BroadCasted

namespace ndarray
namespace broadcast

Functions

ndarray::Shape getBroadCastedShape(const ndarray::Shape &l_shape, const ndarray::Shape &r_shape, const int offset = 0)

make expected broadcasted shape of two given array

Example: l_shape {4,1}, r_shape{1,3} => out_shape{4,3}

Return

Operation output shape.

Parameters

  • l_shape: Left array shape.

  • r_shape: Right array shape.

  • offset: int value

BroadCastedProperty getBroadCastedProperty(const ndarray::Shape &out_shape, const ndarray::Array &A, const ndarray::Array &B, const int offset = 0)

Calculate broadcasted operated array shape and stride. Which is needed for calculate actual index of operation data.

Return

BroadCastedProperty DTO class which have a_shape, b_shape, a_stride, b_stride.

Parameters

  • out_shape: broadcasted output shape.

  • A: ndarray.

  • B: ndarray.

std::vector<int> paddedVector(const std::vector<int> &vec, const size_t size, const int pad_value = 0)

make a vector fill with pad value

Example: vec{3,2,4}, size = 5, pad_value = 100 =>out_vec{100,100,3,2,4}

Return

vector with padded value

Parameters

  • vec: int type vector.

  • size: expected vector size.

  • pad_value: expected value to fill. default: 0

struct BroadCastedProperty
#include <broadCasted.h>

Public Functions

BroadCastedProperty(const ndarray::Shape &a_shape_, const ndarray::Shape &b_shape_, const ndarray::Stride &a_stride_, const ndarray::Stride &b_stride_)

Public Members

ndarray::Shape a_shape
ndarray::Shape b_shape
ndarray::Stride a_stride
ndarray::Stride b_stride