assemble function value in MPI

Question

Hello,

I would lito to solve indeipendently same mesh but different condition on diferent cores, and sum up a funtion value at each core. But it does not work as expected.
In the case of the following demo program 'lift-darg',

#include <dolfin.h>
#include <mpi.h>
#include "Functionals.h"

using namespace dolfin;

// Define sub domain for the dolphin
class Fish : public SubDomain
{
  bool inside(const Array<double>& x, bool on_boundary) const
  {
    return (x[0] > DOLFIN_EPS && x[0] < (1.0 - DOLFIN_EPS) &&
            x[1] > DOLFIN_EPS && x[1] < (1.0 - DOLFIN_EPS) &&
            on_boundary);
  }
};

int main()
{
  // Read mesh from file
  Mesh mesh(MPI_COMM_SELF);
  HDF5File fpH5m(MPI_COMM_SELF,"./dolfin_fine.h5","r");
  fpH5m.read(mesh,"mesh",false);

  // Read velocity field from file
  Functionals::CoefficientSpace_p Vp(mesh);
  Function p(Vp);
  HDF5File fpH5p(MPI_COMM_SELF,"./dolfin_fine_pressure.h5","r");
  fpH5p.read(p,"p");

  // Mark 'fish'
  FacetFunction<std::size_t> markers(mesh, 1);
  Fish fish;
  fish.mark(markers, 1);

  // Functionals for lift and drag
  Functionals::Form_lift L(mesh, p);
  Functionals::Form_drag D(mesh, p);

  // Attach markers to functionals
  L.ds = markers;
  D.ds = markers;

  // Assemble functionals over sub domain
  const double lift = assemble(L);
  const double drag = assemble(D);

  info("Lift: %f", lift);
  info("Drag: %f", drag);
}

Executing this program by "mpirun -n 2 ./myprogram", it sums up p at all process. How can I get sum at 1 process?