ORCA2_LIM3 does not compile on Met Office IBM

[timgraham]

There are some problems in the LIM3 code meaning it won't compile on our system at r4879 (i.e. the revision being used for the merge).

The first bug is in NEMOGCM/NEMO/LIM_SRC_3/limistate.F90. At line 302 the extra commas in the write statements cause a syntax error:

      IF(lwp) THEN
	         WRITE(numout,*), ' ztests : ', ztests
	         IF ( ztests .NE. 4 ) THEN
	            WRITE(numout,*)
	            WRITE(numout,*), ' !!!! ALERT                  !!! '
	            WRITE(numout,*), ' !!!! Something is wrong in the LIM3 initialization procedure '
	            WRITE(numout,*)
	            WRITE(numout,*), ' *** ztests is not equal to 4 '
	            WRITE(numout,*), ' *** ztest_i (i=1,4) = ', ztest_1, ztest_2, ztest_3, ztest_4
	            WRITE(numout,*), ' zat_i_ini : ', zat_i_ini(i_hemis)
	            WRITE(numout,*), ' zht_i_ini : ', zht_i_ini(i_hemis)
	         ENDIF ! ztests .NE. 4
	ENDIF

After fixing this error I've come across the following error:
ORCA2LIM3_LONG/BLD/ppsrc/nemo/limdiahsb.f90", line 89.40: 1513-041 (S) Arguments of the wrong type were specified for the INTRINSIC procedure "sign".

I'm investigating this now.

[timgraham]

The problem arises because the arguments to SIGN have different types. At line 79,

     zinda = MAX( 0.d0 , SIGN( 1.d0 , glob_sum( area(:,:) * tms(:,:) ) - epsi06 ) )

the first argument is double precision but the second is not. The same is true for the MAX function which also fails.

I can get this to compile by changing 1.d0 and 0.d0 to 1.0 and 0.0 but is this an acceptable fix? Any suggestions welcome.

[clem]

It seems that SIGN is not treated the same way depending on the machine (ok on IBM ifort compiler). Your fix seems ok to me, but Gurvan will confirm (?).

[timgraham]

I've made a slight improvement to the fix above by changing the line to

zinda = MAX( 0._wp , SIGN( 1._wp , glob_sum( area(:,:) * tms(:,:) ) - epsi06 ) )

in r4908.

We will need to pick up these changes in our branches next week if we want to test ORCA2_LIM3 on the Met Office IBM.