
===============================================================================
			       Changes in 1.1.0
===============================================================================

- Added two new functions 'HMPI_Group_pauto_create' and 
'HMPI_Group_heuristic_pauto_create', whose interfaces are shown below:

   int
   HMPI_Group_pauto_create
   (
       HMPI_Group* _gid,
       const HMPI_Model* model,
       const int* model_parameters,
       int model_param_count
   )


   int
   HMPI_Group_heuristic_pauto_create
   (
       HMPI_Group* _gid,
       const HMPI_Model* model,
       HMPI_Heuristic_function hfunc,
       const int* model_parameters,
       int model_param_count
   )

These functions are similar to 'HMPI_Group_auto_create' and 
'HMPI_Group_heuristic_auto_create' except that they evaluate the 
process arrangements in parallel. For the functions 'HMPI_Group_auto_create' 
and 'HMPI_Group_heuristic_auto_create', only the host process evaluates all
the process arrangements. 

If the evaluation of a process arrangement takes long time due to
accurate but expensive performance models, the 'pauto' functions must be used. 
During the call to these functions, all the computers/nodes evaluate the 
process arrangements in parallel. The number of process arrangements to evaluate 
is proportional to the speed of the computer/node. So the 'pauto' functions 
are more efficient.

- The following function must be used only in combination with the 'auto' and 
the 'pauto' functions.

   double
   HMPI_Group_timeof
   (
       const HMPI_Group* gid
   )

It returns the estimated time taken to execute the algorithm without real execution 
on the underlying hardware. This function must be called by all the members of the 
group represented by the handle 'gid'.
