syten::DenseIterDual< rank > Struct Template Reference

Iterator over a dense tensor-like object with a secondary index running. More...

#include <dense_iter.h>

Collaboration diagram for syten::DenseIterDual< rank >:

Public Member Functions
template<typename OrdType >
	DenseIterDual (std::array< Index, rank > const &dim_, std::array< OrdType, rank > const &ord_)
	Ctor for nontrivial ordering. More...
Size	dual_offset () const
	Returns the dual offset, i.e. More...
Size	offset () const
	Returns the offset from the start, mostly only valid if the ordering is trivial (but there are cases where it makes sense for non-trivial orderings as well) More...
	operator bool () const
	False if we reached the last element. More...
void	operator++ ()
	Increase the indices in the order specified by ord. More...
DenseIterDual	operator= (DenseIterDual &&)=delete
	Move assignment op deleted, would require pointer fiddling. More...
DenseIterDual	operator= (DenseIterDual const &)=delete
	Copy assignment op deleted, would require pointer fiddling. More...

Protected Attributes
bool	alive
	True unless we have reached the last element. More...
std::array< Index, rank >	cur
	Current value of the array, cur[i] is the value of the i-th increased index. More...
std::array< Index, rank >	dim
	Dimensions of individual indices, dim[i] is the dimension of the i-th increased index. More...
Size	dual_running
	Dual running index. More...
std::array< Index, rank >	factors
	Array of factors. More...
Size	running
	Current primary offset. More...

Detailed Description

template<Rank rank>
struct syten::DenseIterDual< rank >

Iterator over a dense tensor-like object with a secondary index running.

Given a set of dimensions \( D = \left(d_1, d_2, \ldots, d_n \right) \) and a set of orderings \( O = \left(o_1, o_2, \ldots, o_n \right) \) on each call to operator++, the \( o_n \)-th element of \( r \) is increased, or, if this results in \( d_{o_n} \), set to zero and the \( o_{n-1} \) element is increases instead (and similarly if that is then equal to \( d_{o_{n-1}} \)).

One running index, obtainable via offset(), can be used to simply count the number of steps taken. The secondary index obtainable via dual_offset() will be incremented to be the index into a well-ordered dimensions array, which is useful for tensor permutations.

This iterator does not provide a properly-ordered coordinate array.

Remarks

When using this iterator to transpose tensors, I don't know whether it is better to read or write in sequential order.

---

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

• inc/dense/dense_iter.h