TasGrid Namespace Reference

Encapsulates the Tasmanian Sparse Grid module. More...

Namespaces
	AccelerationMeta
	Common methods for manipulating acceleration options and reading CUDA environment properties.
	AccHandle
	Type tags for the different handle types.
	HierarchyManipulations
	Collection of algorithm to manipulate multi-indexes.
	IO
	Collection of I/O handling templates.
	Maths
	Math functions and constants.
	MultiIndexManipulations
	Collection of algorithm to manipulate multi-indexes.
	OneDimensionalMeta
	Metadata for one dimensional rules, number of points, exactness, I/O, etc.
	OneDimensionalNodes
	Contains algorithms for generating Gauss,Chebyshev, and Fourier nodes and weights.
	TasGpu
	Wrappers around custom CUDA kernels to handle domain transforms and basis evaluations, the kernels are instantiated in tsgCudaKernels.cu.
	TasmanianDenseSolver
	Methods for dense linear algebra.
	TasmanianFourierTransform
	Methods for Fast-Fourier-Transform.
	TasmanianTridiagonalSolver
	Methods for tridiagonal eigenvalue problems.
	TasSparse
	Methods for sparse linear algebra.
	Utils
	Miscellaneous utility templates (internal use).

Classes
class	TasmanianSparseGrid
	The master-class that represents an instance of a Tasmanian sparse grid. More...
class	Data2D
	Generic 2D data structure divided into contiguous strips of fixed length (similar to a matrix). More...
class	MultiIndexSet
	Class that stores multi-indexes in sorted (lexicographical) order. More...
class	StorageSet
	Class that stores values, i.e., model outputs, the order of the values is in sync with the order of some MultiIndexSet. More...
struct	HandleDeleter
	Deleter template for the GPU handles, e.g., cuBlas and rocBlas. More...
class	GpuVector
	Template class that wraps around a single GPU array, providing functionality that mimics std::vector. More...
struct	GpuEngine
	Wrapper class around calls GPU accelerated linear algebra libraries. More...
class	AccelerationDomainTransform
	Implements the domain transform algorithms in case the user data is provided on the GPU. More...
struct	AccelerationContext
	Wrapper class around GPU device ID, acceleration type and GpuEngine. More...
class	CacheLagrange
	Cache that holds the values of 1D Lagrange polynomials. More...
class	CacheLagrangeDerivative
	Cache that holds the derivatives of 1D Lagrange polynomials. Uses the same interface as CacheLagrange. . More...
class	CustomTabulated
	Class providing manipulation of custom tabulated rules, file I/O and structured access to the points, weights, and meta-data. More...
class	OneDimensionalWrapper
	A class to cache one dimensional rules, nodes, weight, meta-data, etc. More...
class	TableGaussPatterson
	Rule with hard-corded tabulated points and weights. More...
struct	NodeData
	Holds the pair of point index and model value, the struct is used in a std::forward_list. More...
struct	TensorData
	Holds the description of a single tensor candidate for inclusion into the grid. More...
class	DynamicConstructorDataGlobal
	Helper class that stores data from dynamic construction of a Global grid. More...
struct	SimpleConstructData
	Holds a std::forward_list of pairs of points indexes and values, and a MultiIndexSet of initial nodes. More...
class	CandidateManager
	Manages candidate points. More...
class	CompleteStorage
	Stores complete set of points before adding to the sparse grid. More...
class	VectorToStreamBuffer
	Coverts a vector to basic stream-buffer. More...

Typedefs
using	int_gpu_lapack = int
	Defines the integer used by the LAPACK methods, usually int but DPC++ uses int64_t. More...
using	ModelSignature = std::function< void(std::vector< double > const &x, std::vector< double > &y, size_t thread_id)>
	Signature of a model function to be used in the construction procedures. More...
using	ModelSignatureMPI = std::function< void(std::vector< double > const &x, std::vector< double > &y)>
	Signature of a model function to be used in the MPI construction procedures. More...

