Submit MPI data migration asynchronously, so it's all overlapped

doc/doxygen/chapters/08mpi_support.doxy

@@ -339,7 +339,7 @@ migrate the data, and register the new location.
                                               (uintptr_t)NULL, sizeof(unsigned));
             if (data_handles[x][y]) {
                 /* Migrate the data */
-                starpu_mpi_get_data_on_node(MPI_COMM_WORLD, data_handles[x][y], mpi_rank);
+                starpu_mpi_get_data_on_node_detached(MPI_COMM_WORLD, data_handles[x][y], mpi_rank, NULL, NULL);
                 /* And register the new rank of the matrix */
                 starpu_data_set_rank(data_handles[x][y], mpi_rank);
             }
@@ -367,7 +367,7 @@ latest value on the original home node.
             if (data_handles[x][y]) {
                 int mpi_rank = my_distrib(x, y, size);
                 /* Get back data to original place where the user-provided buffer is.  */
-                starpu_mpi_get_data_on_node(MPI_COMM_WORLD, data_handles[x][y], mpi_rank);
+                starpu_mpi_get_data_on_node_detached(MPI_COMM_WORLD, data_handles[x][y], mpi_rank, NULL, NULL);
                 /* And unregister it */
                 starpu_data_unregister(data_handles[x][y]);
             }

mpi/examples/stencil/stencil5.c

@@ -175,7 +175,7 @@ int main(int argc, char **argv)
 			if (data_handles[x][y] && mpi_rank != starpu_data_get_rank(data_handles[x][y]))
 			{
 				/* Migrate the data */
-				starpu_mpi_get_data_on_node(MPI_COMM_WORLD, data_handles[x][y], mpi_rank);
+				starpu_mpi_get_data_on_node_detached(MPI_COMM_WORLD, data_handles[x][y], mpi_rank, NULL, NULL);
 				/* And register new rank of the matrix */
 				starpu_data_set_rank(data_handles[x][y], mpi_rank);
 			}
@@ -208,7 +208,7 @@ int main(int argc, char **argv)
 			{
 				int mpi_rank = my_distrib(x, y, size);
 				/* Get back data to original place where the user-provided buffer is. */
-				starpu_mpi_get_data_on_node(MPI_COMM_WORLD, data_handles[x][y], mpi_rank);
+				starpu_mpi_get_data_on_node_detached(MPI_COMM_WORLD, data_handles[x][y], mpi_rank, NULL, NULL);
 				/* Register original rank of the matrix (although useless) */
 				starpu_data_set_rank(data_handles[x][y], mpi_rank);
 				/* And unregister it */