[MITgcm-support] G95 build issue: f2c name mangling

chris hill cnh at mit.edu
Wed Oct 11 21:56:13 EDT 2006


Mark,

  I made the T3E declaration change.

  You can either send build options files to this list as attachments or 
we can can give you check in access to CVS as long as
  (1) you're happy to be careful  i.e. do regression tests if you do 
futz with something, put up with being on our mitgcm-cvs and 
mitgcm-devel mailing list that shows checkins etc....
  (2) you won't get offended if/when you get flamed for doing the wrong 
thing - even if the right thing wasn't clearly documented anywhere!

Chris
Mark Hadfield wrote:
> Mark Hadfield wrote:
>> Oh what the hell. Below (following my signature) is my attempt at a 
>> rewrite of subroutine TIMER_GET_TIME in timers.F.
> The code I posted the other day included the following in the 
> declarations section:
> 
> # elif (defined (TARGET_T3E) || defined (TARGET_CRAY_VECTOR))
> C     No declarations necessary
> 
> I assumed this was OK because there were no declarations in the original 
> for the functions in question (second & secondr). I thought they might 
> be Cray-specific intrinsic functions.
> 
> The compiler on the NIWA T3E begs to differ and insists that 
> declarations *are* necessary, as these are external functions. This is 
> what the above code should look like (with some redundant parentheses 
> also removed).
> 
> # elif (defined TARGET_T3E || defined TARGET_CRAY_VECTOR)
>      real second, secondr
>      external second, secondr
> 
> I haven't tested this with TARGET_CRAY_VECTOR. From the "man second" 
> page I think "real" is the correct data type. (There is only one real 
> data type on T3Es.)
> 
> By the way, I have build-options files for a few new platforms to 
> submit. Where should I send them? Should I attach them to an email to 
> the list?
> 




More information about the MITgcm-support mailing list