Bring back the constructor of MPI_InitFinalize.

I had removed it in a previous patch, thinking that it was deprecated,
but in fact it wasn't. Rather than undoing the change, provide the
third argument of another constructor with a default value.

diff --git a/doc/news/changes.h b/doc/news/changes.h
index e28e3b94bd187215240753be81296851ec6a5d29..4ca43dc56a4d1a862939024a4dff60ff0c086731 100644 (file)
--- a/doc/news/changes.h
+++ b/doc/news/changes.h
@@ -38,6 +38,22 @@ inconvenience this causes.
 </p>
 
 <ol>
+  <li> Changed: The MPI_InitFinalize class had a constructor that took
+  two arguments (<code>argc,argv</code>). This constructor has been removed
+  in favor of the one that takes three arguments where the last one is
+  the number of threads to be used for the current MPI process. This last
+  argument has received a default value so that the old, two-argument
+  calling syntax continues to work. However, whereas the old syntax with
+  two arguments indicated that the user wanted only one thread per process,
+  the default value is set so that every processor core on the system is
+  used. If you run as many MPI processes as there are processor cores, then
+  this means one thread per MPI process, as before. On the other hand, if you
+  start fewer MPI processes than there are cores, then your program will now
+  be allowed to use more than one thread, speeding up a number of operations.
+  <br>
+  (Wolfgang Bangerth, 2015/01/14)
+  </li>
+
   <li> Removed: TrilinosWrappers::SparseMatrix copy constructor got removed
   to be in line with PETSc and dealii::SparseMatrix. You can use reinit()
   and copy_from().
@@ -116,7 +132,6 @@ inconvenience this causes.
     approach to getting notified about what happens to triangulations
     has been superseded by the signals defined by the triangulation
     class.
-  - The deprecated constructor of MPI_InitFinalize
   - The MPI support functions in namespace Utilities and Utilities::System.
   - Deprecated members of namespace types.
   - Namespace deal_II_numbers.

diff --git a/include/deal.II/base/mpi.h b/include/deal.II/base/mpi.h
index 25934a1dcbf979ba2668cee102d2e2c5e67c9774..726229d0142a2b0eb61d0821575a87fff1e1f85b 100644 (file)
--- a/include/deal.II/base/mpi.h
+++ b/include/deal.II/base/mpi.h
@@ -279,7 +279,8 @@ namespace Utilities
        * main.
        * @param[in] max_num_threads The maximal number of threads this MPI
        * process should utilize. If this argument is set to
-       * numbers::invalid_unsigned_int, the number of threads is determined
+       * numbers::invalid_unsigned_int (the default value),
+       * then the number of threads is determined
        * automatically in the following way: the number of threads to run on
        * this MPI process is set in such a way that all of the cores in your
        * node are spoken for. In other words, if you have started one MPI
@@ -307,7 +308,7 @@ namespace Utilities
        */
       MPI_InitFinalize (int    &argc,
                         char ** &argv,
-                        const unsigned int max_num_threads);
+                        const unsigned int max_num_threads = numbers::invalid_unsigned_int);
 
       /**
        * Destructor. Calls <tt>MPI_Finalize()</tt> in case this class owns the