tenex::Divide< T1, T2, Args2 > Struct Template Reference

Defines the tensor expression representing the quotient of one tensor expression divided by another tensor expression that evaluates to a rank 0 tensor. More...

#include <Divide.hpp>

Public Types
using	type = typename detail::get_binop_datatype< typename T1::type, typename T2::type >::type
	The type of the data being stored in the result of the expression.
using	symmetry = typename T1::symmetry
	The Symmetry of the result of the expression.
using	index_list = typename T1::index_list
	The list of TensorIndexTypes of the result of the expression.
using	args_list = typename T1::args_list
	The list of generic TensorIndexs of the result of the expression.

Public Member Functions
	Divide (T1 t1, T2 t2)
template<typename LhsTensor >
void	assert_lhs_tensor_not_in_rhs_expression (const gsl::not_null< LhsTensor * > lhs_tensor) const
	Assert that the LHS tensor of the equation does not also appear in this expression's subtree.
template<typename LhsTensorIndices , typename LhsTensor >
void	assert_lhs_tensorindices_same_in_rhs (const gsl::not_null< LhsTensor * > lhs_tensor) const
	Assert that each instance of the LHS tensor in the RHS tensor expression uses the same generic index order that the LHS uses. More...
size_t	get_rhs_tensor_component_size () const
	Get the size of a component from a Tensor in this expression's subtree of the RHS TensorExpression More...
decltype(auto)	get (const std::array< size_t, num_tensor_indices > &result_multi_index) const
	Return the value of the component of the quotient tensor at a given multi-index. More...
template<typename ResultType >
decltype(auto)	get_primary (const ResultType &result_component, const std::array< size_t, num_tensor_indices > &result_multi_index) const
	Return the value of the component of the quotient tensor at a given multi-index. More...
template<typename ResultType >
void	evaluate_primary_children (ResultType &result_component, const std::array< size_t, num_tensor_indices > &result_multi_index) const
	Evaluate the LHS Tensor's result component at this subtree by evaluating the two operand's subtrees separately and dividing. More...
template<typename ResultType >
void	evaluate_primary_subtree (ResultType &result_component, const std::array< size_t, num_tensor_indices > &result_multi_index) const
	Successively evaluate the LHS Tensor's result component at each leg in this expression's subtree. More...

Static Public Attributes
static constexpr auto	num_tensor_indices
	The number of tensor indices in the result of the expression. More...
static constexpr auto	op2_num_tensor_indices
	The number of tensor indices in the left operand expression. More...
static constexpr auto	op2_multi_index
	The multi-index for the denominator. More...
static constexpr size_t	num_ops_left_child = T1::num_ops_subtree
	The number of arithmetic tensor operations done in the subtree for the left operand.
static constexpr size_t	num_ops_right_child = T2::num_ops_subtree
	The number of arithmetic tensor operations done in the subtree for the right operand.
static constexpr size_t	num_ops_subtree
	The total number of arithmetic tensor operations done in this expression's whole subtree. More...
static constexpr size_t	height_relative_to_closest_tensor_leaf_in_subtree
	The height of this expression's node in the expression tree relative to the closest TensorAsExpression leaf in its subtree. More...
static constexpr bool	is_primary_end = T1::is_primary_start
	If on the primary path, whether or not the expression is an ending point of a leg.
static constexpr size_t	num_ops_to_evaluate_primary_left_child
	If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done in the subtree of the child along the primary path, given that we will have already computed the whole subtree at the next lowest leg's starting point. More...
static constexpr size_t	num_ops_to_evaluate_primary_right_child
	If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done in the right operand's subtree. No splitting is currently done, so this is just num_ops_right_child. More...
static constexpr size_t	num_ops_to_evaluate_primary_subtree
	If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done for this expression's subtree, given that we will have already computed the subtree at the next lowest leg's starting point. More...
static constexpr bool	is_primary_start
	If on the primary path, whether or not the expression is a starting point of a leg. More...
static constexpr bool	evaluate_children_separately
	When evaluating along a primary path, whether each operand's subtrees should be evaluated separately. Since DataVector expression runtime scales poorly with increased number of operations, evaluating the two expression subtrees separately like this is beneficial when at least one of the subtrees contains a large number of operations. More...
static constexpr bool	primary_child_subtree_contains_primary_start
	If on the primary path, whether or not the expression's child along the primary path is a subtree that contains a starting point of a leg along the primary path. More...
static constexpr bool	primary_subtree_contains_primary_start
	If on the primary path, whether or not this subtree contains a starting point of a leg along the primary path. More...

Detailed Description

template<typename T1, typename T2, typename... Args2>
struct tenex::Divide< T1, T2, Args2 >

Defines the tensor expression representing the quotient of one tensor expression divided by another tensor expression that evaluates to a rank 0 tensor.

