heterocl.nparray

Helper functions for NumPy arrays.

asarray(arr, dtype=None, ctx=cpu(0))[source]

Convert a NumPy array to a HeteroCL array.

The converted array will be inputs to the executable.

Parameters

  • arr (ndarray) – The array to be converted

  • dtype (Type, optional) – The target data type

  • ctx (TVMContext) – The target context

Returns

Return type

NDArray

Examples

np_A = numpy.zeros(10)
hcl_A = np_A.asarray()
cast_np(np_in, dtype)[source]

Cast a NumPy array to a specified data type.

Parameters

  • np_in (ndarray) – The array to be cast

  • dtype (Type) – The target data type

Returns

Return type

ndarray

Examples

A_float = numpy.random.rand(10)
A_fixed = hcl.cast_np(A_float, hcl.Fixed(10, 8))
pack_np(np_in, dtype_in, dtype_out)[source]

Pack a NumPy array according to the specified data types.

Now we only support packing and unpacking for a 1-dimensional array.

Parameters

  • np_in (ndarray) – The array to be packed

  • dtype_in (Type) – The data type of the input array

  • dtype_out (Type) – The target data type

Returns

Return type

ndarray

Examples

a = numpy.random.randint(16, size=(10,))
packed_a = hcl.pack_np(np_in, hcl.UInt(8), hcl.UInt(32))
unpack_np(np_in, dtype_in, dtype_out)[source]

Unpack a NumPy array according to the specified data types.

Now we only support packing and unpacking for a 1-dimensional array.

Parameters

  • np_in (ndarray) – The array to be unpacked

  • dtype_in (Type) – The data type of the input array

  • dtype_out (Type) – The target data type

Returns

Return type

ndarray

Examples

a = numpy.random.randint(4, size=(10,))
unpacked_a = hcl.pack_np(np_in, hcl.UInt(32), hcl.UInt(8))