Release notes for tiegcm1.94.2
January 5, 2012

Summary of changes from v1.94.1 (r674, 7/28/2011) to v1.94.2 (r739, 1/5/2012):

 - Introducing Intel 2011 Cluster Studio ifort compiler and mpi
   library. See scripts/Make.intel_hao64 to use the hao installation.
   The file scripts/Make.intel11.1_hao64 is retained in case anyone needs
   to use the old compiler. On 64-bit Linux systems at hao, the Intel
   compilers represent a significant performance improvement over PGI.

 - Fixed memory accumulation bug with ifort compiler by changing use of
   f90 intrinsic "trim" function in strloc (util.F), addfld.F, and elsewhere.
   The Intel ifort compiler was doing mallocs when the trim function
   was being used to compare strings (e.g., if (trim(string1)==trim(string2))
   Used memory checker in valgrind to find this problem.

 - Changes to dynamic memory allocation in mpi routines (mpi.F), to
   allocate send and receive buffers once per run per subroutine,
   instead of alloc/dealloc at every call. These modifications are
   for performance purposes only, and do not affect model results.

 - Provide no-dynamo option. To run the model without electro-dynamo
   set namelist read parameter DYNAMO = 0. In this case, advance calls
   sub nodynamo instead of sub dynamo. Sub nodynamo zeros out the 3 components 
   of the electric field (magnetic grid), and the 3d electric potential
   in both geographic and magnetic. POTEN, UI,VI,WI will be all zero
   on the histories.

 - Allow seamless switching between model resolutions, using $modelres
   in the job script. This is accomplished by comparing defs.h in the
   execdir with defs2.5 and defs5.0 in the srcdir, and copying the
   appropriate one from srcdir to execdir when necessary. Note change
   of names from defs.2.5 and defs.5.0 to defs2.5 and defs5.0.

 - Add sub rdfilename (nchist.F) to read data file names from source
   history. This allows changes in parameter mxlen_filename between
   releases, and fixes a bug that was occaisionally expressed when
   making continuation runs from an "old" history.
  
 - Corrected sign error in zigmc and zigm2 in southern hemisphere
   when difference coefficients are set up. This error was occurring
   only when the current calculation was turned on (icalkqlam==1).
   This fix affects Kqphi by approximately 1%.

 - Rename sub getpid to getprocessid (util.F) to avoid conflict with
   PVM lib when the model is coupled into CMIT.

 - Fixed problem with i/o timing calls in advance.F, and added copying
   of source history for initial runs to primary history timing.

 - Added SECSTART and SECSTOP to the Userguide namelist parameter 
   descriptions (doc/userguide/namelist.rst).

 - Move parameter joulefac ("joule heating factor") from dt.F to sub qjoule_tn 
   in qjoule.F, where it is applied to qji_tn during the joule heating calculation.
   This is before it is optionally saved as a secondary history diagnostic (QJOULE 
   and/or QJOULE_INTEG). Sub qjoule_tn is called before dt, so joulefac is no longer
   applied in dt. Also added joulefac to history files as a scalar per history. 
   Changed qjoule.F to a module, w/ joulefac in module data, so it can be use-associated
   by output.F (the routines are use-associated by dynamics.F for the calls).
   This will NOT affect model results, EXCEPT for the joule heating diagnostics,
   which will be 1.5x larger than in previous revisions (joulefac==1.5).

 - Move sub calccloc from wei01cm.F to util.F, modified to use default values for
   convection radius and center location of the convection pattern. The defaults
   are normally used for Bz northward conditions. Sub calccloc has been tested
   with Weimer01, but is NOT called at all in this revision. A new subroutine 
   wei01loc (based on wei05loc) is called for Weimer01. Both wei01loc and wei05loc 
   were modified to use byloc, a subset of byimf, which is used for the cusp location
   with phid. Sub calccloc will be used later for imposed potential patters
   for CMIT/LFM or AMIE. For more details see svn commit log r738 by Emery (1/5/2012).