gko::dim<N>

The extent of an N-dimensional tensor is represented by a gko::dim<N> object. In nearly all instances N is exclusively set to N=2, which corresponds to elements from \(\mathbb{R}^{n\times m}\).

A gko::dim<N> object is created by passing in the extents for each dimensions. For N=2 this corresponds to passing in the number of rows and number of columns of a matrix.

gko::size_type n1, n2, n3, n4, num_rows, num_cols;
gko::dim<4> dim4(n1, n2, n3, n4);
gko::dim<2> size(num_rows, num_cols);

The gko::dim<N> provides access to the extents through the [] operator. The extent of the i-th dimension (starting with zero) is given by size[i].

gko::dim<2> size(num_rows, num_cols);
auto extent = size[1];
assert(extent == num_cols);

A gko::dim<N> object can be converted to bool. It evaluates to true if and only if all extents are larger than zero. As soon as any extent is zero, it will evaluate to false. This allows using gko::dim<N> in if expression, perhaps to catch edge cases.

gko::dim<2> size(0, 0);
if(!size){
  // handle empty size case
}

The gko::dim<N> class also supports equality checking of two objects. They are the same if and only if all their extents are the same.