Enumerations
enum	TypeIndexRelation { type_abeforeb , type_bbeforea , type_asameb }
	Describes the relation between two multi-indexes when compared during sorting. More...
enum	TypeDepth {   type_none , type_level , type_curved , type_hyperbolic ,   type_iptotal , type_qptotal , type_ipcurved , type_qpcurved ,   type_iphyperbolic , type_qphyperbolic , type_tensor , type_iptensor ,   type_qptensor }
	Used by Global Sequence and Fourier grids, indicates the selection criteria. More...
enum	TypeOneDRule {   rule_none , rule_clenshawcurtis , rule_clenshawcurtis0 , rule_fejer2 ,   rule_chebyshev , rule_chebyshevodd , rule_leja , rule_lejaodd ,   rule_rleja , rule_rlejadouble2 , rule_rlejadouble4 , rule_rlejaodd ,   rule_rlejashifted , rule_rlejashiftedeven , rule_rlejashifteddouble , rule_maxlebesgue ,   rule_maxlebesgueodd , rule_minlebesgue , rule_minlebesgueodd , rule_mindelta ,   rule_mindeltaodd , rule_gausslegendre , rule_gausslegendreodd , rule_gausspatterson ,   rule_gausschebyshev1 , rule_gausschebyshev1odd , rule_gausschebyshev2 , rule_gausschebyshev2odd ,   rule_gaussgegenbauer , rule_gaussgegenbauerodd , rule_gaussjacobi , rule_gaussjacobiodd ,   rule_gausslaguerre , rule_gausslaguerreodd , rule_gausshermite , rule_gausshermiteodd ,   rule_customtabulated , rule_localp , rule_localp0 , rule_semilocalp ,   rule_localpb , rule_wavelet , rule_fourier }
	Used to specify the one dimensional family of rules that induces the sparse grid. More...
enum	TypeRefinement {   refine_classic , refine_parents_first , refine_direction_selective , refine_fds ,   refine_stable , refine_none }
	Refinement strategy for local polynomial and wavelet grids. More...
enum	TypeAcceleration {   accel_none , accel_cpu_blas , accel_gpu_default , accel_gpu_cublas ,   accel_gpu_cuda , accel_gpu_magma }
	Modes of acceleration. More...

