Global Functions

index_iterator make_iterator(const Lengths &len, const Strides&... strides)

Create an index_iterator from the given set of lengths and zero or more sets of strides.

std::vector<int> lengths{4, 5, 2};
std::vector<int> row_strides{10, 2, 1};
std::vector<int> col_strides{1, 4, 20};

auto it = make_iterator(lengths, row_strides, col_strides);
auto row_off = 0, col_off = 0;

while (it.next(row_off, col_off))
{
    ...
}

Parameters:

len – The lengths of each dimension. Can be any container type whose values are convertible to len_type.
strides – A parameter pack of strides. There must be exactly N parameters in the pack. Can be any container type whose values are convertible to stride_type. The size of each set of strides must be the same as the size of len.

Returns:

An index_iterator.

template<typename Type, int NDim, typename Allocator> void swap(marray<Type, NDim, Allocator> &a, marray<Type, NDim, Allocator> &b)

Swap the shape, data, and layout of two tensors.

Parameters:

a – The first tensor to swap.
b – The second tensor to swap.

template<int N> immutable_view cview(tensor_or_view x)

Return an immutable view of the given tensor.

Parameters:: x – The tensor to view.
Template Parameters:: N – The number of dimensions in the resulting view or [DYNAMIC](MArray::DYNAMIC). The default is the same number of dimensions in the tensor (if fixed), or DYNAMIC (if not).
Returns:: An immutable view.

template<int N> possibly_mutable_view view(tensor_or_view x)

Return a view of the given tensor.

Parameters:: x – The tensor to view.
Template Parameters:: N – The number of dimensions in the resulting view or [DYNAMIC](MArray::DYNAMIC). The default is the same number of dimensions in the tensor (if fixed), or DYNAMIC (if not).
Returns:: A possibly-mutable tensor view. For a tensor ([marray](MArray::marray)), the returned view is immutable if the instance is const-qualified. For a tensor view ([marray_view](MArray::marray_view)), the returned view is mutable if the value type is not const-qualified.

std::ostream &operator<<(std::ostream &os, tensor_or_view x)

Write out a representation of the tensor or tensor view to the given output stream.

The format resembles the following (i.e., a valid C/C++ nested array initializer):

{
 {0, 1, 2},
 {3, 4, 5},
 {6, 7, 9}
}

Parameters:

os – The output stream to write to.
x – The tensor or tensor view to write.

Returns:

The output stream os.

template<typename Type, int NDim> void swap(marray_view<Type, NDim> &a, marray_view<Type, NDim> &b)

Swap two views.

Parameters:

a – The first view to swap.
b – The second view to swap.