

A few suggestions to make progress:

In which subroutine does the compiler error occur? (Try deleting whole routines one at a time, starting from the end of the module, and see if the error goes away. You'll also need to delete any corresponding references to these routines at the start of the module).

Does the error go away if you replace -O2 by -O0 on the command line for RtmMod.F90 ?

If so, try deleting whole blocks or entire loops from the routine in which the error occurs, starting from the end, and see if the error persists. Check that you can still recompile successfully at -O0, to ensure that you haven't introduced any syntax errors.  Even if you can't create a standalone reproducer on your own, this might localize the problem and reduce the external dependencies sufficiently that we could construct one.

View solution in original post

0 Kudos
5 Replies
TimP
Honored Contributor III
1,565 Views

It may be better to file a full report on premier.intel.com, as requested, as it appears that source code for the .mod files will be needed (and more?).  Could you try the 15.0.1 update?

0 Kudos
Guowei_H_
Beginner
1,565 Views

Dear Tim Prince,

I'd love to try the update. But it is on a supercomputer, and I've no permission to do that. Do I need any extra resource to access premier.intel.com? It said "You do not currently have access to the requested site."

0 Kudos
Steven_L_Intel1
Employee
1,565 Views

You would have to find who is assigned the role of license administrator for your installation and have them submit this for you. If you are able to provide all the sources in a gzipped tar file here, we could look at it, but I would ask that you try to cut down the source to the minimum needed to show the problem.

0 Kudos
Martyn_C_Intel
Employee
1,566 Views

A few suggestions to make progress:

In which subroutine does the compiler error occur? (Try deleting whole routines one at a time, starting from the end of the module, and see if the error goes away. You'll also need to delete any corresponding references to these routines at the start of the module).

Does the error go away if you replace -O2 by -O0 on the command line for RtmMod.F90 ?

If so, try deleting whole blocks or entire loops from the routine in which the error occurs, starting from the end, and see if the error persists. Check that you can still recompile successfully at -O0, to ensure that you haven't introduced any syntax errors.  Even if you can't create a standalone reproducer on your own, this might localize the problem and reduce the external dependencies sufficiently that we could construct one.

0 Kudos
Martyn_C_Intel
Employee
1,565 Views

An Intel Premier Support issue was created that provided source code. This could be compiled successfully using the ifort driver outside of the mpif90 wrapper. My understanding is that compilation of RtmMod with mpif90 was successful using a rebuilt MPI library. I think, therefore, that the original problem is resolved.

If you see any further internal compiler errors, please provide the details.

0 Kudos
Reply