Functions
TasmanianSparseGrid	makeEmpty ()
	Returns an empty sparse grid. More...
TasmanianSparseGrid	makeGlobalGrid (int dimensions, int outputs, int depth, TypeDepth type, TypeOneDRule rule, std::vector< int > const &anisotropic_weights=std::vector< int >(), double alpha=0.0, double beta=0.0, const char *custom_filename=nullptr, std::vector< int > const &level_limits=std::vector< int >())
	Factory method, creates a new grid and calls TasmanianSparseGrid::makeGlobalGrid(). More...
TasmanianSparseGrid	makeSequenceGrid (int dimensions, int outputs, int depth, TypeDepth type, TypeOneDRule rule, std::vector< int > const &anisotropic_weights=std::vector< int >(), std::vector< int > const &level_limits=std::vector< int >())
	Factory method, creates a new grid and calls TasmanianSparseGrid::makeSequenceGrid(). More...
TasmanianSparseGrid	makeLocalPolynomialGrid (int dimensions, int outputs, int depth, int order=1, TypeOneDRule rule=rule_localp, std::vector< int > const &level_limits=std::vector< int >())
	Factory method, creates a new grid and calls TasmanianSparseGrid::makeLocalPolynomialGrid(). More...
TasmanianSparseGrid	makeWaveletGrid (int dimensions, int outputs, int depth, int order=1, std::vector< int > const &level_limits=std::vector< int >())
	Factory method, creates a new grid and calls TasmanianSparseGrid::makeWaveletGrid(). More...
TasmanianSparseGrid	makeFourierGrid (int dimensions, int outputs, int depth, TypeDepth type, std::vector< int > const &anisotropic_weights=std::vector< int >(), std::vector< int > const &level_limits=std::vector< int >())
	Factory method, creates a new grid and calls TasmanianSparseGrid::makeFourierGrid(). More...
TasmanianSparseGrid	readGrid (const char *filename)
	Factory method, creates a new grid and calls TasmanianSparseGrid::read(). More...
TasmanianSparseGrid	readGrid (std::string const &filename)
	Overload using std::string. More...
TasmanianSparseGrid	copyGrid (TasmanianSparseGrid const &source, int outputs_begin=0, int outputs_end=-1)
	Returns a grid that is a copy of the source. More...
template<typename T >
std::vector< T >	spltVector2D (std::vector< T > const &x, size_t stride, int ibegin, int iend)
	Take a vector logically organized into stips and strides and extract a sub-strip from every strip. More...
template<typename >
void	deleteHandle (int *)
	Deletes the handle, specialized for each TPL backend and tag in TasGrid::AccHandle namepace. More...
CustomTabulated	getSubrules (CustomTabulated &ct, int start_index, int stride, std::string description)
	Generates a subset of rules as a CustomTabulated object. The subset has a short description string given by description and takes the levels of the input ct starting from level start_index with distance stride between consecutive levels. More...
template<class T >
std::vector< const T * >	makeReverseReferenceVector (const std::forward_list< T > &list)
	Takes a list and creates a vector of references in reversed order, needed for I/O so that read can be followed by push_front. . More...
template<bool use_ascii>
void	writeNodeDataList (const std::forward_list< NodeData > &data, std::ostream &os)
	Writes a NodeData std::forward_list to a file using either binary or ascii format. . More...
template<typename iomode >
std::forward_list< NodeData >	readNodeDataList (std::istream &is, size_t num_dimensions, size_t num_outputs)
	Reads a NodeData std::forward_list from a file using either binary or ascii format. . More...
template<typename iomode >
std::forward_list< TensorData >	readTensorDataList (std::istream &is, size_t num_dimensions)
	Reads a TensorData std::forward_list from a file using either binary or ascii format. More...
template<class RuleLike >
std::vector< double >	listToNodes (std::forward_list< NodeData > const &node_list, size_t num_dimensions, RuleLike const &rule)
	Using MultiIndexManipulations::indexesToNodes() convert the node_list to actual points according to the rule. More...
template<typename callable_method >
std::vector< double >	listToLocalNodes (std::forward_list< NodeData > const &node_list, size_t num_dimensions, callable_method rule)
	Using MultiIndexManipulations::indexesToNodes() convert the node_list to actual points according to the rule. More...
double	poly_eval (const std::vector< double > &roots, double x)
	Evaluates a polynomial with roots given by roots at the point x. More...
double	lagrange_eval (size_t idx, const std::vector< double > &roots, double x)
	Evaluates a Lagrange basis polynomial at a point x. More...
template<bool is_symmetric>
void	getGaussNodesAndWeights (const int n, const std::vector< double > &ref_points, const std::vector< double > &ref_weights, std::vector< std::vector< double >> &points_cache, std::vector< std::vector< double >> &weights_cache)
	Generates n levels of points and weights using orthogonal polynomials with respect to ref_points and ref_weights. More...
TasGrid::CustomTabulated	getShiftedExoticQuadrature (const int n, const double shift, const std::vector< double > &shifted_weights, const std::vector< double > &ref_points, const char *description, const bool is_symmetric=false)
	Generate the exotic quadrature points and weights and load them into a TasGrid::CustomTabulated object. More...
void	shiftReferenceWeights (std::vector< double > const &vals, double shift, std::vector< double > &ref_weights)
	Multiplies the reference quadrature weights by the shifted values of the weight function. More...
TasGrid::CustomTabulated	getExoticQuadrature (const int num_levels, const double shift, const TasGrid::TasmanianSparseGrid &grid, const char *description, const bool is_symmetric=false)
	Constructs an exotic quadrature from a sparse grid object. More...
TasGrid::CustomTabulated	getExoticQuadrature (const int num_levels, const double shift, std::function< double(double)> weight_fn, const int num_ref_points, const char *description, const bool is_symmetric=false)
	Similar to getExoticQuadrature() but the weight function is defined by a lambda expression. More...
template<bool parallel_construction, bool use_initial_guess>
void	constructCommon (ModelSignature model, size_t max_num_points, size_t num_parallel_jobs, size_t max_samples_per_job, TasmanianSparseGrid &grid, std::function< std::vector< double >(TasmanianSparseGrid &)> candidates, std::string const &checkpoint_filename)
	Construction algorithm using generic candidates procedure. More...
template<bool parallel_construction = TasGrid::mode_parallel, bool initial_guess = no_initial_guess>
void	constructSurrogate (ModelSignature model, size_t max_num_points, size_t num_parallel_jobs, size_t max_samples_per_job, TasmanianSparseGrid &grid, double tolerance, TypeRefinement criteria, int output=-1, std::vector< int > const &level_limits=std::vector< int >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda. More...
