WORK IN PROGRESS: Updates for Standards Conformance (NAG)

[mathomp4]

This PR tracks changes needed to compile GEOSgcm_GridComp with the NAG Compiler. NAG is very strict and will not allow the use of Fortran extensions that are endemic to GEOS.

---

The first category are procedures that are now part of the Fortran Standard that used to be extensions in some compilers:

• iargc() → command_argument_count()
• getarg() → get_command_argument()
• getenv() → get_environment_variable()
• system() → execute_command_line()
• call exit(0) → stop
• call exit(N) → error stop N

NAG also doesn't support (without flags that need to be used everywhere as @tclune can elucidate on) the use of real*4 or real(kind=4). The reason is that NAG does not use 4 for 32-bit real and 8 for 64-bit, instead they use 1 and 2. (The Standard does not specify what "kind" a 32-bit real must be, only that there must be one.) So, many changes are of the type:

• real*4 → real(REAL32)
• real(kind=8) → real(kind=REAL64)

Finally, in this, some catch code uses nan and inf defined like:

  real*8, parameter :: inf8 = O'0777600000000000000000'
  real*8, parameter :: nan8 = O'0777610000000000000000'
  real*4, parameter :: inf4 = O'17740000000'
  real*4, parameter :: nan4 = O'17760000000'
  real,   parameter :: inf = inf4
  real,   parameter :: nan = nan4

NAG does not allow this as it sees these as, well, not numbers at compile-time. So with @tclune we made functions to return these values via ieee_arithmetic a la:

   function nan() result(nan_32)
      real(REAL32) :: nan_32

      nan_32 = ieee_value(nan_32,  ieee_quiet_nan)

   end function nan

   function inf() result(inf_32)
      real(REAL32) :: inf_32

      inf_32 = ieee_value(inf_32,  ieee_positive_inf)

   end function inf

[sdrabenh]

@mathomp4 mind if I close this for now? It looks very old

[mathomp4]

@mathomp4 mind if I close this for now? It looks very old

@sdrabenh Actually, keep it open for now and let me see if I can merge in develop. If it doesn't work, I'll do it. This branch was too much work for me to let go of now!

[mathomp4]

Okay. The merge worked, which is good. For now I'll close the PR and can always bring it back if needed.