CondaInstall
Why use conda?
Conda is a cross-platform package manager (supporting Linux, macOS, and Windows) that makes it easy to install GNU Radio, its dependencies, and out-of-tree modules in a self-contained environment. Conda lets you create independent environments with their own sets of packages, and those environments are separate from your system installation and other package managers. If you've struggled with installing GNU Radio by other methods, you want to use GNU Radio with other bleeding-edge software available through conda-forge, or you want to try out a new version without affecting your tried-and-true system installation, conda may be right for you! In addition to GNU Radio, there are also related software packages you can install that may be of interest.
Installation using conda
The conda packages for GNU Radio are unofficial and supported through volunteer effort within the conda-forge community. Conda-forge is a community-maintained collection of conda packages where anyone with interest can help improve or add packages. Please report and discuss any packaging-specific issues with the maintainers at the gnuradio feedstock. If you'd like to help or are interested in adding other related software to conda-forge (e.g. out-of-tree modules), get in touch with the maintainers through Github.
Step 1: Install conda itself
First, you will need to install a base conda distribution so that you can use the conda
program to create an environment and install the GNU Radio packages. If you already have this and can run conda
on a command line or use the Anaconda Navigator to install packages, skip to the next step.
Alternatively, there is a full conda distribution called radioconda available at [1] that includes GNU Radio. You can follow the instructions there and then skip all of the installation instructions here. If you don't know much about conda
and just want to get going, that's a good place to start. Otherwise, continue on.
If this is your first time using conda, we recommend miniforge for the base installation. This provides a minimal environment to run conda
that is already set up to pull packages from conda-forge (which is where you can find GNU Radio). Download and install miniforge by following the instructions at [2]. Installers for many different platforms are available there, but the most relevant are the following:
For Linux and macOS, run the downloaded installer from a console and follow the prompts:
- (Linux)
bash Miniforge3-Linux-x86_64.sh
- (macOS)
bash Miniforge3-MacOSX-x86_64.sh
By default, this will install miniforge to your home directory at ~/miniforge3
and allow you to use the conda
command from any console.
For Windows, double click on the downloaded installer executable and follow the prompts. By default, this will install miniforge to your user directory at C:\Users\USERNAME\miniforge3
and add a "Miniforge Prompt" entry to your start menu. Running the "Miniforge Prompt" should bring up a console window where your base conda environment is activated and you can use the conda
command.
This guide provides a nice introduction to using conda generally; it may help you better understand the rest of these instructions, but it is not necessary to read.
(Optional, experts only) It may be convenient to now install mamba
in your base environment and use it in place of the conda
command. Mamba is a third-party alternative to conda that is written in C instead of Python, and its main benefit is faster dependency resolution when installing packages. If you do install mamba, you can simply use the mamba
command in place of conda
for all subsequent steps.
Step 2: Create an environment for GNU Radio
It's almost never a good idea to install additional packages to your base conda environment because their dependencies may interfere with conda's own dependencies. We recommend creating an environment specifically for GNU Radio and related packages. From a console with the base conda environment activated, run
conda create -n gnuradio
to create an empty environment called "gnuradio". Enter the environment by activating it with the command:
conda activate gnuradio
Before installing the GNU Radio packages, we need to ensure that the environment is set up to look for packages from conda-forge:
conda config --env --add channels conda-forge conda config --env --set channel_priority strict
Now you're ready to actually install GNU Radio!
Step 3: Install GNU Radio from conda-forge
GNU Radio on conda-forge is split into a few subpackages. Most users will be happy with the full installation provided by the gnuradio
metapackage. From within your activated "gnuradio" environment (previous step), run the command:
conda install gnuradio
This will install the latest available version of GNU Radio along with the latest version of Python. If you want a specific version of the gnuradio
package (get a list of possibilities from conda search gnuradio
), you can specify it like
conda install gnuradio=3.8.2
If you want a specific version of Python, you can install it before gnuradio
or specify them together like
conda install gnuradio python=3.8
To upgrade all non-Python packages in your environment to their latest available versions, use the upgrade command
conda upgrade --all
The gnuradio
metapackage installs all of the following subpackages:
gnuradio-core
gnuradio-grc
gnuradio-qtgui
gnuradio-uhd
gnuradio-video-sdl
gnuradio-zeromq
If you don't want all of those and their dependencies, you can install the ones you'd like individually like
conda install gnuradio-uhd
to get only the core package (always a dependency) and UHD support without any GUI elements.
Using GNU Radio from conda
You can use GNU Radio from a conda environment in mostly the same way that it is normally used, provided that you have activated your "gnuradio" environment (or whatever name you might give it). So from a fresh console, you'll need to run
conda activate gnuradio
and then use whatever GNU Radio scripts or programs you want, for instance
gnuradio-companion
Windows users will find that GNU Radio Companion can also be launched from the Start Menu. Running it from there will first activate the appropriate environment, so all you need to do is click the icon!
