VOLK_GNSSSDR is a sub-project of GNSS-SDR. This library provides a set of extra kernels that can be used stand-alone or in combination with VOLK's. Please see https://www.libvolk.org for documentation, source code, and contact information about the original VOLK library.
The boilerplate of this code was initially generated with volk_modtool, an
application provided by VOLK that creates the skeleton that can then be filled
with custom kernels. Some modifications were added to accommodate the
specificities of Global Navigation Satellite Systems (GNSS) signal processing.
Those changes are clearly indicated in the source code, and do not break
compatibility.
This library contains kernels of hand-written SIMD code for different
mathematical operations, mainly with 8-bit and 16-bit real and complex data
types, offering a platform/architecture agnostic version that will run in all
machines, plus other versions for different SIMD instruction sets. Then, the
application volk_gnsssdr_profile runs some iterations of all versions that
your machine can execute and annotates which is the fastest, which will then be
selected at runtime when executing GNSS-SDR. In this way, we can address at the
same time portability (by
creating executables that will run in nearly all processor architectures) and
efficiency (by providing
custom implementations specially designed to take advantage of the specific
processor that is running the code).
These kernels have some specific features (e.g., saturation arithmetics) that are aimed to GNSS signal processing, but could make them not suitable for its general use in other applications. Check out the documentation generated by Doxygen and the generic (that is, plain C) implementation to see what each kernel is actually doing.
This figure shows the role of some VOLK_GNSSSDR kernels in the context of a GNSS baseband processor:
This library is automatically built and installed along with GNSS-SDR if it is not found by CMake on your system at configure time.
However, you can install and use VOLK_GNSSSDR kernels as you use VOLK's, independently of GNSS-SDR.
If you want to install the VOLK and VOLK_GNSSSDR libraries in the same machine, please install VOLK before building VOLK_GNSSSDR in order to avoid a clash of the cpu_features library, used internally by both. The CMake script will detect the presence of cpu_features and will make use of it if already installed, thus avoiding to install it twice.
First, make sure that the required dependencies are installed in your machine:
$ sudo apt-get install build-essential python3-mako cmake git ca-certificates \
libboost-dev libboost-filesystem-dev libboost-system-dev
Please note that if you are using a compiler supporting the C++17 standard (for
instance, gcc >= 8.0), specifically the std::filesystem library, packages
libboost-dev, libboost-filesystem-dev and libboost-system-dev are no
longer required dependencies. The CMake script will detect that availability for
you.
In order to get the source code, build and install the library, do:
$ git clone https://github.com/gnss-sdr/gnss-sdr
$ cd gnss-sdr
$ git checkout next
$ cd build
$ cmake ../src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr
$ make
$ sudo make install
$ sudo ldconfig
That's it!
Before its first use, please execute volk_gnsssdr_profile to let your system
know which is the fastest available implementation for each kernel. This only
has to be done once:
$ volk_gnsssdr_profile
From now on, GNSS-SDR (and any other program of your own that makes use of VOLK_GNSSSDR) will benefit from the acceleration provided by SIMD instructions available in your processor.
The execution of volk_gnsssdr_profile can be set automatically after building,
leaving your system ready to use:
$ git clone https://github.com/gnss-sdr/gnss-sdr
$ cd gnss-sdr
$ git checkout next
$ cd build
$ cmake -DENABLE_PROFILING=ON ../src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr
$ make
$ sudo make install
$ sudo ldconfig
To build for these boards you need specify the correct CMake toolchain file for best performance.
- Raspberry Pi 4:
arm_cortex_a72_hardfp_native.cmake - Raspberry Pi 3:
arm_cortex_a53_hardfp_native.cmake
Example for Raspberry Pi 4:
$ git clone https://github.com/gnss-sdr/gnss-sdr
$ cd gnss-sdr
$ git checkout next
$ cd build
$ cmake -DCMAKE_TOOLCHAIN_FILE=../cmake/Toolchains/arm_cortex_a72_hardfp_native.cmake \
../src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr
$ make
$ make test
$ sudo make install
$ volk_gnsssdr_profile
$ sudo ldconfig
With Visual Studio 2019, in a Powershell run by administrator:
$ git clone https://github.com/gnss-sdr/gnss-sdr
$ cd gnss-sdr
$ git checkout next
$ cd build
$ cmake -G "Visual Studio 16 2019" ..\src\algorithms\libs\volk_gnsssdr_module\volk_gnsssdr
$ cd ..
$ cmake --build build --config Release
$ cd build
$ ctest -C Release
$ cd ..
$ cmake --install build
If you use VOLK_GNSSSDR in your research and/or software, please cite the following paper:
- C. Fernández-Prades, J. Arribas, P. Closas, Accelerating GNSS Software Receivers, in Proc. of the 29th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2016), pp. 44-61, Portland, Oregon, September 12-16, 2016. DOI: 10.33012/2016.14576
BibTeX entry:
@inproceedings{Fernandez16-ion,
author = {C.~{Fern\'{a}ndez--Prades} and J.~Arribas and P.~Closas},
title = {Accelerating {GNSS} Software Receivers},
booktitle = {Proc. 29th Int. Tech. Meeting Sat. Div. Inst. Navig.},
pages = {44--61},
year = {2016},
address = {Portland, OR},
month = {Sep.},
note = {{DOI}: 10.33012/2016.14576}
}
Citations are useful for the continued development and maintenance of the library.
VOLK_GNSSSDR was originally created by Andres Cecilia Luque in the framework of the Summer Of Code In Space (SOCIS 2014) program organized by the European Space Agency, and then evolved and maintained by Carles Fernández-Prades and Javier Arribas. This software is released under the GNU General Public License version 3, see the file COPYING.
This project is managed by Centre Tecnològic de Telecomunicacions de Catalunya.