Installation

This section describes how to download and install the ADACS version of the LSC Algorithm Library Suite (LALSuite) as well as the Python package lal_cuda developed to ease testing during its development and to illustrate the use of the LALSimulation routines addressed by this project (including some minor changes to the LALSuite API).

Installing LALSuite

The installation of the ADACS branch of LALSuite is essentially the same as the standard version it branched from (see here for more information). In brief: to download the ADACS version of LALSuite (to /path/to/src/dir; adjust according to your needs) and compile/install the resulting libraries (to /path/to/dir/install; adjust according to your needs), perform the following:

$ cd /path/to/src/dir
$ git clone --single-branch -b ADACS https://github.com/ADACS-Australia/ADACS-SS18A-RSmith.git
$ cd ADACS-SS18A-RSmith
$ ./00boot
$ ./configure --enable-cuda --enable-python --prefix=/path/to/dir/install
$ make -j 24 install

Note

The GPU implementation of the code is in the ‘ADACS’ branch of this repository. The git clone command above will ensure that you are working with it before compiling.

Note

The –enable-cuda option is required for GPU acceleration. However, it can be omitted if an NVidia GPU is not available.

Note

The –enable-python option is required to use the lal_cuda Python package.

Installing the lal_cuda Python package

This package assumes that you have a working copy of the LSC Algorithm Library Suite (LALSuite) installed (see here for instructions). Furthermore, by default it assumes that you are working with the GPU-enabled version developed by ADACS, which can be obtained here . If this version is not being used, make sure all scripts in this package are run with the ‘–legacy’ flag and that the ‘legacy’ setting in the tests is enabled, if you want to use those. Furthermore, before working with the lal_cuda library make sure that the LALSuite SWIG libraries are added to your Python path using the approach for setting environment variables appropriate for your shell. For example, for bash this can be done as follows (using the install directory in the example above; adjust accordingly):

$ export PYTHONPATH=${PYTHONPATH}:/path/to/dir/install/lib/python2.7/site-packages/

To install lal_cuda, it needs to be downloaded and installed into your Python environment as follows:

$ cd /path/to/src/dir
$ git clone https://github.com/ADACS-Australia/ADACS-SS18A-RSmith-python
$ cd ADACS-SS18A-RSmith-python
$ make init
$ make install

Note

A makefile is provided in the project directory to ease the use of this software project. Type make help for a list of options.

Warning

If the ADACS version of LALSuite is not being used with the executables in this package, make sure to use the ‘–legacy’ flag and – if you intend to use them – enable the ‘legacy’ setting in the tests.

Warning

Make sure that the make init line is run first-thing before installing. It will ensure that all needed dependencies are present in your current Python environment. Make sure to re-run this line before re-installing, if you change Python environments.