A new major revision of the software has been developed and is available at https://github.com/bek0s/gbkfit
GBKFIT is a high-performance open-source software for modelling galaxy kinematics from 3D spectroscopic observations.
For instructions on how to install GBKFIT read here.
GBKFIT is a new software and thus it has not been tested in many different platforms. If you find problems or bugs related to the installation process please send me an e-mail or create a new GitHub issue. Your help is greatly appreciated.
For instructions on how to use GBKFIT read here.
GBKFIT is developed by me (Georgios Bekiaris) during my PhD studies at the Swinburne University of Technology and my visit to University of Toronto.
If you use GBKFIT in a publication please cite: Bekiaris et al. 2016.
GBKFIT is a high-performance open-source software for modelling galaxy kinematics from 3D spectroscopic observations. It is written in C++, and uses the CMake build system.
GBKFIT features a modular architecture which allows it to use a variety of data models, galaxy models, and optimization techniques. It also provides a clean object-oriented interface which enables programmers to create and add their own custom models and optimization techniques into the software.
GBKFIT models observations with a combination of two models: a Data Model
(DModel), and a Galaxy Model (GModel). The former is used to describe the data
structure of the observation, while the latter is used to describe the
observed galaxy. By convention, the name of the data and galaxy models start
with gbkfit.dmodel. and gbkfit.gmodel. respectively.
In GBKFIT, the optimization techniques are called fitters, and by convention,
their names start with gbkfit.fitter..
Galaxy kinematic modelling is a computationally intensive process and it can result very long run times. GBKFIT tackles this problem by utilizing the many-core architectures of modern computers. GBKFIT can accelerate the likelihood evaluation step of the fitting procedure on the Graphics Processing Unit (GPU) using CUDA. If there is no GPU available on the system, it can use all the cores available on the Central Processing Unit (CPU) through OpenMP.
GBKFIT comes with the following data models:
gbkfit.dmodel.mmaps_<device_api>: This model is used to describe moment maps extracted from a spectral data cube. Thus, this model should be used to perform 2D fits to velocity and velocity dispersion maps. Flux maps are also supported but they are currently experimental and should not be used.gbkfit.dmodel.scube_<device_api>: This model is used to describe spectral data cubes. Thus, this model should be used to perform 3D fits to spectral data cubes. Support for 3D fitting is experimental and should be avoided for now.
<device_api> can be either omp (for multi-threaded CPU acceleration) or
cuda (for GPU acceleration).
GBKFIT comes with the following galaxy models:
-
gbkfit.gmodel.gmodel1_<device_api>: This model is a combination of a thin and flat disk, a surface brightness profile, a rotation curve, and an intrinsic velocity dispersion which is assumed to be constant across the galactic disk.The following surface brightness profiles are supported:
- Exponential disk
The following rotation curve profiles are supported:
- Linear ramp (Wright et al. 2007)
- Arctan (Courteau 1997)
- Boissier (Boissier et al. 2003)
- Epinat (Epinat et al. 2008)
<device_api> can be either omp (for multi-threaded CPU acceleration) or
cuda (for GPU acceleration).
GBKFIT comes with the following fitters:
gbkfit.fitter.mpfit: This fitter employs the Levenberg-Marquardt Algorithm through the MPFIT library.gbkfit.fitter.multinest: This fitter employs a modified version of the Nested Sampling technique through the MultiNest library.
GBKFIT supports the following Point Spread Function (PSF) models: 2D elliptical Gaussian, 2D elliptical Lorentzian, and 2D elliptical Moffat. Alternatively, the user can supply a 2D image.
GBKFIT supports the following Line Spread Function (LSF) models: 1D Gaussian, 1D Lorentzian, and 1D Moffat. Alternatively, the user can supply an 1D image.