|
Mallard 0.8.0
Unstructured, GPU accelerated, high-order finite volume CFD solver
Mallard website |
Finite-rate chemistry of the cells of a mesh: each cell that needs it is an adiabatic constant-volume reactor over a splitting step, integrated by one thread or by the vector lanes of a team. More...
#include <cstdint>#include <memory>#include <Kokkos_Core.hpp>#include "kinetics.h"#include "mixture.h"#include "reactor.h"#include "sparse_lu.h"#include "state.h"Go to the source code of this file.
Classes | |
| struct | CellChemistryOptions |
| Options of CellChemistry: the reactor's tolerances, the temperature below which cells are frozen, the vector lanes and threads of the team of each cell (0: automatic; lanes = 1: one thread per cell; automatic threads on GPUs: one, and several warps for the cells whose last call was expensive), whether a team keeps its factors and vectors in scratch (shared) memory where they fit (1; default no), and whether cells integrated by one thread are ordered by their last cost (binning; on GPUs only). More... | |
| struct | ChemistryForcing |
| A constant forcing of a chemistry call: the rates dU/dt of every conservative variable of each cell (rho, rho u, rho E, rho Y_k), e.g. the transport tendency of balanced splitting or of simplified spectral deferred corrections, and a state of the same size for scratch. More... | |
| struct | ChemistryCells |
| The per-cell views a chemistry call reads and writes: the state, the temperature seed, the last sub-step, the sub-steps of the step and of the last call, the optional multiplier of dt, and with a forcing its species rates and U + dt rate, whose partial densities receive the result instead of rhoY (all empty without one). More... | |
| class | CellChemistry |
| Work memory and kernels that advance the cells of a state as adiabatic constant-volume reactors at fixed (rho, e). More... | |
| struct | CellChemistry::Statistics |
Finite-rate chemistry of the cells of a mesh: each cell that needs it is an adiabatic constant-volume reactor over a splitting step, integrated by one thread or by the vector lanes of a team.