types

group Type system

Objects for the specification and management of types.

Enums

enum class ReductionOpKind : std::int32_t

Enum for reduction operator kinds.

Values:

enumerator ADD

Addition

enumerator MUL

Multiplication

enumerator MAX

Binary maximum operator

enumerator MIN

Binary minimum operator

enumerator OR

Bitwise OR

enumerator AND

Bitwse AND

enumerator XOR

Bitwas XOR

Functions

Type primitive_type(Type::Code code)

Creates a metadata object for a primitive type.

Parameters:

code – Type code

Returns:

Type object

Type string_type()

Creates a metadata object for the string type.

Returns:

Type object

Type binary_type(std::uint32_t size)

Creates an opaque binary type of a given size.

Parameters:

size – Element size

Returns:

Type object

FixedArrayType fixed_array_type(

const Type &element_type,

std::uint32_t N

)

Creates a metadata object for a fixed-size array type.

Parameters:

• element_type – Type of the array elements

• N – Size of the array

Returns:

FixedArrayType object

StructType struct_type(

const std::vector<Type> &field_types,

bool align = true

)

Creates a metadata object for a struct type.

Parameters:

• field_types – A vector of field types

• align – If true, fields in the struct are aligned

Returns:

StructType object

ListType list_type(const Type &element_type)

Creates a metadata object for a list type.

Parameters:

element_type – Type of the list elements

Returns:

ListType object

template<typename ...Args>
std::enable_if_t<std::conjunction_v<std::is_convertible<std::decay_t<Args>, Type>...>, StructType> struct_type(

bool align,

Args&&... field_types

)

Creates a metadata object for a struct type.

Parameters:

• align – If true, fields in the struct are aligned

• field_types – Field types

Returns:

StructType object

std::ostream &operator<<(

std::ostream &ostream,

const Type::Code &code

)

std::ostream &operator<<(std::ostream &ostream, const Type &type)

Type bool_()

Creates a boolean type.

Returns:

Type object

Type int8()

Creates a 8-bit signed integer type.

Returns:

Type object

Type int16()

Creates a 16-bit signed integer type.

Returns:

Type object

Type int32()

Creates a 32-bit signed integer type.

Returns:

Type object

Type int64()

Creates a 64-bit signed integer type.

Returns:

Type object

Type uint8()

Creates a 8-bit unsigned integer type.

Returns:

Type object

Type uint16()

Creates a 16-bit unsigned integer type.

Returns:

Type object

Type uint32()

Creates a 32-bit unsigned integer type.

Returns:

Type object

Type uint64()

Creates a 64-bit unsigned integer type.

Returns:

Type object

Type float16()

Creates a half-precision floating point type.

Returns:

Type object

Type float32()

Creates a single-precision floating point type.

Returns:

Type object

Type float64()

Creates a double-precision floating point type.

Returns:

Type object

Type complex64()

Creates a single-precision complex number type.

Returns:

Type object

Type complex128()

Creates a double-precision complex number type.

Returns:

Type object

FixedArrayType point_type(std::uint32_t ndim)

Creates a point type.

Parameters:

ndim – Number of dimensions

Returns:

FixedArrayType object

StructType rect_type(std::uint32_t ndim)

Creates a rect type.

Parameters:

ndim – Number of dimensions

Returns:

StructType object

Type null_type()

Creates a null type.

Returns:

Type object

bool is_point_type(const Type &type)

Checks if the type is a point type.

Parameters:

type – Type to check

Returns:

true If the type is a point type

Returns:

false Otherwise

bool is_point_type(const Type &type, std::uint32_t ndim)

Checks if the type is a point type of the given dimensionality.

Parameters:

• type – Type to check

• ndim – Number of dimensions the point type should have

Returns:

true If the type is a point type

Returns:

false Otherwise

std::int32_t ndim_point_type(const Type &type)

Returns the number of dimensions of a given point type.

Parameters:

type – Point type

Throws:

std::invalid_argument – IF the type is not a point type

Returns:

Number of dimensions

bool is_rect_type(const Type &type)

Checks if the type is a rect type.

Parameters:

type – Type to check

Returns:

true If the type is a rect type

Returns:

false Otherwise

bool is_rect_type(const Type &type, std::uint32_t ndim)

Checks if the type is a rect type of the given dimensionality.

Parameters:

• type – Type to check

• ndim – Number of dimensions the rect type should have

Returns:

true If the type is a rect type

Returns:

false Otherwise

std::int32_t ndim_rect_type(const Type &type)

Returns the number of dimensions of a given rect type.

Parameters:

type – Rect type

Throws:

std::invalid_argument – IF the type is not a rect type

Returns:

Number of dimensions

class Type

#include <legate/type/types.h>

A base class for data type metadata.

