vsergeev / volk

The Vector Optimized Library of Kernels

Home Page:http://libvolk.org

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Build Status Build status Check PR Formatting Run VOLK tests

VOLK Logo

Welcome to VOLK!

VOLK is a sub-project of GNU Radio. Please see http://libvolk.org for bug tracking, documentation, source code, and contact information about VOLK. See https://www.gnuradio.org/ for information about GNU Radio.

VOLK is the Vector-Optimized Library of Kernels. It is a library that contains kernels of hand-written SIMD code for different mathematical operations. Since each SIMD architecture can be very different and no compiler has yet come along to handle vectorization properly or highly efficiently, VOLK approaches the problem differently.

For each architecture or platform that a developer wishes to vectorize for, a new proto-kernel is added to VOLK. At runtime, VOLK will select the correct proto-kernel. In this way, the users of VOLK call a kernel for performing the operation that is platform/architecture agnostic. This allows us to write portable SIMD code.

Bleeding edge code can be found in our git repository at https://www.gnuradio.org/git/volk.git/.

How to use VOLK:

For detailed instructions see http://libvolk.org/doxygen/using_volk.html

See these steps for a quick build guide.

Building on most x86 (32-bit and 64-bit) platforms

$ mkdir build
$ cd build
$ cmake ..
$ make
$ make test
$ sudo make install

# You might want to explore "make -j[SOMEVALUE]" options for your multicore CPU.

# Perform post-installation steps

# Linux OS: Link and cache shared library
$ sudo ldconfig

# macOS/Windows: Update PATH environment variable to point to lib install location

# volk_profile will profile your system so that the best kernel is used
$ volk_profile

Missing submodule

We use cpu_features as a git submodule to detect CPU features, e.g. AVX. There are two options to get the required code:

git clone --recursive https://github.com/gnuradio/volk.git

will automatically clone submodules as well. In case you missed that, you can just run:

git submodule update --init --recursive

that'll pull in missing submodules.

Building on Raspberry Pi and other ARM boards (32 bit)

To build for these boards you need specify the correct cmake toolchain file for best performance.

Note: There is no need for adding extra options to the compiler for 64 bit.

  • Raspberry Pi 4 arm_cortex_a72_hardfp_native.cmake
  • Raspberry Pi 3 arm_cortex_a53_hardfp_native.cmake
$ mkdir build && cd build
$ cmake -DCMAKE_TOOLCHAIN_FILE=../cmake/Toolchains/arm_cortex_a72_hardfp_native.cmake ..
# make -j4 might be faster
$ make
$ make test
$ sudo make install

# volk_profile will profile your system so that the best kernel is used
$ volk_profile

Supported platforms

VOLK aims to be portable to as many platforms as possible. We can only run tests on some platforms.

Hardware architectures

Currently VOLK aims to run with optimized kernels on x86 with SSE/AVX and ARM with NEON.

OS / Distro

We run tests on a variety of Ubuntu versions and aim to support as many current distros as possible. The same goal applies to different OSes. Although this does only rarely happen, it might occur that VOLK does not work on obsolete distros, e.g. Ubuntu 12.04.

Compilers

We want to make sure VOLK works with C/C++ standard compliant compilers. Of course, as an open source project we focus on open source compilers, most notably GCC and Clang. We want to make sure VOLK compiles on a wide variety of compilers. Thus, we target AppleClang and MSVC as well. Mind that MSVC lacks aligned_alloc support for aligned arrays. We use MSVC specific instructions in this case which cannot be free'd with free.

License

VOLK is migrating from the GNU General Public License v3.0 or later (GPL-3.0-or-later) to the GNU Lesser General Public License v3.0 or later (LGPL-3.0-or-later). At this point in time, much of the code in the repository is still GPL-licensed. New contributors are required to use the LGPLv3+ for their code contributions. Existing contributors are very kindly requested to re-submit their GPL-3.0-or-later license code contributions to LGPL-3.0-or-later by adding their name, GitHub handle, and email address(es) used for VOLK commits to the file AUTHORS_RESUBMITTING_UNDER_LGPL_LICENSE.md.

About

The Vector Optimized Library of Kernels

http://libvolk.org

License:Other


Languages

Language:C++ 87.9%Language:CMake 5.2%Language:C 2.8%Language:Python 2.1%Language:Assembly 1.5%Language:Shell 0.3%Language:Csound 0.2%