template<bool parallel_construction = TasGrid::mode_parallel, bool initial_guess = no_initial_guess>
void	constructSurrogate (ModelSignature model, size_t max_num_points, size_t num_parallel_jobs, size_t max_samples_per_job, TasmanianSparseGrid &grid, TypeDepth type, std::vector< int > const &anisotropic_weights=std::vector< int >(), std::vector< int > const &level_limits=std::vector< int >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda. More...
template<bool parallel_construction = TasGrid::mode_parallel, bool initial_guess = no_initial_guess>
void	constructSurrogate (ModelSignature model, size_t max_num_points, size_t num_parallel_jobs, size_t max_samples_per_job, TasmanianSparseGrid &grid, TypeDepth type, int output, std::vector< int > const &level_limits=std::vector< int >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda. More...
bool	hasGPUBasis (TasmanianSparseGrid const &grid)
	Returns true if the grid has a GPU algorithm for computing the hierarchical basis. More...
template<typename scalar_type >
void	generateCoefficientsGPU (double const data_points[], int num_data, scalar_type model_values[], double tolerance, TasmanianSparseGrid &grid)
	Generates the coefficients and solves the least-squares problem. More...
template<typename scalar_type >
Data2D< scalar_type >	generateCoefficients (double const data_points[], int num_data, double const model_values[], double tolerance, TasmanianSparseGrid &grid)
	Generates the coefficients and solves the least-squares problem. More...
template<typename scalar_type >
void	loadUnstructuredDataL2tmpl (double const data_points[], int num_data, double const model_values[], double tolerance, TasmanianSparseGrid &grid)
	Template implementation that handles the case of Fourier grids vs. all other types. More...
void	loadUnstructuredDataL2 (double const data_points[], int num_data, double const model_values[], double tolerance, TasmanianSparseGrid &grid)
	Construct a sparse grid surrogate using a least-squares fit. More...
void	loadUnstructuredDataL2 (std::vector< double > const &data_points, std::vector< double > const &model_values, double tolerance, TasmanianSparseGrid &grid)
	Overload that used vectors and infers the number of data points from the size of the vectors. More...
template<bool use_initial_guess>
void	mpiConstructCommon (ModelSignatureMPI model, int num_dimensions, int num_outputs, size_t max_num_points, size_t max_samples_per_job, size_t max_num_ranks, int tagx, int tagy, int root, MPI_Comm comm, TasmanianSparseGrid &grid, std::function< std::vector< double >(TasmanianSparseGrid &)> candidates, std::string const &checkpoint_filename)
	MPI construction algorithm using generic candidates procedure. More...
template<bool use_initial_guess = no_initial_guess>
void	mpiConstructSurrogate (ModelSignatureMPI model, int num_dimensions, int num_outputs, size_t max_num_points, size_t max_samples_per_job, size_t max_num_ranks, int tagx, int tagy, int root, MPI_Comm comm, TasmanianSparseGrid &grid, double tolerance, TypeRefinement criteria, int output=-1, std::vector< int > const &level_limits=std::vector< int >(), std::vector< double > const &scale_correction=std::vector< double >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda, MPI version. More...
template<bool use_initial_guess = no_initial_guess>
void	mpiConstructSurrogate (ModelSignatureMPI model, int num_dimensions, int num_outputs, size_t max_num_points, size_t max_samples_per_job, size_t max_num_ranks, int tagx, int tagy, int root, MPI_Comm comm, TasmanianSparseGrid &grid, TypeDepth type, std::vector< int > const &anisotropic_weights=std::vector< int >(), std::vector< int > const &level_limits=std::vector< int >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda, MPI version. More...
template<bool use_initial_guess = no_initial_guess>
void	mpiConstructSurrogate (ModelSignatureMPI model, int num_dimensions, int num_outputs, size_t max_num_points, size_t max_samples_per_job, size_t max_num_ranks, int tagx, int tagy, int root, MPI_Comm comm, TasmanianSparseGrid &grid, TypeDepth type, int output, std::vector< int > const &level_limits=std::vector< int >(), std::string const &checkpoint_filename=std::string())
	Construct a sparse grid surrogate to the model defined by the lambda, MPI version. More...
template<bool use_initial_guess = no_initial_guess>
void	mpiConstructWorker (ModelSignatureMPI model, int num_dimensions, int num_outputs, size_t max_samples_per_job, size_t max_num_ranks, int tagx, int tagy, int root, MPI_Comm comm)
	Executes the worker (i.e., non-root) portion of the MPI sampling. More...
int	getMPIRank (MPI_Comm comm)
	Utility to return the rank within the given comm. More...
template<bool binary = TasGrid::mode_binary>
int	MPIGridSend (TasmanianSparseGrid const &grid, int destination, int tag, MPI_Comm comm)
	Send a grid to another process in the MPI comm. More...
template<bool binary = TasGrid::mode_binary>
int	MPIGridRecv (TasmanianSparseGrid &grid, int source, int tag, MPI_Comm comm, MPI_Status *status=MPI_STATUS_IGNORE)
	Receive a grid from another process in the MPI comm. More...
template<bool binary = TasGrid::mode_binary>
int	MPIGridBcast (TasmanianSparseGrid &grid, int root, MPI_Comm comm)
	Broadcast a grid to all processes in an MPI comm. More...
template<bool binary = TasGrid::mode_binary>
int	MPIGridScatterOutputs (TasmanianSparseGrid const &source, TasmanianSparseGrid &destination, int root, int tag, MPI_Comm comm)
	Split the grid across the comm where each rank receives an equal portion of the total outputs. More...
template<bool parallel_construction = true, bool overwrite_loaded = false>
void	loadNeededValues (std::function< void(double const x[], double y[], size_t thread_id)> model, TasmanianSparseGrid &grid, size_t num_threads)
	Loads the current grid with model values, does not perform any refinement. More...
template<bool parallel_construction = true, bool overwrite_loaded = false>
void	loadNeededValues (std::function< void(std::vector< double > const &x, std::vector< double > &y, size_t thread_id)> model, TasmanianSparseGrid &grid, size_t num_threads)
	Overload that uses vectors for the model inputs and outputs. More...
template<bool parallel_construction = true, bool overwrite_loaded = false>
void	loadNeededPoints (std::function< void(double const x[], double y[], size_t thread_id)> model, TasmanianSparseGrid &grid, size_t num_threads)
	Alias to loadNeededValues(), array variant.
template<bool parallel_construction = true, bool overwrite_loaded = false>
void	loadNeededPoints (std::function< void(std::vector< double > const &x, std::vector< double > &y, size_t thread_id)> model, TasmanianSparseGrid &grid, size_t num_threads)
	Alias to loadNeededValues(), vector variant.

