Installation

Once you have the downloaded the package, you will need to compile WSClean. WSClean requires:

• Casacore, for opening measurement sets. Version >=2.0 is required, not lower. Casacore is required even if you already have Casa installed. Casacore needs to be compiled with C++11 support.
• FFTW version 3.3.5 or newer, used to perform Fourier transformations.
• Boost, used for threading, date and time calculations and some other general functionalities.
• CFITSIO, for reading and writing FITS files.
• GSL, the GNU Scientific Library, used for certain computations.

WSClean uses C++11 features. Because of this, building WSClean with GCC requires at least GCC version 4.8.1.

To use the image domain gridder (a fast GPU gridder), you will need to install the IDG libraries from https://gitlab.com/astron-idg/idg .

After you have installed these libraries, you can compile WSClean with the following commands:

mkdir -p build
cd build
cmake ../
make -j 4

And optionally:

sudo make install

If cmake reports errors, you might have to install or specify your libraries in the call to cmake if they are not in standard directories, e.g.:

cmake ../ -DCMAKE_PREFIX_PATH=/opt/cep/casacore/

to add that directory to the search path. To add multiple directories to the search path, put the paths between double quotes and separate them with a semicolon:

cmake ../ -DCMAKE_PREFIX_PATH="/path/to/casacore;/path/to/cfitsio"

If you want to make full-sky images or use the recentering technique, you will probably want to use the chgcentre tool too. This tool changes can change the phase centre of a measurement set, and is available in a separate package available on this website.

On Ubuntu and Debian

Binary packages are available on Ubuntu and Debian, and can be installed with sudo apt-get install wsclean. In case that version is new enough for your purpose, you're all done. If you want to compile WSClean from source, the following packages need to be installed:

apt-get install casacore-dev libgsl-dev libhdf5-dev libfftw3-dev libboost-dev \
libboost-date-time-dev libboost-filesystem-dev libboost-program-options-dev \
libboost-system-dev libboost-thread-dev libcfitsio-dev cmake g++

The LOFAR beam and IDG libraries are optional, but need to be installed manually from source if they are required (see elsewhere on this page).

On Red Hat

The following document lists some instructions that can be helpful for instaling WSClean on Red Hat: Installing WSClean on rhel 7.6 (Thanks to Leonardo Saavedra from NRAO).

Linking errors

If you get undefined reference errors in casacore-code when you compile WSClean, e.g. similar to "undefined reference to casa::ArrayColumn<float>::get(unsigned int) const", it probably means that you did not compile Casacore with C++11 support. When you compile Casacore, you need to turn this on explicitly:

anoko@leopard:~/casacore/build$ cmake ../ -DCXX11="ON"

after which cmake should respond with a list of enabled features, including:

-- C++11 support ......... = ON

Whether it is necessary to switch on C++11 depends on the version of the compiler -- with newer compilers it is turned on by default.

Compiling platform independently / portability

By default, cmake will create a binary that is optimized for the machine that it is compiled on, and will only work on machines that contain the same instruction set as the compiling machine. In other words, the same binary might not work on other machines, because it might use advanced instructions such as AVX instructions that might not be available on another machine. It has been reported that this can e.g. lead to an "Illegal instruction" error (see ticket 50).

If you want to make a binary that can be used on different platforms, you can add -DPORTABLE=True to cmake:

cmake ../ -DPORTABLE=True

You should note that this makes the binary in general slower, so you should not use it unless you have to.

Using the LOFAR beam

NB: WSClean used to use the LOFAR beam from the LOFAR repository, but versions after 2.6 require the "LOFARBeam" package from Github instead. The LOFARBeam package can be find here: https://github.com/lofar-astron/LOFARBeam

To have the LOFAR beam available in WSClean, CMake needs to find the LOFARBeam installation on your computer. If you have installed LOFARBeam in a custom directory, you can add it to your search path. For example, if LOFARBeam been installed to ~/Software/LOFARBeam-install, this cmake command will use it:

cmake ../ -DCMAKE_PREFIX_PATH="~/Software/LOFARBeam-install/"

CMake will tell whether the LOFAR tree was found:

cmake ../ -DCMAKE_PREFIX_PATH="~/Software/LOFARBeam-install/"
[..]
LOFAR beam library found.

CMake searches for the file StationResponse/ITRFConverter.h and the library libstationresponse.a or libstationresponse.so.

Extra paths can be added to e.g. also include IDG. Paths can be separated with a ; (semicolon).

Using the MWA beam

The MWA beam requires that a .h5 file with beam coefficients is present in your path, which you can find in the MWA software tools. This file needs to be in your Python search path. WSClean will iterate over your Python search paths by executing the following Python program:

import sys
  for a in sys.path
  print a