Details

For details on aliases and members defined in this class, as well as general TensorExpression terminology used in its members' documentation, see documentation for TensorExpression.

Template Parameters

T1	the numerator operand expression of the division expression
T2	the denominator operand expression of the division expression
Args2	the generic indices of the denominator expression

Member Function Documentation

assert_lhs_tensorindices_same_in_rhs()

template<typename T1 , typename T2 , typename... Args2>

template<typename LhsTensorIndices , typename LhsTensor >

void tenex::Divide< T1, T2, Args2 >::assert_lhs_tensorindices_same_in_rhs ( const gsl::not_null< LhsTensor * >  lhs_tensor ) const

inline

Assert that each instance of the LHS tensor in the RHS tensor expression uses the same generic index order that the LHS uses.

Template Parameters

LhsTensorIndices

the list of generic TensorIndexs of the LHS result Tensor being computed

Parameters

lhs_tensor

the LHS result Tensor being computed

evaluate_primary_children()

template<typename T1 , typename T2 , typename... Args2>

template<typename ResultType >

void tenex::Divide< T1, T2, Args2 >::evaluate_primary_children ( ResultType &  result_component, const std::array< size_t, num_tensor_indices > &  result_multi_index  ) const

inline

Evaluate the LHS Tensor's result component at this subtree by evaluating the two operand's subtrees separately and dividing.

Details

This function takes into account whether we have already computed part of the result component at a lower subtree. In recursively computing this quotient, the current result component will be substituted in for the most recent (highest) subtree below it that has already been evaluated.

The left and right operands' subtrees are evaluated successively with two separate assignments to the LHS result component. Since DataVector expression runtime scales poorly with increased number of operations, evaluating the two expression subtrees separately like this is beneficial when at least one of the subtrees contains a large number of operations. Instead of evaluating a larger expression with their combined total number of operations, we evaluate two smaller ones.

Parameters

result_component	the LHS tensor component to evaluate
result_multi_index	the multi-index of the component of the result tensor to evaluate

evaluate_primary_subtree()

template<typename T1 , typename T2 , typename... Args2>

template<typename ResultType >

void tenex::Divide< T1, T2, Args2 >::evaluate_primary_subtree ( ResultType &  result_component, const std::array< size_t, num_tensor_indices > &  result_multi_index  ) const

inline

Successively evaluate the LHS Tensor's result component at each leg in this expression's subtree.

Details

This function takes into account whether we have already computed part of the result component at a lower subtree. In recursively computing this quotient, the current result component will be substituted in for the most recent (highest) subtree below it that has already been evaluated.

Parameters

result_component	the LHS tensor component to evaluate
result_multi_index	the multi-index of the component of the result tensor to evaluate

get()

template<typename T1 , typename T2 , typename... Args2>

decltype(auto) tenex::Divide< T1, T2, Args2 >::get ( const std::array< size_t, num_tensor_indices > &  result_multi_index ) const

inline

Return the value of the component of the quotient tensor at a given multi-index.

Parameters

result_multi_index

the multi-index of the component of the quotient / tensor to retrieve

Returns: the value of the component in the quotient tensor at result_multi_index

get_primary()

template<typename T1 , typename T2 , typename... Args2>

template<typename ResultType >

decltype(auto) tenex::Divide< T1, T2, Args2 >::get_primary ( const ResultType &  result_component, const std::array< size_t, num_tensor_indices > &  result_multi_index  ) const

inline

Return the value of the component of the quotient tensor at a given multi-index.

Details

This function differs from get in that it takes into account whether we have already computed part of the result component at a lower subtree. In recursively computing this quotient, the current result component will be substituted in for the most recent (highest) subtree below it that has already been evaluated.

Parameters

result_component	the LHS tensor component to evaluate
result_multi_index	the multi-index of the component of the quotient / tensor to retrieve

Returns: the value of the component in the quotient tensor at result_multi_index

get_rhs_tensor_component_size()

template<typename T1 , typename T2 , typename... Args2>

size_t tenex::Divide< T1, T2, Args2 >::get_rhs_tensor_component_size ( ) const

inline

Get the size of a component from a Tensor in this expression's subtree of the RHS TensorExpression

Returns: the size of a component from a Tensor in this expression's subtree of the RHS TensorExpression

Member Data Documentation

evaluate_children_separately

template<typename T1 , typename T2 , typename... Args2>

constexpr bool tenex::Divide< T1, T2, Args2 >::evaluate_children_separately

staticconstexpr

Initial value:

=
      is_primary_start and (num_ops_to_evaluate_primary_left_child >=
                                detail::max_num_ops_in_sub_expression<type> or
                            num_ops_to_evaluate_primary_right_child >=
                                detail::max_num_ops_in_sub_expression<type>)

When evaluating along a primary path, whether each operand's subtrees should be evaluated separately. Since DataVector expression runtime scales poorly with increased number of operations, evaluating the two expression subtrees separately like this is beneficial when at least one of the subtrees contains a large number of operations.

