Opm::gpuistl::detail::FlexibleSolverWrapper< Matrix, Vector, Comm > Class Template Reference

FlexibleSolverWrapper is compilational trick to reduce compile time overhead. More...

#include <FlexibleSolverWrapper.hpp>

Public Types
using	AbstractSolverType = Dune::InverseOperator<Vector, Vector>
using	AbstractPreconditionerType = Dune::PreconditionerWithUpdate<Vector, Vector>
using	AbstractOperatorType = Dune::AssembledLinearOperator<Matrix, Vector, Vector>
using	AbstractOperatorPtrType = std::unique_ptr<AbstractOperatorType>
using	AbstractSolverPtrType = std::unique_ptr<AbstractSolverType>
using	GpuCommunicationType

Public Member Functions
	FlexibleSolverWrapper (const Matrix &matrix, bool parallel, const PropertyTree &prm, std::size_t pressureIndex, const std::function< Vector()> &weightCalculator, bool forceSerial, const Comm *comm)
void	update ()
void	apply (Vector &x, Vector &y, Dune::InverseOperatorResult &result)

Detailed Description

template<class Matrix, class Vector, class Comm>
class Opm::gpuistl::detail::FlexibleSolverWrapper< Matrix, Vector, Comm >

FlexibleSolverWrapper is compilational trick to reduce compile time overhead.

Essentially this class acts as a direct interface towards the FlexibleSolver class, but we want to have a clear wall between the ISTLSolverGPUISTL and the FlexibleSolver to avoid compilational overhead (the former is based on TypeTag the latter on operators).

This wrapper handles the mapping from matrices to concrete operator types.

Member Typedef Documentation

GpuCommunicationType

template<class Matrix, class Vector, class Comm>

using Opm::gpuistl::detail::FlexibleSolverWrapper< Matrix, Vector, Comm >::GpuCommunicationType

Initial value:

---

The documentation for this class was generated from the following files:

• opm/simulators/linalg/gpuistl/detail/FlexibleSolverWrapper.hpp
• opm/simulators/linalg/gpuistl/detail/FlexibleSolverWrapper.cpp