NDArray

Lazily evaluated array. An array is a stateful (partial) closure consisting of three basic components:

1. a formal argument list

2. an expression

3. an environment

1) and 3) are stored in a single capture list, where the value None represents items in the argument list. Arguments can be instantiated and moved into its environment. This process is monotonic, that is, the environment cannot be modified by operations other than instantiation of the arguments.

lappy.ndarray.ndim

number of dimensions specified as an Unsigned. At the moment, only concrete ndim is supported, i.e., ndim.value must be known.

lappy.ndarray.shape

overall shape of array expression. Tuple of str with length equal to ndim. Each component of the tuple is the name of a lazy object in the namespace.

lappy.ndarray.inames

index names for iterating over the array. Tuple of pymbolic.Variable’s.

lappy.ndarray.domain_expr

set expression for constructing the loop domain. A set expression has leafs of lappy.core.primitives.PwAff variables and integers. The domain expression is trivial if it equals to the set of indices of the array elements (no “hidden” axes).

lappy.ndarray.dtype

data type, either None or convertible to loopy type. Lappy does not allow implicit type casting.

lappy.ndarray.is_integral

a bool indicating whether array elements are integers

lappy.ndarray.integer_domain

an ISL set if is_integral is True, otherwise None. The set represents the domain for the full array. It is used to express conditions like non-negative integers, multiples of two, etc.