Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN)  0.95.0
Performance library for Deep Learning
 All Classes Files Functions Variables Typedefs Enumerations Enumerator Groups Pages
Intel(R) MKL-DNN Documentation

The Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN) is an open-source C/C++ performance library for Deep Learning (DL) applications primarily intended for acceleration of DL frameworks on Intel(R) architecture and Intel(R) Processor Graphics Architecture. Intel MKL-DNN includes highly optimized implementations of computational operations used in convolutional and recurrent neural networks covering a wide range of applications, including image recognition, object detection, semantic segmentation, neural machine translation, and speech recognition.

Building and Linking

Programming Model

Primitives

Compute intensive operations:

Memory bandwidth limited operations:

Data manipulation:

Performance Benchmarking and Inspection

Advanced topics

Examples

Scenario Platform C++ API C API
Introduction CPU Getting started
GPU Getting started on GPU Getting started on GPU
fp32 inference CPU CNN fp32 inference example CNN fp32 inference example
RNN fp32 inference example
int8 inference CPU CNN int8 inference example
RNN int8 inference example
training CPU CNN fp32 training example CNN fp32 training example
CPU RNN fp32 training example

Legal information