Replace "derf()" with standard "erf()" in SBM MP scheme

[kkeene44]

TYPE: bug fix

KEYWORDS: PGI, compile, mp_SBM_polar_radar, phys, erf, derf

SOURCE: Bart Brashers (Ramboll)

DESCRIPTION OF CHANGES: A fortran function (DErF) was being used in the module_mp_SBM_polar_radar file. This was causing a compiling fail for PGI (tested with versions
19.10.0 on user's machine and 19.3 on Cheyenne). According to this article:
https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/DErF-Intrinsic.html
the DErF function is archaic. Modifying this function to ErF corrects the problem and allows the
PGI compiler to build the WRF executable from source.

LIST OF MODIFIED FILES:
M phys/module_mp_SBM_polar_radar.F

TESTS CONDUCTED:

1. Verified that code does not compile in original state, but does with the modified code when
   using the PGI compiler.
2. Reggie should be OK.

RELEASE NOTE: A minor modification has been introduced into the SBM Polar Radar microphysics code to allow the WRF code to build with PGI compilers. The nonstandard derf() intrinsic has been replaced by the 2008 standard compliant erf() intrinsic.

[kkeene44]

I clearly did something wrong with my fork - 532 changed files. Will correct and submit a new PR.

[davegill]

Kelly,
Version dyslexia: v4.1.2 vs v4.2.1!

[kkeene44]

@davegill Good eye - thanks!!

[davegill]

@kkeene44 @weiwangncar @dudhia
Folks,
There are a few places where we have this "derf" used:

> find . -name \*.F -exec grep -iw derf {} \; -print
         t=derf((1.0d0-fract_volume_water)/fract_volume_water-0.2d0)
./phys/module_mp_SBM_polar_radar.F

!      real derf,derfc, erf_alt
./phys/module_mp_morr_two_moment_aero.F

      function derf(x)
      derf = y
      end function derf
#define ERF derf
      real(r8) derf,derfc, erf_alt
!++ag replace sg erf with hm derf pre 1.68
!               nact=nact+0.5*(1.-derf(x))*na(m)
!               nact=nact+0.5*(1.-derf(x))*na(m)  
!       write(*,*)'nact',nact,derf(x),na(m),m
./phys/module_cu_mskf.F

             dum1  = nanew1*0.5*(1.-derf(uu1)) ! activated number in kg-1 mode 1
             dum2  = nanew2*0.5*(1.-derf(uu2)) ! activated number in kg-1 mode 2
 real function DERF(X)
      DERF = Y
 end function DERF
./phys/module_mp_p3.F

   module procedure derf
!   subprograms: ERF, ERFC, and ERFCX (or DERF, DERFC, and DERFCX),
!                   Y=ERF(X)     (or   Y=DERF(X)),
FUNCTION DERF(X)
   real(rk) :: DERF
   CALL CALERF_r8(X, DERF, JINT)
END FUNCTION DERF
./phys/module_cam_error_function.F

   real(r8) derf
#define ERF derf
      real(r8) derf,derfc
      real(r8) function derf(x)
      derf = y
      end function derf
./phys/module_cam_mp_microp_aero.F

Clearly, every other place has been patched up with various local definitions of this gaussian.

The erf is an intrinsic from Fortran 2008 and on.
GNU ERF doc
There is no need to ever say derf (which is an extension). However, there is no need to put the letter d in front of a zillion functions in the physics directory: dsin, dcos, dexp, dlog, etc.

Due to precedent, I am OK with this being the only change made to the WRF source.

[dudhia]

Why has this only shown up now when it was also in some older schemes?

…

On Mon, Apr 27, 2020 at 4:35 PM Dave Gill ***@***.***> wrote: @kkeene44 <https://github.com/kkeene44> @weiwangncar <https://github.com/weiwangncar> @dudhia <https://github.com/dudhia> Folks, There are a few places where we have this "derf" used: > find . -name \*.F -exec grep -iw derf {} \; -print t=derf((1.0d0-fract_volume_water)/fract_volume_water-0.2d0) ./phys/module_mp_SBM_polar_radar.F ! real derf,derfc, erf_alt ./phys/module_mp_morr_two_moment_aero.F function derf(x) derf = y end function derf #define ERF derf real(r8) derf,derfc, erf_alt !++ag replace sg erf with hm derf pre 1.68 ! nact=nact+0.5*(1.-derf(x))*na(m) ! nact=nact+0.5*(1.-derf(x))*na(m) ! write(*,*)'nact',nact,derf(x),na(m),m ./phys/module_cu_mskf.F dum1 = nanew1*0.5*(1.-derf(uu1)) ! activated number in kg-1 mode 1 dum2 = nanew2*0.5*(1.-derf(uu2)) ! activated number in kg-1 mode 2 real function DERF(X) DERF = Y end function DERF ./phys/module_mp_p3.F module procedure derf ! subprograms: ERF, ERFC, and ERFCX (or DERF, DERFC, and DERFCX), ! Y=ERF(X) (or Y=DERF(X)), FUNCTION DERF(X) real(rk) :: DERF CALL CALERF_r8(X, DERF, JINT) END FUNCTION DERF ./phys/module_cam_error_function.F real(r8) derf #define ERF derf real(r8) derf,derfc real(r8) function derf(x) derf = y end function derf ./phys/module_cam_mp_microp_aero.F Clearly, every other place has been patched up with various local definitions of this gaussian. The 'erfis an intrinsic from Fortran 2008 and on. [GNU ERF doc]( https://gcc.gnu.org/onlinedocs/gfortran/ERF.html) There is no need to ever sayderf(which is an extension). However, there is no need to put the letterdin front of a zillion functions in the physics directory:dsin, dcos, dexp, dlog`, etc. Due to precedent, I am OK with this being the only change made to the WRF source. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1183 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEIZ77C3V7GFO4ZLP2HUE2TROYCBXANCNFSM4MSJK6KQ> .

[davegill]

@dudhia @kkeene44 @weiwangncar

Why has this only shown up now when it was also in some older schemes?

Jimy,
All of the other instances inside of WRF of the use of the "derf()" intrinsic function were essentially user-defined. The only raw "derf()" function is the single one that Kelly is modifying.

[dudhia]

thanks

…

On Tue, Apr 28, 2020 at 9:37 AM Dave Gill ***@***.***> wrote: @dudhia <https://github.com/dudhia> @kkeene44 <https://github.com/kkeene44> @weiwangncar <https://github.com/weiwangncar> Why has this only shown up now when it was also in some older schemes? Jimy, All of the other instances inside of WRF of the use of the "derf()" intrinsic function were essentially user-defined. The only raw "derf()" function is the single one that Kelly is modifying. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1183 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEIZ77EBRDO2FFTK63J7DXDRO3Z5HANCNFSM4MSJK6KQ> .

[davegill]

Approved

[davegill]

@nbald
Thanks for spotting the double PR.
I am closing #1246, but stealing the best parts of its commit message.