ProgramContext – the C++-side facade a generated problem.so calls to run a compiled time Program during sim.step(dt) (epic ADC-399, ADC-401 Phase 2b). More...

#include <chrono>
#include <cstdint>
#include <functional>
#include <optional>
#include <stdexcept>
#include <string>
#include <utility>
#include <pops/core/foundation/types.hpp>
#include <pops/runtime/program/profiler.hpp>
#include <pops/coupling/schur/core/schur_condensation.hpp>
#include <pops/mesh/boundary/physical_bc.hpp>
#include <pops/mesh/execution/for_each.hpp>
#include <pops/mesh/geometry/geometry.hpp>
#include <pops/mesh/storage/fab2d.hpp>
#include <pops/mesh/storage/mf_arith.hpp>
#include <pops/mesh/storage/multifab.hpp>
#include <pops/numerics/elliptic/interface/elliptic_problem.hpp>
#include <pops/numerics/elliptic/mg/geometric_mg.hpp>
#include <pops/numerics/elliptic/poisson/poisson_operator.hpp>
#include <pops/numerics/linalg/lorentz_eliminator.hpp>
#include <pops/runtime/config/runtime_params.hpp>
#include <pops/runtime/context/grid_context.hpp>
#include <pops/runtime/program/cache_manager.hpp>
#include <pops/runtime/system.hpp>

Include dependency graph for program_context.hpp:

Go to the source code of this file.

Classes
struct	pops::runtime::program::detail::SchurOperatorCoeffKernelC
	Aux-component-aware variants of the native Schur kernels (coupling/schur/core/schur_condensation.hpp. More...

struct	pops::runtime::program::detail::SchurExplicitFluxKernelC
	out = B^{-1} (mx, my) at the center (Fx in comp 0, Fy in comp 1): the explicit flux F = rhoB^{-1}v. More...

struct	pops::runtime::program::detail::SchurRhsAssembleKernelC
	rhs = -Lap phi^n - g*div(F), the centered FV divergence of the explicit flux F packed in ONE 2-component buffer (Fx in comp 0, Fy in comp 1 – the layout schur_explicit_flux writes), fused with the already-negated -Lap phi^n. More...

struct	pops::runtime::program::detail::SchurReconstructKernelC
	Reconstruct v^{n+theta} = B^{-1}(v^n - thetadtgrad phi) and write mom = rho^n*v (rho frozen). More...

struct	pops::runtime::program::detail::SchurEnergyKernelC
	Condensed-Schur kinetic-energy increment (ADC-427), mirroring the native detail::SchurEnergyKernel: E^{n+1} = E^n + (1/2) rho (\|v^{n+1}\|^2 - \|v^n\|^2), v = (mx, my)/rho. More...

class	pops::runtime::program::ProgramContext

Namespaces
namespace	pops

namespace	pops::runtime

namespace	pops::runtime::program

namespace	pops::runtime::program::detail

Detailed Description

ProgramContext – the C++-side facade a generated problem.so calls to run a compiled time Program during sim.step(dt) (epic ADC-399, ADC-401 Phase 2b).

It REIMPLEMENTS NOTHING. Each method forwards to an existing pops::System primitive: install(fn) -> System::install_program_step(fn) (registers the macro-step body) solve_fields() -> System::solve_fields() (elliptic solve + aux at current U) solve_fields_from_state(b, U) -> System::solve_fields_from_state(b, U) (aux at a stage state) n_blocks() -> System::n_blocks() state(b) -> System::block_state(b) (the block's live MultiFab, zero-copy) rhs_into(b, U, R) -> System::block_rhs_into(b, U, R) (R <- -div F + S, Poisson frozen) neg_div_flux_default_into(b, U, R) -> System::block_neg_div_flux_into (R <- -div F, NO source) axpy(U, a, R) -> pops::saxpy(U, a, R) (U <- U + a R, device-dispatched)

The Program composes the chain (e.g. Forward Euler = solve_fields(); for each block: rhs_into(b, U, R); axpy(U, dt, R)) and installs it via install(...). The .so NEVER touches System::Impl / Array4 / fill_boundary / the elliptic solver / Kokkos / MPI / CFL / substeps.

IDIOM: ProgramContext is a plain (non-template) class holding a System*. A generated .so receives the System as a flat void* across the dlopen boundary (like the native loader's void* self) and wraps it here; it reaches per-block storage through the System's public accessors because System::Impl is private to the _pops translation unit.

include/pops/runtime/program/program_context.hpp File Reference¶

Classes

Namespaces

Detailed Description