Hello,
I'm trying to optimize the stitching_helper() function in SerialMesh, which
is currently a bottleneck in the work I'm doing (stitching together a large
number of small components to assemble a full mesh).
I've found out that over 85% of the running time of this function is
actually spent in UnstructuredMesh::find_neighbors(). This function is
called twice:
once through UnstructuredMesh::copy_nodes_and_elements() and a second time
through a call to MeshBase::prepare_for_use().
The first call doesn't seem to be necessary, since removing works perfectly
fine and doesn't alter the solution of a subsequent solve, while decreasing
the stitching time by 40%.
Do you believe the other call to find_neighbors() could be avoided somehow
by manually performing the required actions directly in stitch_meshes()
with the information available there? Chances are this would dramatically
improve the efficiency of mesh stitching in libMesh.
Thanks,
Dana
PS: The attached file stitches three cubes and can be used to investigate
the question.
// C++ include files that we need
#include <iostream>
#include <sstream>
#include <ctime>
#include <cstdlib>
//Basic include file needed for the mesh functionality.
#include "libmesh/libmesh.h"
#include "libmesh/mesh.h"
#include "libmesh/gmv_io.h"
#include "libmesh/libmesh_logging.h"
#include "libmesh/getpot.h"
// Include file that defines various mesh generation utilities
#include "libmesh/mesh_generation.h"
// Bring in everything from the libMesh namespace
using namespace libMesh;
int main (int argc, char** argv)
{
// Initialization
LibMeshInit init (argc, argv);
// Skip this 2D example if libMesh was compiled as 1D-only.
libmesh_example_assert(3 <= LIBMESH_DIM, "3D support");
// A brief message to the user to inform her of the
// exact name of the program being run, and its command line.
std::cout << "Running " << argv[0];
for (int i=1; i<argc; i++)
std::cout << " " << argv[i];
std::cout << std::endl << std::endl;
GetPot command_line (argc, argv);
unsigned int n = 0;
if ( command_line.search(1, "-n") )
n = command_line.next(n);
// Create 3 meshes, with dimensions to be overridden later, distributed
// across the default MPI communicator.
SerialMesh mesh0(init.comm());
SerialMesh mesh1(init.comm());
SerialMesh mesh2(init.comm());
// Use the MeshTools::Generation mesh generator to create uniform
// 3D grids: one unit cube on [0,1]^3, another one shifted by 1 in
// the +X direction and a third one shifted by 1 in both +X and +Z
// (L-shape)
MeshTools::Generation::build_cube(mesh0, n, n, n, 0., 1., 0., 1., 0., 1.);
MeshTools::Generation::build_cube(mesh1, n, n, n, 1., 2., 0., 1., 0., 1.);
MeshTools::Generation::build_cube(mesh2, n, n, n, 1., 2., 0., 1., 1., 2.);
// Stitch the second cube to the first one and the third one to the first one as well using the N^2 method
START_LOG("Stitching", "main");
mesh0.stitch_meshes(mesh1, 2, 4, TOLERANCE);
mesh0.stitch_meshes(mesh2, 5, 0, TOLERANCE);
STOP_LOG("Stitching", "main");
// Save the mesh to disk
GMVIO(mesh0).write("stitched_mesh.gmv");
//mesh0.write("stitched_mesh.exo");
return 0;
}
------------------------------------------------------------------------------
WatchGuard Dimension instantly turns raw network data into actionable
security intelligence. It gives you real-time visual feedback on key
security issues and trends. Skip the complicated setup - simply import
a virtual appliance and go from zero to informed in seconds.
http://pubads.g.doubleclick.net/gampad/clk?id=123612991&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
