numba_dpex.core.boxing

Contains the box and unbox functions for numba_dpex types that are passable as arguments to a kernel or dpjit decorated function.

Overview

Classes

NdRangeType

Numba-dpex type corresponding to

RangeType

Numba-dpex type corresponding to

NdRange

Analogue to the sycl::nd_range class.

Range

Analogue to the sycl::range class.

USMNdArray

A type class to represent dpctl.tensor.usm_ndarray.

NdRange

Analogue to the sycl::nd_range class.

Range

Analogue to the sycl::range class.
Function

unbox_range(typ, obj, c)

Converts a Python Range object to numba-dpex’s native struct representation

unbox_ndrange(typ, obj, c)

Converts a Python Range object to numba-dpex’s native struct representation

box_range(typ, val, c)

Convert a native range structure to a Range object.

box_ndrange(typ, val, c)

Convert a native range structure to a Range object.

unbox_dpnp_nd_array(typ, obj, c)

Converts a dpctl.tensor.usm_ndarray/dpnp.ndarray object to a Numba-dpex

box_array(typ, val, c)

Boxes a NativeValue representation of USMNdArray/DpnpNdArray type into a

Classes

class NdRangeType(ndim: int)

Bases: numba.core.types.Type

Numba-dpex type corresponding to numba_dpex.kernel_api.ranges.NdRange

class RangeType(ndim: int)

Bases: numba.core.types.Type

Numba-dpex type corresponding to numba_dpex.kernel_api.ranges.Range

class NdRange(global_size, local_size)

Analogue to the sycl::nd_range class.

The NdRange defines the index space for a work group as well as the global index space. It is passed to parallel_for to execute a kernel on a set of work items.

This class basically contains two Range object, one for the global_range and the other for the local_range. The global_range parameter contains the global index space and the local_range parameter contains the index space of a work group. This class mimics the behavior of sycl::nd_range class.

Overview

Methods

get_global_range()

Returns a Range defining the index space.

get_local_range()

Returns a Range defining the index space of a work group.

Members

get_global_range()

Returns a Range defining the index space.

Returns:

A Range object defining the index space.

Return type:

Range

get_local_range()

Returns a Range defining the index space of a work group.

Returns:

A Range object to specify index space of a work group.

Return type:

Range

class Range

Bases: tuple

Analogue to the sycl::range class.

The range is an abstraction that describes the number of elements in each dimension of buffers and index spaces. It can contain 1, 2, or 3 numbers, depending on the dimensionality of the object it describes.

This is just a wrapper class on top of a 3-tuple. The kernel launch parameter is consisted of three int’s. This class basically mimics the behavior of sycl::range.

Overview

Methods

get(index)

Returns the range of a single dimension.

size()

Returns the size of a range.

Members

get(index)

Returns the range of a single dimension.

Parameters:

index (int) – The index of the dimension, i.e. [0,2]

Returns:

The range of the dimension indexed by index.

Return type:

int

size()

Returns the size of a range.

Returns the size of a range by multiplying the range of the individual dimensions.

Returns:

The size of a range.

Return type:

int

class USMNdArray(ndim, layout='C', dtype=None, usm_type='device', device=None, queue=None, readonly=False, name=None, aligned=True, addrspace=address_space.GLOBAL.value)

Bases: numba.core.types.npytypes.Array

A type class to represent dpctl.tensor.usm_ndarray.

Overview

Methods

copy(dtype, ndim, layout, readonly, addrspace, device, usm_type)

-

unify(typingctx, other)

Unify this with the other USMNdArray.

can_convert_to(typingctx, other)

Convert this USMNdArray to the other.

Members

copy(dtype=None, ndim=None, layout=None, readonly=None, addrspace=None, device=None, usm_type=None)
unify(typingctx, other)

Unify this with the other USMNdArray.

can_convert_to(typingctx, other)

Convert this USMNdArray to the other.

class NdRange(global_size, local_size)

Analogue to the sycl::nd_range class.

The NdRange defines the index space for a work group as well as the global index space. It is passed to parallel_for to execute a kernel on a set of work items.

This class basically contains two Range object, one for the global_range and the other for the local_range. The global_range parameter contains the global index space and the local_range parameter contains the index space of a work group. This class mimics the behavior of sycl::nd_range class.

Overview

Methods

get_global_range()

Returns a Range defining the index space.

get_local_range()

Returns a Range defining the index space of a work group.

Members

get_global_range()

Returns a Range defining the index space.

Returns:

A Range object defining the index space.

Return type:

Range

get_local_range()

Returns a Range defining the index space of a work group.

Returns:

A Range object to specify index space of a work group.

Return type:

Range

class Range

Bases: tuple

Analogue to the sycl::range class.

The range is an abstraction that describes the number of elements in each dimension of buffers and index spaces. It can contain 1, 2, or 3 numbers, depending on the dimensionality of the object it describes.

This is just a wrapper class on top of a 3-tuple. The kernel launch parameter is consisted of three int’s. This class basically mimics the behavior of sycl::range.

Overview

Methods

get(index)

Returns the range of a single dimension.

size()

Returns the size of a range.

Members

get(index)

Returns the range of a single dimension.

Parameters:

index (int) – The index of the dimension, i.e. [0,2]

Returns:

The range of the dimension indexed by index.

Return type:

int

size()

Returns the size of a range.

Returns the size of a range by multiplying the range of the individual dimensions.

Returns:

The size of a range.

Return type:

int

Functions

unbox_range(typ, obj, c)

Converts a Python Range object to numba-dpex’s native struct representation for RangeType.

unbox_ndrange(typ, obj, c)

Converts a Python Range object to numba-dpex’s native struct representation for NdRangeType.

box_range(typ, val, c)

Convert a native range structure to a Range object.

box_ndrange(typ, val, c)

Convert a native range structure to a Range object.

unbox_dpnp_nd_array(typ, obj, c)

Converts a dpctl.tensor.usm_ndarray/dpnp.ndarray object to a Numba-dpex internal array structure.

Parameters:

  • typ – The Numba type of the PyObject

  • obj – The actual PyObject to be unboxed

  • c – The unboxing context

Returns: A NativeValue object representing an unboxed

dpctl.tensor.usm_ndarray/dpnp.ndarray

box_array(typ, val, c)

Boxes a NativeValue representation of USMNdArray/DpnpNdArray type into a dpctl.tensor.usm_ndarray/dpnp.ndarray PyObject

Parameters:

  • typ – The representation of the USMNdArray/DpnpNdArray type.

  • val – A native representation of a Numba USMNdArray/DpnpNdArray type object.

  • c – The boxing context.

Returns: A Pyobject for a dpctl.tensor.usm_ndarray/dpnp.ndarray boxed from

the Numba-dpex native value.