TM5-MP

This issue was already reported when testing the updated polar cap photolysis code (r4981-5). I post it here as a reminder, since it has been a known issue for a long time and I tend to forget this kind of specifics if not written down. I was able to reproduce it with the trunk at both r899 and r910, when running the code compiled with check-all (i.e. -check bounds,uninit -ftrapuv -fpe0), In each cases, it crashes at the same time step.

>>> dynamics step from : 2006/01/02 22:40
>>> dynamics step from : 2006/01/02 23:00
forrtl: error (72): floating overflow
Image              PC                Routine            Line        Source             
libintlc.so.5      00002AAAAC315CF1  tbk_trace_stack_i     Unknown  Unknown
libintlc.so.5      00002AAAAC313E2B  tbk_string_stack_     Unknown  Unknown
tm5mp.x            00000000015C49A4  Unknown               Unknown  Unknown
tm5mp.x            00000000015C47B6  Unknown               Unknown  Unknown
tm5mp.x            000000000156C3B9  Unknown               Unknown  Unknown
tm5mp.x            0000000001570DDC  Unknown               Unknown  Unknown
libpthread-2.11.3  00002AAAAD8AC850  Unknown               Unknown  Unknown
tm5mp.x            0000000000CD2F78  mdf_mp_mdf_put_va       22829  mdf.F90
tm5mp.x            0000000001461D62  user_output_pdump        4107  user_output_pdump.F90
tm5mp.x            000000000149020E  user_output_pdump        1122  user_output_pdump.F90
tm5mp.x            0000000001301F45  user_output_mp_us         968  user_output.F90
tm5mp.x            0000000000E6AB0B  modelintegration_         821  modelIntegration.F90
tm5mp.x            0000000000E68F95  modelintegration_         540  modelIntegration.F90
tm5mp.x            0000000001015F82  tm5_mp_tm5_model_        1204  tm5.F90
tm5mp.x            00000000010C6146  MAIN__                     63  tracer.F90
tm5mp.x            000000000041021E  Unknown               Unknown  Unknown
libc-2.11.3.so     00002AAAADD52C36  __libc_start_main     Unknown  Unknown
tm5mp.x            0000000000410129  Unknown               Unknown  Unknown

If the user_output_pdump.F90 is not compiled with Intel -fpe0, the temperature written to the file is correct. This was already the case before, but was not documented. Until further notice, the recommendation is: do not use pdump when running a check-all tests. One day run may be ok. Keep in mind this is for the ECMWF CRAY machine. Behavior on other platforms is not known.