/* ************************************************************************ * Copyright (C) 2018-2021 Advanced Micro Devices, Inc. All rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to deal * in the Software without restriction, including without limitation the rights * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell * copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in * all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN * THE SOFTWARE. * * ************************************************************************ */ #ifndef ROCALUTION_PRECONDITIONER_AI_HPP_ #define ROCALUTION_PRECONDITIONER_AI_HPP_ #include "../solver.hpp" #include "preconditioner.hpp" #include "rocalution/export.hpp" namespace rocalution { /** \ingroup precond_module * \class AIChebyshev * \brief Approximate Inverse - Chebyshev Preconditioner * \details * The Approximate Inverse - Chebyshev Preconditioner is an inverse matrix * preconditioner with values from a linear combination of matrix-valued * Chebyshev polynomials. * \cite chebpoly * * \tparam OperatorType - can be LocalMatrix * \tparam VectorType - can be LocalVector * \tparam ValueType - can be float, double, std::complex or std::complex */ template class AIChebyshev : public Preconditioner { public: ROCALUTION_EXPORT AIChebyshev(); ROCALUTION_EXPORT virtual ~AIChebyshev(); ROCALUTION_EXPORT virtual void Print(void) const; ROCALUTION_EXPORT virtual void Solve(const VectorType& rhs, VectorType* x); /** \brief Set order, min and max eigenvalues */ ROCALUTION_EXPORT void Set(int p, ValueType lambda_min, ValueType lambda_max); ROCALUTION_EXPORT virtual void Build(void); ROCALUTION_EXPORT virtual void Clear(void); protected: virtual void MoveToHostLocalData_(void); virtual void MoveToAcceleratorLocalData_(void); private: OperatorType AIChebyshev_; int p_; ValueType lambda_min_, lambda_max_; }; /** \ingroup precond_module * \class FSAI * \brief Factorized Approximate Inverse Preconditioner * \details * The Factorized Sparse Approximate Inverse preconditioner computes a direct * approximation of \f$M^{-1}\f$ by minimizing the Frobenius norm \f$||I - GL||_{F}\f$, * where \f$L\f$ denotes the exact lower triangular part of \f$A\f$ and \f$G:=M^{-1}\f$. * The FSAI preconditioner is initialized by \f$q\f$, based on the sparsity pattern of * \f$|A^{q}|\f$. However, it is also possible to supply external sparsity patterns in form * of the LocalMatrix class. * \cite kolotilina * * \note * The FSAI preconditioner is only suited for symmetric positive definite matrices. * * \tparam OperatorType - can be LocalMatrix * \tparam VectorType - can be LocalVector * \tparam ValueType - can be float, double, std::complex or std::complex */ template class FSAI : public Preconditioner { public: ROCALUTION_EXPORT FSAI(); ROCALUTION_EXPORT virtual ~FSAI(); ROCALUTION_EXPORT virtual void Print(void) const; ROCALUTION_EXPORT virtual void Solve(const VectorType& rhs, VectorType* x); /** \brief Set the power of the system matrix sparsity pattern */ ROCALUTION_EXPORT void Set(int power); /** \brief Set an external sparsity pattern */ ROCALUTION_EXPORT void Set(const OperatorType& pattern); ROCALUTION_EXPORT virtual void Build(void); ROCALUTION_EXPORT virtual void Clear(void); /** \brief Set the matrix format of the preconditioner */ ROCALUTION_EXPORT void SetPrecondMatrixFormat(unsigned int mat_format, int blockdim = 1); protected: virtual void MoveToHostLocalData_(void); virtual void MoveToAcceleratorLocalData_(void); private: OperatorType FSAI_L_; OperatorType FSAI_LT_; VectorType t_; int matrix_power_; bool external_pattern_; const OperatorType* matrix_pattern_; // Keep the precond matrix in CSR or not bool op_mat_format_; // Precond matrix format unsigned int precond_mat_format_; // Matrix format block dimension int format_block_dim_; }; /** \ingroup precond_module * \class SPAI * \brief SParse Approximate Inverse Preconditioner * \details * The SParse Approximate Inverse algorithm is an explicitly computed preconditioner for * general sparse linear systems. In its current implementation, only the sparsity * pattern of the system matrix is supported. The SPAI computation is based on the * minimization of the Frobenius norm \f$||AM - I||_{F}\f$. * \cite grote * * \tparam OperatorType - can be LocalMatrix * \tparam VectorType - can be LocalVector * \tparam ValueType - can be float, double, std::complex or std::complex */ template class SPAI : public Preconditioner { public: ROCALUTION_EXPORT SPAI(); ROCALUTION_EXPORT virtual ~SPAI(); ROCALUTION_EXPORT virtual void Print(void) const; ROCALUTION_EXPORT virtual void Solve(const VectorType& rhs, VectorType* x); ROCALUTION_EXPORT virtual void Build(void); ROCALUTION_EXPORT virtual void Clear(void); /** \brief Set the matrix format of the preconditioner */ ROCALUTION_EXPORT void SetPrecondMatrixFormat(unsigned int mat_format, int blockdim = 1); protected: virtual void MoveToHostLocalData_(void); virtual void MoveToAcceleratorLocalData_(void); private: OperatorType SPAI_; // Keep the precond matrix in CSR or not bool op_mat_format_; // Precond matrix format unsigned int precond_mat_format_; // Matrix format block dimension int format_block_dim_; }; /** \ingroup precond_module * \class TNS * \brief Truncated Neumann Series Preconditioner * \details * The Truncated Neumann Series (TNS) preconditioner is based on * \f$M^{-1} = K^{T} D^{-1} K\f$, where \f$K=(I-LD^{-1}+(LD^{-1})^{2})\f$, with the * diagonal \f$D\f$ of \f$A\f$ and the strictly lower triangular part \f$L\f$ of * \f$A\f$. The preconditioner can be computed in two forms - explicitly and implicitly. * In the explicit form, the full construction of \f$M\f$ is performed via matrix-matrix * operations, whereas in the implicit from, the application of the preconditioner is * based on matrix-vector operations only. The matrix format for the stored matrices can * be specified. * * \tparam OperatorType - can be LocalMatrix * \tparam VectorType - can be LocalVector * \tparam ValueType - can be float, double, std::complex or std::complex */ template class TNS : public Preconditioner { public: ROCALUTION_EXPORT TNS(); ROCALUTION_EXPORT virtual ~TNS(); ROCALUTION_EXPORT virtual void Print(void) const; /** \brief Set implicit (true) or explicit (false) computation */ ROCALUTION_EXPORT void Set(bool imp); ROCALUTION_EXPORT virtual void Solve(const VectorType& rhs, VectorType* x); ROCALUTION_EXPORT virtual void Build(void); ROCALUTION_EXPORT virtual void Clear(void); /** \brief Set the matrix format of the preconditioner */ ROCALUTION_EXPORT void SetPrecondMatrixFormat(unsigned int mat_format, int blockdim = 1); protected: virtual void MoveToHostLocalData_(void); virtual void MoveToAcceleratorLocalData_(void); private: OperatorType L_; OperatorType LT_; OperatorType TNS_; VectorType Dinv_; VectorType tmp1_; VectorType tmp2_; // Keep the precond matrix in CSR or not bool op_mat_format_; // Precond matrix format unsigned int precond_mat_format_; // Matrix format block dimension int format_block_dim_; // implicit (true) or explicit (false) computation bool impl_; }; } // namespace rocalution #endif // ROCALUTION_PRECONDITIONER_AI_HPP_