GNU Radio will have access to the Python installed into the conda environment, and not any other installation. If you need specific Python packages to use with GNU Radio, just install them into the environment using conda install
. As a last resort, pip
or manual installation (python setup.py install
) can also be used, but these will not be managed by conda and you'll have to be careful to manage your environment.
Beyond core GNU Radio and the in-tree modules, some out-of-tree modules and related software have also been packaged for conda. This is a non-exhaustive list of other packages that you can install using the conda install
command:
gnuradio-osmosdr
gnuradio-satellites
gnuradio-soapy
gqrx
limesuite
rtl-sdr
soapysdr-module-lms7
soapysdr-module-plutosdr
soapysdr-module-remote
soapysdr-module-rtlsdr
soapysdr-module-uhd
uhd
volk
If you have other hardware or OOT modules that you'd like to see supported, first search Anaconda Cloud to see if a package not listed above already exists on conda-forge or another channel. If it doesn't, you can create your own package recipe and submit it to conda-forge (see the documentation at [3] for details) and/or get in touch with the maintainers of the existing packages to see if they would be willing to help.
Building OOT modules to use with conda-installed GNU Radio
In the event that the OOT module you want to use is not packaged for conda, it is possible to build it from source and use it within your conda environment. If you do this, please consider contributing your recipe/procedure so that others can benefit. See the note at the end of this section.
First, if you're doing this on Windows, you will need to have Visual Studio installed, and it will probably need to include MSVC v141, e.g. by selecting the "MSVC v141 - VS2017 C++ x64/x86 build tools (v14.16)" component under the list of "Individual Components".
Then, some special setup is required to ensure that the CMake packaging tool knows where to find dependencies and where to install the OOT module. Install the packaging tools we'll need to the conda environment by installing the purpose-built meta-package:
conda install gnuradio-build-deps
Also make sure you have all of the OOT module's dependencies installed which might not be needed by GNU Radio itself.
Re-activate your environment if you just installed cxx-compiler
for the first time to set some environment variables:
(on Linux/macOS)
conda activate $CONDA_DEFAULT_ENV
(on Windows)
conda activate %CONDA_DEFAULT_ENV%
Then run CMake like you normally would (e.g. from a build directory), adding the following flags to your cmake
command, so cmake ..
becomes:
(on Linux/macOS)
cmake -G Ninja -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DLIB_SUFFIX="" ..
(on Windows)
cmake -G Ninja -DCMAKE_INSTALL_PREFIX="%CONDA_PREFIX%\Library" -DCMAKE_PREFIX_PATH="%CONDA_PREFIX%\Library" -DGR_PYTHON_DIR="%CONDA_PREFIX%\Lib\site-packages" ..
It is possible that not all of these added arguments are necessary, and you might need more to help CMake locate particular libraries.
Once this cmake
step succeeds, proceed with building the OOT as you would outside of the conda environment:
cmake --build . cmake --build . --target install
Document your results to help others: if you've succeeded in building an OOT in this way, please consider documenting your process in an issue filed at the radioconda GitHub! Helpful information to document includes the names of any extra conda packages that you needed to install as dependencies, and the exact CMake and build commands that you used.
Extra credit: if you're successful and want to be extra helpful, you're halfway to creating a conda recipe! Consider following the conda-build documentation to finish the job in creating a recipe and then submit it to conda-forge so others can benefit. You can use existing recipes (like gnuradio-satellites) as an example.
Installing pre-release versions
Sometimes, pre-release versions of GNU Radio will also be available through conda-forge (on a best effort basis!). So these don't get installed over release versions, they are provided on a separate channel label. For a release candidate ("RC" pre-release), that channel is conda-forge/label/gnuradio_rc
. You can install from a special channel by using the -c
flag with conda install
, as in
conda install -c conda-forge/label/gnuradio_rc gnuradio
Since extra packages are unlikely to also be available and compatible with a pre-release version, we recommend creating a separate environment first and installing the pre-release there. So for example,
conda create -n grtest -c conda-forge/label/gnuradio_rc -c conda-forge gnuradio conda activate grtest
Building GNU Radio from source within a conda environment
Take note that this is an advanced topic that is likely to be of interest to developers only. It is possible to install all of GNU Radio's dependencies into a conda environment and then build GNU Radio from source. This can be useful for development, where the source is being updated and rebuilds are frequent.
