Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN)

Intel(R) Math Kernel Library for Deep Neural Networks (Intel(R) MKL-DNN) is an open source performance library for Deep Learning (DL) applications intended for acceleration of DL frameworks on Intel(R) architecture. Intel(R) MKL-DNN includes highly vectorized and threaded building blocks for implementation of convolutional neural networks (CNN) with C and C++ interfaces. We created this project to help DL community innovate on Intel(R) processors.

Intel MKL-DNN functionality shares implementation with Intel(R) Math Kernel Library (Intel(R) MKL), but is not API compatible with Intel MKL 2017. We will be looking into ways to converge API in future releases of Intel MKL.

This release is a technical preview with functionality limited to AlexNet and VGG topologies forward path. While this library is in technical preview phase, its API may change without considerations of backward compatibility.

License

Intel MKL-DNN is licensed under Apache License Version 2.0.

Documentation

You can find the latest Intel MKL-DNN documentation at GitHub pages.

Support

Please report issues and suggestions via GitHub issues or start a topic on Intel MKL forum.

How to Contribute

We welcome community contributions to Intel MKL-DNN. If you have an idea how to improve the product:

• Let us know about your proposal via GitHub issues.

• Make sure you can build the product and run all the examples with your patch

• In the case of a larger feature, create a test

• Submit a pull request

We will review your contribution and, if any additional fixes or modifications are necessary, may give some feedback to guide you. When accepted, your pull request will be merged into our internal and GitHub repositories.

System Requirements

Intel MKL-DNN supports Intel(R) 64 architecture processors and is optimized for

• Intel(R) Xeon(R) processor E5-xxxx v3 (codename Haswell)
• Intel(R) Xeon(R) processor E5-xxxx v4 (codename Broadwell)

Processors without Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) are not supported in this release.

Software dependencies:

• Cmake 2.8.0 or later
• Doxygen 1.8.5 or later
• C++ compiler with C++11 standard support

The software was validated on RedHat* Enterprise Linux 7 with

• GNU* Compiler Collection 4.8
• GNU* Compiler Collection 6.1
• Clang* 3.8.0
• Intel(R) C/C++ Compiler 16.0 or later

The implementation relies on OpenMP* SIMD extensions, and we recommend using Intel(R) compiler for the best performance results.

Installation

Download Intel MKL-DNN source code or clone the repository to your system

	git clone https://github.com/01org/mkl-dnn.git

Before the installation, make sure that all the dependencies are available and have correct versions. Intel MKL-DNN uses optimized matrix-matrix multiplication (GEMM) routine from Intel MKL. Dynamic library with this functionality is included with the Intel MKL-DNN release. Before building the project download the library using provided script

	cd scripts && ./prepare_mkl.sh && cd ..

or download manually and unpack to external directory in the repository root.

Intel MKL-DNN uses CMake-based build system

	mkdir -p build && cd build && cmake .. && make

Intel MKL-DNN includes unit tests implemented using the googletest framework. To validate the build, run:

	make test

Documentation is provided inline and can be generated in HTML format with Doxygen:

	make doc

Documentation will be created in build/reference/html folder.

Finally,

	make install

will put header files, libraries and documentation to /usr/local. To change installation path use the option -DCMAKE_INSTALL_PREFIX=<prefix> when invoking CMake.