oneAPI Deep Neural Network Library (oneDNN)
Performance library for Deep Learning

Compute primitives. More...


 Common operations to create, destroy and inspect primitives.
 A container for parameters that extend primitives behavior.
 A primitive to copy data between two memory objects.
 A primitive to concatenate data by arbitrary dimension.
 A primitive to sum multiple tensors.
 A primitive to perform tensor operations over two tensors.
 A primitive to perform 1D, 2D or 3D convolution.
 A primitive to perform 1D, 2D or 3D deconvolution.
 A primitive to shuffle tensor data along an axis.
 A primitive to perform elementwise operations such as the rectifier linear unit (ReLU).
 A primitive to perform softmax.
 A primitive to perform logsoftmax.
 A primitive to perform max or average pooling.
 PReLU primitive A primitive to perform PReLU (leaky ReLU with trainable alpha parameter)
 A primitive to perform local response normalization (LRN) across or within channels.
 Batch Normalization
 A primitive to perform batch normalization.
 Layer Normalization
 A primitive to perform layer normalization.
 Inner Product
 A primitive to compute an inner product.
 A primitive to compute recurrent neural network layers.
 Matrix Multiplication
 A primitive to perform matrix-matrix multiplication.
 A primitive to compute resampling operation on 1D, 2D or 3D data tensor using Nearest Neighbor, or Linear (Bilinear, Trilinear) interpolation method.
 A primitive to compute reduction operation on data tensor using min, max, mul, sum, mean and norm_lp operations.

Detailed Description

Compute primitives.

See also
Basic Concepts