IILM Error

I recently encountered a new error for me. Now, I was able to compile the file this morning by just doing a clean and a re-make, but I was wondering, what exactly does this error mean when I get it? I’ve never seen it before and, amazingly, neither has Google.

To wit:

...AO-Baselibs-4_0_2/x86_64-unknown-linux-gnu/pgfortran_12.8-openmpi_1.6.4/Linux/include/esmf -fast -Kieee -Mcuda=nofma,ptxinf
o,4.2,cc20 -ta=nvidia:wait,nofma,4.2,cc20 -Minfo=accel,par,ccff       -Minfo -fpic -Mbackslash -Ktrap=fp  -tp=nehalem-64   
          GEOS_MoistGridComp.F90
IILM file version error
Expecting AST2ILM version 1/23
      got i6 version 0/0
gmake[5]: *** [GEOS_MoistGridComp.o] Error 2
gmake[5]: Leaving directory `/nobackupp6/mathomp4/Models/Ganymed-4_0_UNSTABLE-PGI12-GPU/GEOSagcm/src/GEOSgcs_GridComp/GEOSg
cm_GridComp/GEOSagcm_GridComp/GEOSphysics_GridComp/GEOSmoist_GridComp'

Is this a PGI error? Lustre?

Thanks,
Matt

Hi Matt,

This does look like a PGI error, but it’s very unusual. It could occur if the front and back end compilers (pgf901 and pgf902) we’re different versions. That could only occur if someone manually overwrote one of them. It’s also possible, but highly unlikely, that the intermediary ILM file somehow got corrupted between the time it was produced by pgf901 and read in by pgf902.

Is this error reproducible?

  • Mat

Mat,

No, it’s not reproducible; in fact, it was the only time I’ve ever seen it. My guess is that the disks or something on our cluster had a hiccup and sent the compiler in a tailspin. I was just curious to know, if I ever did see it again, what might cause such an error.

Thanks,
Matt