[MITgcm-devel] authors testing

[Ed Hill]

Hi folks,

I installed daily testing scripts on "eddy" (which is running Fedora
Core 5) and the results can be seen at the usual location:

  http://mitgcm.org/testing.html

I intend to switch over our main testing machine from faulks to eddy in
a week or so since FC-3 (what faulks is now running) reached end-of-
life months ago.  In the mean time, here are some notes and please feel
free to compare the new compilers with the older ones.

And are there any objections to this switch-over?

Ed

compiler notes:

  g77:
    - The output is _exactly_ the same (according to "diff -u") on 
      eddy (FC-5) as it is on faulks (FC-3).  This is to be expected 
      since its an identical compiler, just a different underlying 
      OS.

  gfortran:
    - The namelist handling still produces problems however this 
      bug was recently fixed in GCC CVS so it will get corrected
      in forthcoming releases:
         http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24459

    - There is a problem with the adjoint-only code where, for 
      example, in grdchk_main.f there are statements such as:

        print ('(2a)'), 'a string', 'another string'

      and I think we can easily fix this syntax.  The good 
      news here is that gfortran does compile the STAF output 
      ("ad_taf_output.f")

  ifort:
    - The new Intel v9.x compiler works nicely on FC-5
    - The older Intel v8.x compiler (which is currently the default 
      on authors) does not work with FC-5, at least in its current
      configuration.  Given that v8.x and v9.x are pretty similar,
      I don't want to waste any time trying to get v8.x working 
      with FC-5.

   pgf77:
    - The old PGI v5.2 compiler does not work on FC-5 and is not 
      supported on that OS.
    - The latest PGI v6.1 compiler works fine on FC-5 and gives 
      essentially identical results -- the few failed tests are 
      due to intermittent problems contacting the PGI license 
      manager -- *very* annoying!  I'll try to schedule future 
      tests at a different time so that they do not compete with 
      the tests on the clusters (cg01 and aces) for the limited 
      number of PGI floating licenses.


-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464