Create a development environment
Create and activate a development environment called "grdev":
conda create -n grdev conda activate grdev conda config --env --add channels conda-forge conda config --env --set channel_priority strict
Install all of the dependencies (or a subset of them if you like):
(Linux)
conda install alsa-lib boost-cpp click click-plugins cmake codec2 cppzmq cxx-compiler fftw gtk3 gmp gsl libad9361-iio libcblas libiio libsndfile log4cpp lxml mako ninja numpy pkg-config portaudio pybind11 pybind11-abi pygobject pyqt python pyyaml pyzmq qt qwt sdl soapysdr uhd volk zeromq libice-cos6-x86_64 libselinux-cos6-x86_64 libsm-cos6-x86_64 libx11-cos6-x86_64 libxau-cos6-x86_64 libxcb-cos6-x86_64 libxdamage-cos6-x86_64 libxext-cos6-x86_64 libxfixes-cos6-x86_64 libxrender-cos6-x86_64 libxxf86vm-cos6-x86_64 libuuid-cos6-x86_64 mesa-dri-drivers-cos6-x86_64 mesa-libgl-devel-cos6-x86_64 libudev-devel-cos6-x86_64
(macOS)
conda install boost-cpp click click-plugins cmake codec2 cppzmq cxx-compiler fftw gtk3 gmp gsl libad9361-iio libcblas libiio libsndfile log4cpp lxml mako ninja numpy pkg-config pybind11 pybind11-abi pygobject pyqt python pyyaml qt qwt sdl soapysdr uhd volk zeromq
(Windows)
conda install boost-cpp click click-plugins cmake codec2 cppzmq cxx-compiler fftw gtk3 gsl libad9361-iio libcblas libiio libsndfile log4cpp lxml mako mpir ninja numpy pkg-config pybind11 pybind11-abi pygobject pyqt python pyyaml qt qwt sdl soapysdr uhd volk zeromq
On Linux only, remove a symlink to the system udev that will get in the way of the linker:
rm $CONDA_PREFIX/lib/libudev.so.0
Pass necessary arguments to CMake
With the development environment activated and from a build directory within the source directory, call cmake
with the following arguments:
(Linux)
cmake -G "Ninja" -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -DLIB_SUFFIX="" ..
(macOS)
cmake -G "Ninja" -DCMAKE_PREFIX_PATH=$CONDA_PREFIX -DCMAKE_INSTALL_PREFIX=$CONDA_PREFIX -DLIB_SUFFIX="" ..
(Windows)
cmake -G "Ninja" -DCMAKE_INSTALL_PREFIX="%CONDA_PREFIX%\Library" -DCMAKE_PREFIX_PATH="%CONDA_PREFIX%\Library" -DPYTHON_EXECUTABLE="%CONDA_PREFIX%\python" -DGR_PYTHON_DIR="%CONDA_PREFIX%\Lib\site-packages" ..
In time, hopefully this list of argument can be trimmed down to only specifying the prefix.
Build as normal
Proceed with building and installing as normal:
cmake --build . cmake --build . --target install
FAQ
When I try to run GNU Radio, why do I get a ModuleNotFoundError (no module named ...)?
It is likely that you have another GNU Radio or Python installation interfering with your conda environment. See ModuleNotFoundError for more information on this problem in general. Verify that your Python path points to the correct directory by running
python -c "import sys; print('\n'.join(sys.path))"
and inspecting the output. If other directories come before your conda environment Python directory, you may have to modify your PYTHONPATH
environment variable. If your conda environment Python directory does not appear at all, make sure that your PATH
environment variable is set correctly to run the python
executable from your conda environment and not some other installation of Python.
Why do I get cryptic error messages with a new GNU Radio installation (or new major version)?
There are a few flavors of this, but if you've just installed a new major version of GNU Radio (e.g. 3.8, 3.9), it's possible that some files from an old installation are conflicting with the new version. It can be helpful to ensure that all remnants of the older version have been purged from your system, especially if you've just installed the conda package and have previously used another package or installed from source. In particular, it can be helpful to clear the GNU Radio user preferences directory and start fresh. You can find the location of this directory by running
gnuradio-config-info --userprefsdir
Delete that directory or all of its contents, and you might have better luck starting GNU Radio without errors.
Why can't GNU Radio find my hardware device?
Provided you've installed the appropriate package for your hardware through conda (e.g. uhd
) and the accompanying GNU Radio blocks (e.g. gnuradio-uhd
), it's likely that you still need to follow extra steps to get your hardware running. Consult the documentation for your device and follow the additional steps it recommends beyond installing the basic software. For example:
- For Ettus UHD-supported devices, you'll need to download the images package into the conda environment by running the
uhd_images_downloader
script provided byuhd
. - Using Windows with a USB-connected device, you will probably need to install a USB driver for the device. Search for driver installation instructions for your device and follow those instructions. You will likely have to download and run Zadig to install the latest WinUSB driver for your device.
Whatever you need to do, make sure your device is visible and usable outside of GNU Radio before trying to track down any GR-specific problem.