FIX: F1001 "@(#)fltreg.c:1.31" When Compiling with /Ox

The information in this article applies to:

• Microsoft FORTRAN PowerStation for MS-DOS, version 1.0

SYMPTOMS

A complex series of expressions may generate the Internal Compiler Error F1001 in file "@(#)fltreg.c:1.31" when compiled with full optimization (/Ox).

RESOLUTION

Adding the compiler switch /Op for floating-point consistency may avoid this error message. Removing the compiler switch /Ox may also resolve this problem. Breaking down the complex expressions using temporary variables is also a recommended solution.

STATUS

Microsoft has confirmed this to be a problem in FORTRAN PowerStation for MS-DOS, version 1.0. This problem has been resolved with FORTRAN powerstation maintenance release version 1.0a for MS-DOS.

FORTRAN PowerStation version 1.0 can be differentiated from maintenance release version 1.0a by invoking the linker. Typing "Link32 | more" from \F32\BIN directory will show version 2.8 for FORTRAN Powerstation version 1.0, and it will show version 1.0f for the maintenance release version 1.0a.

MORE INFORMATION

Some code compiled with the /Ox switch may generate the following error message:

   fatal error F1001: INTERNAL COMPILER ERROR
   compiler file '@(#)fltreg.c:1.31', line 4792

To duplicate this error message, compile the following code with only the /Ox option. This sample produces the line 4792 message.

To resolve the problem, add the /Op option or remove the /Ox option. Also, almost any simplification or restructuring of the equations will remove the error. For example, uncomment the "temp =" line.

This code compiles without any error with Microsoft FORTRAN 5.1 or earlier.

Sample Code

C Compiler options required: /Ox and not /Op

      DIMENSION BN(30)
      COMMON ALPA, DN, THT, NP1, K, BN

      SR=ALPA**(DN*THT)+BN(NP1)*0.5D0*ALPA**((DN+1.0D0)*THT)
      UTH=DSIN(THT)-BN(NP1)*0.5D0/ALPA**((DN+1.0D0)*THT)
      URB=ALPA**DCOS(DN*THT)*(DN-1.0D0)
      SRK2=(1.0D0)**ALPA**(K-1)
      UTHK2=(-1.0D0)**ALPA**K
      END