Subclassed by legate::FixedArrayType, legate::ListType, legate::StructType

Public Types

enum class Code : std::int32_t

Enum for type codes.

Values:

enumerator BOOL

Boolean type

enumerator INT8

8-bit signed integer type

enumerator INT16

16-bit signed integer type

enumerator INT32

32-bit signed integer type

enumerator INT64

64-bit signed integer type

enumerator UINT8

8-bit unsigned integer type

enumerator UINT16

16-bit unsigned integer type

enumerator UINT32

32-bit unsigned integer type

enumerator UINT64

64-bit unsigned integer type

enumerator FLOAT16

Half-precision floating point type

enumerator FLOAT32

Single-precision floating point type

enumerator FLOAT64

Double-precision floating point type

enumerator COMPLEX64

Single-precision complex type

enumerator COMPLEX128

Double-precision complex type

enumerator NIL

Null type

enumerator BINARY

Opaque binary type

enumerator FIXED_ARRAY

Fixed-size array type

enumerator STRUCT

Struct type

enumerator STRING

String type

enumerator LIST

List type

Public Functions

Code code() const

Code of the type.

Returns:

Type code

std::uint32_t size() const

Size of the data type in bytes.

Returns:

Data type size in bytes

std::uint32_t alignment() const

Alignment of the type.

Returns:

Alignment in bytes

std::uint32_t uid() const

Unique ID of the data type.

Returns:

Unique ID

bool variable_size() const

Indicates whether the data type is of variable size elements.

Returns:

true Elements can be variable size

Returns:

false Elements have fixed size

std::string to_string() const

Converts the data type into a string.

Returns:

A string of the data type

bool is_primitive() const

Indicates whether the type is a primitive type.

Returns:

true If the type is a primitive type

Returns:

false Otherwise

FixedArrayType as_fixed_array_type() const

Dynamically casts the type into a fixed size array type.

If the type is not a fixed size array type, an exception will be raised.

Returns:

Type object

StructType as_struct_type() const

Dynamically casts the type into a struct type.

If the type is not a struct type, an exception will be raised.

Returns:

Type object

ListType as_list_type() const

Dynamically casts the type into a struct type.

If the type is not a struct type, an exception will be raised.

Returns:

Type object

void record_reduction_operator(

std::int32_t op_kind,

GlobalRedopID global_op_id

) const

Records a reduction operator.

The global ID of the reduction operator is issued when that operator is registered to the runtime.

Parameters:

• op_kind – Reduction operator kind

• global_op_id – Global reduction operator ID

void record_reduction_operator(

ReductionOpKind op_kind,

GlobalRedopID global_op_id

) const

Records a reduction operator.

The global ID of the reduction operator is issued when that operator is registered to the runtime.

Parameters:

• op_kind – Reduction operator kind

• global_op_id – Global reduction operator ID

GlobalRedopID find_reduction_operator(std::int32_t op_kind) const

Finds the global operator ID for a given reduction operator kind.

Raises an exception if no reduction operator has been registered for the kind.

Parameters:

op_kind – Reduction operator kind

Returns:

Global reduction operator ID

GlobalRedopID find_reduction_operator(ReductionOpKind op_kind) const

Finds the global operator ID for a given reduction operator kind.

Raises an exception if no reduction operator has been registered for the kind.

Parameters:

op_kind – Reduction operator kind

Returns:

Global reduction operator ID

bool operator==(const Type &other) const

Equality check between types.

Note that type checks are name-based; two isomorphic fixed-size array types are considered different if their uids are different (the same applies to struct types).

Parameters:

other – Type to compare

Returns:

true Types are equal

Returns:

false Types are different

class FixedArrayType : public legate::Type

#include <legate/type/types.h>

A class for fixed-size array data types.

Public Functions

std::uint32_t num_elements() const

Returns the number of elements.

Returns:

Number of elements

Type element_type() const

Returns the element type.

Returns:

Element type

class StructType : public legate::Type

#include <legate/type/types.h>

A class for struct data types.

Public Functions

std::uint32_t num_fields() const

Returns the number of fields.

Returns:

Number of fields

Type field_type(std::uint32_t field_idx) const

Returns the element type.

Parameters:

field_idx – Field index. Must be within the range

Returns:

Element type

bool aligned() const

Indicates whether the fields are aligned.

Returns:

true Fields are aligned

Returns:

false Fields are compact

std::vector<std::uint32_t> offsets() const

Returns offsets to fields.

Returns:

Field offsets in a vector

class ListType : public legate::Type

#include <legate/type/types.h>

A class for list types.

Public Functions

Type element_type() const

Returns the element type.

Returns:

Element type