height_relative_to_closest_tensor_leaf_in_subtree

template<typename T1 , typename T2 , typename... Args2>

constexpr size_t tenex::Divide< T1, T2, Args2 >::height_relative_to_closest_tensor_leaf_in_subtree

staticconstexpr

Initial value:

=
      T2::height_relative_to_closest_tensor_leaf_in_subtree <=
              T1::height_relative_to_closest_tensor_leaf_in_subtree
          ? (T2::height_relative_to_closest_tensor_leaf_in_subtree !=
                     std::numeric_limits<size_t>::max()
                 ? T2::height_relative_to_closest_tensor_leaf_in_subtree + 1
                 : T2::height_relative_to_closest_tensor_leaf_in_subtree)
          : T1::height_relative_to_closest_tensor_leaf_in_subtree + 1

The height of this expression's node in the expression tree relative to the closest TensorAsExpression leaf in its subtree.

is_primary_start

template<typename T1 , typename T2 , typename... Args2>

constexpr bool tenex::Divide< T1, T2, Args2 >::is_primary_start

staticconstexpr

Initial value:

=
      num_ops_to_evaluate_primary_subtree >=
      detail::max_num_ops_in_sub_expression<type>

If on the primary path, whether or not the expression is a starting point of a leg.

num_ops_subtree

template<typename T1 , typename T2 , typename... Args2>

constexpr size_t tenex::Divide< T1, T2, Args2 >::num_ops_subtree

staticconstexpr

Initial value:

=
      num_ops_left_child + num_ops_right_child + 1

The total number of arithmetic tensor operations done in this expression's whole subtree.

num_ops_to_evaluate_primary_left_child

template<typename T1 , typename T2 , typename... Args2>

constexpr size_t tenex::Divide< T1, T2, Args2 >::num_ops_to_evaluate_primary_left_child

staticconstexpr

Initial value:

=
      is_primary_end ? 0 : T1::num_ops_to_evaluate_primary_subtree

If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done in the subtree of the child along the primary path, given that we will have already computed the whole subtree at the next lowest leg's starting point.

num_ops_to_evaluate_primary_right_child

template<typename T1 , typename T2 , typename... Args2>

constexpr size_t tenex::Divide< T1, T2, Args2 >::num_ops_to_evaluate_primary_right_child

staticconstexpr

Initial value:

=
      num_ops_right_child

If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done in the right operand's subtree. No splitting is currently done, so this is just num_ops_right_child.

num_ops_to_evaluate_primary_subtree

template<typename T1 , typename T2 , typename... Args2>

constexpr size_t tenex::Divide< T1, T2, Args2 >::num_ops_to_evaluate_primary_subtree

staticconstexpr

Initial value:

=
      num_ops_to_evaluate_primary_left_child +
      num_ops_to_evaluate_primary_right_child + 1

If on the primary path, this is the remaining number of arithmetic tensor operations that need to be done for this expression's subtree, given that we will have already computed the subtree at the next lowest leg's starting point.

num_tensor_indices

template<typename T1 , typename T2 , typename... Args2>

constexpr auto tenex::Divide< T1, T2, Args2 >::num_tensor_indices

staticconstexpr

Initial value:

=
      tmpl::size<typename T1::index_list>::value

The number of tensor indices in the result of the expression.

op2_multi_index

template<typename T1 , typename T2 , typename... Args2>

constexpr auto tenex::Divide< T1, T2, Args2 >::op2_multi_index

staticconstexpr

Initial value:

=
      make_array<op2_num_tensor_indices, size_t>(0)

The multi-index for the denominator.

op2_num_tensor_indices

template<typename T1 , typename T2 , typename... Args2>

constexpr auto tenex::Divide< T1, T2, Args2 >::op2_num_tensor_indices

staticconstexpr

Initial value:

=
      tmpl::size<typename T2::index_list>::value

The number of tensor indices in the left operand expression.

primary_child_subtree_contains_primary_start

template<typename T1 , typename T2 , typename... Args2>

constexpr bool tenex::Divide< T1, T2, Args2 >::primary_child_subtree_contains_primary_start

staticconstexpr

Initial value:

=
      T1::primary_subtree_contains_primary_start

If on the primary path, whether or not the expression's child along the primary path is a subtree that contains a starting point of a leg along the primary path.

primary_subtree_contains_primary_start

template<typename T1 , typename T2 , typename... Args2>

constexpr bool tenex::Divide< T1, T2, Args2 >::primary_subtree_contains_primary_start

staticconstexpr

Initial value:

=
      is_primary_start or primary_child_subtree_contains_primary_start

If on the primary path, whether or not this subtree contains a starting point of a leg along the primary path.

---

The documentation for this struct was generated from the following file:

• src/DataStructures/Tensor/Expressions/Divide.hpp