evolution::initial_data Namespace Reference

Namespace for things related to initial data used for evolution systems. More...

Namespaces
namespace	OptionTags
	Option tags for initial data of evolution system.
namespace	Tags
	Tags for initial data of evolution system.

Classes
class	InitialData
	The abstract base class for initial data of evolution systems. All analytic solutions and analytic data must virtually inherit from this class. More...
class	NumericData
	Load numeric data from volume data files. More...
class	WithNoise
	Wraps any analytic InitialData and adds uniform random noise to selected evolution variables after the inner solution has initialized them. More...

Enumerations
enum class	NoiseAmplitudeType : uint8_t { Absolute , Relative }
	How the noise amplitude passed to add_noise_to_tensor is interpreted. More...

Functions
bool	operator== (const NumericData &lhs, const NumericData &rhs)
bool	operator!= (const NumericData &lhs, const NumericData &rhs)
std::ostream &	operator<< (std::ostream &os, NoiseAmplitudeType value)
template<typename TensorType>
void	add_noise_to_tensor (gsl::not_null< TensorType * > tensor, double amplitude, NoiseAmplitudeType amplitude_type, size_t element_seed, size_t component_offset)
	Add uniform random noise to all stored components of tensor.
template<size_t Dim>
size_t	make_element_seed (size_t base_seed, const tnsr::I< DataVector, Dim, Frame::Inertial > &inertial_coords)
	Combine base_seed with the first grid point of inertial_coords to produce a per-element RNG seed.

Detailed Description

Namespace for things related to initial data used for evolution systems.

Enumeration Type Documentation

NoiseAmplitudeType

enum class evolution::initial_data::NoiseAmplitudeType : uint8_t

strong

How the noise amplitude passed to add_noise_to_tensor is interpreted.

Enumerator
Absolute	The noise is drawn from \([-A,\,A]\).
Relative	The noise is drawn from \([-A M,\,A M]\), where \(M\) is the maximum absolute value over all stored components and grid points of the tensor being perturbed.

Function Documentation

add_noise_to_tensor()

template<typename TensorType>

void evolution::initial_data::add_noise_to_tensor	(	gsl::not_null< TensorType * >	tensor,
		double	amplitude,
		NoiseAmplitudeType	amplitude_type,
		size_t	element_seed,
		size_t	component_offset )

Add uniform random noise to all stored components of tensor.

Each independent stored component receives noise drawn independently from \(\text{Uniform}[-A,\,A]\). For NoiseAmplitudeType::Absolute, \(A\) is amplitude. For NoiseAmplitudeType::Relative, \(A\) is amplitude times the maximum absolute value over all stored components and grid points of tensor, so a tensor that is identically zero receives no noise. The RNG seed for component \(i\) is hash(element_seed, component_offset + i), so calling this function with different component_offset values keeps separate tensor fields independent even when element_seed is the same.

Parameters

tensor	Tensor to perturb in place.
amplitude	Sets \(\pm A\) of the noise interval (absolute or relative, depending on amplitude_type).
amplitude_type	Whether amplitude is absolute or relative to the maximum absolute value of tensor.
element_seed	Per-element seed, typically constructed by hashing a base seed with the inertial coordinates of the element's first grid point.
component_offset	Offset added to the component index before hashing, used to keep different tensor fields independent.

make_element_seed()

template<size_t Dim>

size_t evolution::initial_data::make_element_seed	(	size_t	base_seed,
		const tnsr::I< DataVector, Dim, Frame::Inertial > &	inertial_coords )

Combine base_seed with the first grid point of inertial_coords to produce a per-element RNG seed.

Because most domains have elements have distinct first grid points, this gives independent noise on almost every element even when base_seed is the same.