Skip to content

Latest commit

 

History

History
286 lines (185 loc) · 7.39 KB

dependencies.md

File metadata and controls

286 lines (185 loc) · 7.39 KB
Raw

Dependencies

Lensed has a number of compile-time dependencies, namely

If these dependencies are installed in a system-wide, default-accessible path, Lensed should be able to find them without any further intervention.

For XPA support and integration with SAOImage DS9, there is a further dependency on

For region file support, there is a further dependency on the

XPA and region file support must be enabled explicitly in the build options.

CFITSIO

The CFITSIO library is the standard for reading and writing FITS files.

Binary releases are statically linked against a recent version of CFITSIO and therefore do not depend on it. When building from source, it is necessary to have both the CFITSIO library and the header files available.

Packages

Because CFITSIO is a mature library, pre-compiled binary versions are often found in the repositories of package management systems. If it is possible to install new packages into the system, this is the easiest way to get CFITSIO.

Ubuntu Linux

For running a binary release, it should suffice to install the CFITSIO runtime:

$ sudo apt-get install libcfitsio3

When building Lensed from source, the CFITSIO development package must be used:

$ sudo apt-get install libcfitsio3-dev

Mac OS X (Homebrew)

The CFITSIO library can be found in the default Homebrew repository.

$ brew install cfitsio

Sources

In case it is not possible to install CFITSIO from a package, it can be built from source. This is straightforward:

$ curl ftp://heasarc.gsfc.nasa.gov/software/fitsio/c/cfitsio3370.tar.gz | tar -xz
$ cd cfitsio
$ ./configure  # --prefix=... and other options as necessary
$ make

If possible, CFITSIO should be installed into the system, using

$ sudo make install

or similar, depending on the platform at hand.

MultiNest

The MultiNest library is a modern sampler for posterior distributions and Bayesian inference problems. It can be downloaded from the website after (free) registration. There are two different versions of the library available,

  • a version using the CMake build system, and
  • a standard Makefile version.

Downloading and installing the CMake version is highly recommended, as it will automatically detect the build configuration and offers a way to install the library into the system.

MultiNest has a number of requirements of its own, among those are

  • a modern Fortran compiler, compatible with the 2003 standard, and
  • an implementation of the LAPACK library, which might need BLAS as well.

It is assumed that these requirements are met, as they depend heavily on the individual system. As a guideline for workstation machines running Linux, a combination of gfortran and OpenBLAS might be a good start. On Mac OS X, the shortest route to a working setup is installing gfortran via Homebrew, and the system already provides LAPACK in the Accelerate framework.

MultiNest with CMake

After downloading and extracting the MultiNest_v3.9_CMake.tar.gz archive, the building and installation should be straightforward:

$ cd MultiNest_v3.9_CMake/multinest
$ cd build
$ cmake .. -G "Unix Makefiles"
$ make

If possible, it is a good idea to install MultiNest into the system, so that it can be picked up easily by the build system. This can be achieved with

$ sudo make install

or a similar command, depending on the platform.

MultiNest with make

If CMake is not available, it is possible to build the Makefile version of MultiNest. After extracting the MultiNest_v3.9.tar.gz archive, it might be necessary to edit the contained Makefile. The default settings are for a system using gfortran and -llapack, respectively. Alternative settings for Intel's ifort compiler are also present, but commented out. Once the Makefile is set up correctly, the library can be built using

$ make

as usual.

Please note that the Makefile version of MultiNest does not provide facilities for system-wide installation. The name of the library differs as well, using the name libnest3 instead of libmultinest. Finally, the header file for C is not directly exposed, but contained within the folder of the C example. All of these things have to be kept in mind when preparing Lensed for building.

OpenCL

Lensed was designed from the start for heterogeneous computing environments, using OpenCL to communicate with both CPU and GPU devices through a unified programming platform.

A OpenCL runtime library for the compute devices must present in the system. Such a library usually comes with the driver of an OpenCL-enabled device. When building Lensed from source, it is further necessary to have the OpenCL headers installed.

Packages

Ubuntu Linux

For the OpenCL runtime library, it is necessary to install the device drivers for the CPUs/GPUs present in the system. Please refer to the Ubuntu manual for information.

The OpenCL headers necessary for compiling Lensed can be found in package opencl-headers.

$ sudo apt-get install opencl-headers

Linux (general)

The device vendors usually provide Linux drivers and SDKs for their platforms. Some useful links are

Mac OS X

Mac OS X ships with the libraries and headers required to build OpenCL programs by default.

XPA

For XPA support and DS9 integration, the

is required.

Sources

The XPA library and command line tools can be installed quickly from the GitHub repository:

$ curl -L https://github.com/ericmandel/xpa/archive/master.tar.gz | tar xz
$ cd xpa-master
$ ./configure  # --prefix=... and other options as necessary
$ make

For use with Lensed, it is not necessary to have X11 or Tcl integration.

$ ./configure --without-x --without-tcl  # --prefix=...
$ make

If possible, the XPA libraries and command line tools should be installed into the system, using

$ sudo make install

or similar, depending on the platform at hand.

Packages

Mac OS X (Homebrew)

Under Mac OS X, the library can be installed using Homebrew from the official X11 repository. Please note that this currently requires an installation of XQuartz, even though building from source does not (see above).

$ brew install homebrew/x11/xpa

Regions

For region file support, the

is required.

Sources

The regions library and command line tools can be installed quickly from the GitHub repository:

$ curl -L https://github.com/ericmandel/regions/archive/master.tar.gz | tar xz
$ cd regions-master
$ ./configure  # --prefix=... and other options as necessary
$ make

If possible, the regions library and command line tools should be installed into the system, using

$ sudo make install

or similar, depending on the platform at hand.