As part of my research for my Ph.D. I am on a team that is currently developing an adjoint of the EPA’s CMAQ air quality model. In the process of integrating all parts of the model into the full adjoint model, I ran into an error that was rather difficult to resolve.
Running the model would result in many occurances of the following error:
forrtl: error (63): output conversion error, unit -5, file Internal Formatted Write
Image PC Routine Line Source
ADJOINT_FWD 00000000009B34BD Unknown Unknown Unknown
ADJOINT_FWD 00000000009B1FC5 Unknown Unknown Unknown
ADJOINT_FWD 0000000000969210 Unknown Unknown Unknown
ADJOINT_FWD 000000000092AADF Unknown Unknown Unknown
ADJOINT_FWD 000000000092A312 Unknown Unknown Unknown
ADJOINT_FWD 000000000095305A Unknown Unknown Unknown
ADJOINT_FWD 00000000005D9F94 ckdesc3_ 138 ckdesc3.f
ADJOINT_FWD 00000000005A9FD1 open3_ 216 open3.F
ADJOINT_FWD 000000000047B395 chk_files_impl_mp 170 CHK_FILES_IMPL.F
ADJOINT_FWD 0000000000485060 chk_files1_mp_chk 347 CHK_FILES.F
ADJOINT_FWD 00000000005666CB vdiff_ 369 vdiffacm2.F
ADJOINT_FWD 0000000000496B7E sciproc_ 228 sciproc.F
ADJOINT_FWD 000000000048DDB5 MAIN__ 205 driver_fwd.F
ADJOINT_FWD 0000000000404A1C Unknown Unknown Unknown
libc.so.6 0000003FE9E1D994 Unknown Unknown Unknown
ADJOINT_FWD 0000000000404929 Unknown Unknown Unknown
>>> WARNING in subroutine CRTFIL3 <<<
Error creating netCDF variable for file ADJ_VDIFF_CHK
Illegal data type 0
*** ERROR ABORT in subroutine CHK_FILE_CREATE_
Could not open ADJ_VDIFF_CHK file
Date and time 13:00:00 July 22, 2001 (2001203:130000)
I spent a lot of time searching online, however I was unable to find a solution for my problem. After days of debugging, I finally found the source of the problem.
The “output conversion error” comes from a WRITE command that uses an incorrect format for the number to print. It is often caused by numbers being bigger than expected. In the case of CMAQ adjoint, the problem was within the ckdesc.f file for ioapi.
The problematic section of the code is shown below:
DO U = 2, NVARS3D
DO V = 1, U-1
IF( VNAME3D( U ) .EQ. VNAME3D( V ) ) THEN
WRITE( MESG, 94030 )
& 'Variable name VNAME3D(', U, ') = "' //
& VNAME3D( U )( 1:TRIMLEN( VNAME3D( U ) ) ) //
& '" duplicates VNAME3D(', V, ') = "' //
& VNAME3D( V )( 1:TRIMLEN( VNAME3D( V ) ) ) //
& '" in file "' // FNAME( 1:TRIMLEN( FNAME ) ) // '"'
CALL M3MSG2( MESG ) EFLAG = .TRUE.
END IF
END DO END DO
94030 FORMAT( 4 ( A, I2, :, 2X ) )
The reason for the error was that the variable U had a max value of 125 in our case, while the format descriptor only allowed for a 2 digit integer in the WRITE statement. Changing I2 to I3 fixed the problem.
This was actually a bug in the IOAPI release, and has subsequently been resolved.
I really like reading a post that can make people think.
Also, thanks for allowing me to comment!
I have just encountered this error from my fortran compiler and I dont know how to solve it. I will appreciate any help for a possible solution.
Segmentation fault
0.618u 2.397s 0:03.11 96.4% 0+0k 5136+32io 21pf+0w