divergence_kerngen.hpp File Reference

Team-based matrix-free Divergence operator for the spherical shell. More...

#include "../../quadrature/quadrature.hpp"
#include "communication/shell/communication.hpp"
#include "communication/shell/communication_plan.hpp"
#include "dense/vec.hpp"
#include "fe/wedge/integrands.hpp"
#include "fe/wedge/kernel_helpers.hpp"
#include "grid/shell/spherical_shell.hpp"
#include "linalg/operator.hpp"
#include "linalg/trafo/local_basis_trafo_normal_tangential.hpp"
#include "linalg/vector.hpp"
#include "linalg/vector_q1.hpp"
#include "util/timer.hpp"

Go to the source code of this file.

Classes
class	terra::fe::wedge::operators::shell::DivergenceKerngen< ScalarT >

Namespaces
namespace	terra
namespace	terra::fe
namespace	terra::fe::wedge
	Features for wedge elements.
namespace	terra::fe::wedge::operators
namespace	terra::fe::wedge::operators::shell

Detailed Description

Team-based matrix-free Divergence operator for the spherical shell.

This is a performance-oriented variant of Divergence (fe/wedge/operators/shell/divergence.hpp) that transfers the optimisation techniques used in EpsilonDivDivKerngen:

• Kokkos::TeamPolicy with backend-aware tile sizing.
• Per-team shared-memory staging of coords, radii and velocity src.
• Host-side BC-aware kernel path dispatch (Slow / FastDirichletNeumann / FastFreeslip) with no in-kernel branching on the path.
• LaunchBounds<128, 5> for occupancy tuning on CUDA.
• ShellBoundaryCommPlan-based halo exchange.

The kernel path math is intentionally kept identical to the original Divergence: every cell assembles the same local 6x18 element matrix (dense::Mat< ScalarT, 6, 18 > A[2]), applies the same boundary mask / freeslip rotation, and scatters through the same atomically_add_local_wedge_scalar_coefficients helper onto the coarse pressure grid (x/2, y/2, r/2). The only difference in the fast paths is that per-cell input data is read from team shared memory rather than from global memory. This gives a safe, first-pass optimisation that keeps correctness provable against the existing operator.