Variables
constexpr TypeAcceleration	accel_gpu_hip = accel_gpu_cuda
	At the front API, the HIP and CUDA options are equivalent, see TasGrid::TypeAcceleration.
constexpr TypeAcceleration	accel_gpu_rocblas = accel_gpu_cublas
	At the front API, the HIP and CUDA options are equivalent, see TasGrid::TypeAcceleration.
constexpr bool	mode_ascii = false
	Constant allowing for more expressive selection of ascii and binary mode in IO methods.
constexpr bool	mode_binary = true
	Constant allowing for more expressive selection of ascii and binary mode in IO methods.
constexpr bool	mode_parallel = true
	Allows for expressive calls to TasGrid::constructSurrogate().
constexpr bool	mode_sequential = false
	Allows for expressive calls to TasGrid::constructSurrogate().
constexpr bool	with_initial_guess = true
	Allows for expressive calls to TasGrid::constructSurrogate().
constexpr bool	no_initial_guess = false
	Allows for expressive calls to TasGrid::constructSurrogate().

Detailed Description

Encapsulates the Tasmanian Sparse Grid module.

Typedef Documentation

int_gpu_lapack

using TasGrid::int_gpu_lapack = typedef int

Defines the integer used by the LAPACK methods, usually int but DPC++ uses int64_t.