Differentiation::AD Namespace Reference

Namespaces
namespace	internal

Classes
struct	ADNumberTraits
class	CellLevelBase
class	EnergyFunctional
class	HelperBase
struct	is_ad_number
struct	is_adolc_number
struct	is_adolc_taped_number
struct	is_adolc_tapeless_number
struct	is_complex_valued_ad_number
struct	is_real_valued_ad_number
struct	is_sacado_dfad_number
struct	is_sacado_number
struct	is_sacado_rad_number
struct	is_taped_ad_number
struct	is_tapeless_ad_number
struct	Numbers
struct	NumberTraits
class	PointLevelFunctionsBase
class	ResidualLinearization
class	ScalarFunction
struct	TapedDrivers
struct	TapelessDrivers
struct	Types
class	VectorFunction

Enumerations
enum class	NumberTypes {   none , adolc_taped , adolc_tapeless , sacado_dfad ,   sacado_dfad_dfad , sacado_rad , sacado_rad_dfad }

Functions
static ::ExceptionBase &	ExcRequiresADNumberSpecialization ()
static ::ExceptionBase &	ExcRequiresADOLC ()
static ::ExceptionBase &	ExcSupportedDerivativeLevels (std::size_t arg1, std::size_t arg2)

Detailed Description

Wrappers for automatic differentiation libraries. Currently there is support for the following libraries:

• ADOL-C
• Sacado (a component of Trilinos)

Enumeration Type Documentation

NumberTypes

enum class Differentiation::AD::NumberTypes

strong

An enumeration to indicate which type of auto-differentiable number is to be used for computations. If a type that is selected for use is not available in the library, a run-time error will be thrown.

Enumerator
none	A dummy type for floating point numbers (i.e., non-differentiable scalar types). This option exists to facilitate the use of template meta-programming techniques to select, based on this enumeration, which auto-differentiable number type will be used to perform calculations. It will not permit any computations because the underlying number types resulting from its selection are floating point types, rather than auto-differentiable numbers.
adolc_taped	Taped forward and reverse-mode ADOL-C number type (n-differentiable). First derivatives will be computed using reverse mode, while the second derivatives will be computed using forward mode. Even higher-order derivatives can be computed using ADOL-C's own driver functions.
adolc_tapeless	Tapeless dynamic forward-mode ADOL-C number type (once differentiable).
sacado_dfad	Tapeless dynamic forward-mode Sacado number type (once differentiable).
sacado_dfad_dfad	Tapeless nested dynamic forward-mode Sacado number type (twice differentiable). Both the first and second derivatives will be computed using forward mode.
sacado_rad	Tapeless reverse-mode Sacado number type (once differentiable).
sacado_rad_dfad	Tapeless nested reverse-mode and dynamic forward-mode Sacado number type (twice differentiable). First derivatives will be computed using reverse mode, while the second derivatives will be computed using forward mode. Note that the repeated use of the nested reverse-forward mode results in a memory leak described in this Trilinos issue.

Definition at line 34 of file ad_number_types.h.