Main features:                                                                           current version:   2.55beta

  • memory scalability via memory distribution of all main pixel and harmonic domain objects what ensures that the memory required per processor decreases (nearly) inversely proportionally to a number of allocated processors.
  • flexibility in accepting different distribution patterns of the objects over processors depending on the application.
  • nearly linear speed-up with increasing number of processors, i.e., the execution time is (nearly) inversely proportional to a number of allocated processors.
  • efficient performance on a single processor, i.e., comparable to the serial HEALPix 2.01 routines.
  • memory and load-balance optimization for full and partial sky maps.
  • applicability to any iso-latitude pixelization with pixel areas independent on the azimuthal angle.
  • simultaneous, multiple map analysis and synthesis.
  • parallel arbitrary spin transforms. !!! NEW in ver. 2.0 !!!
  • alternative Fast Fourier transform packages. !!! NEW in ver. 2.5 !!!
  • stable with gfortran. !!! from ver. 2.54 !!!

    > Top of the page

    Provenance and technicalities:

  • uses MPI for communication.
  • written in Fortran 90 with a C interface.
  • this software is a derivative of the spherical harmonic transforms included in the HEALPix package(*) and is based on both serial and MPI routines of its version 2.01, however ...
  • ... since version 2.5 this software is fully autonomous of HEALPix and can be compiled and run without the HEALPix library.

    (*) HEALPix has been developed by K. M. Gorski, E. Hivon, B. D. Wandelt, A. J. Banday, M. Bartelmann, H. K. Eriksen, F. K. Hansen, M. Reinecke and is distributed under the GNU General Public Licence.

    > Top of the page


  • The current version of the package is a version v.2.55 beta and can be downloaded from here. The software is distributed with the hope that it will be found useful but with all the caveats as implied by the GNU public license. (Last update: April, 30, 2012 and the log file of the recent changes (since the version 2.00) is here.)
  • Algorithmic approach.
  • On-line documentation (for the version 2.5beta).
  • Performance tests.
  • Validation. (Coming soon !)
  • Example codes.
  • General comments.
  • Known bugs/problems.
  • Release history.

    > Top of the page

    Contact info/bug tracking

    Please contact us and report problems and issues via google bug tracking facility. (external link).

    > Top of the page

    Principal developers

    • Radek Stompor (APC, France);
    • Mikolaj Szydlarski (INRIA, France; ANR-MIDAS'09);
    • Laura Grigori (INRIA, France).
    • see Acknowledgements section below for a list of contributors and validators.

    > Top of the page


    This work has been supported in part by
    • the European Community Marie Curie International Reintergration grant (FP6) from 2006-2008.
    • the Agence National de Recherche (ANR) grant MIDAS'09 awarded under the ANR-Calcul Intensif programme, 2010-2012.
    The code development and validation used high performance computing (HPC) resources of
    Thanx to:
    • the HEALPix guys for HEALPix.
    • Ovidiu Hupca for the first GPU version of the inverse spherical harmonic transform, alm2map.
    • Chris Cantalupo for the C/F90 interface tricks from the M3 Library.
    • Pierre Esterie, and Joel Falcou for help with the development of multi-core and CUDA versions.
    • Matt Tristram, Julien Grain and Giulio Fabbian for help in debugging and testing the software.
    • Ted Kisner for help with debugging, suggestions of improvements, as well as for maintaining the library on the supercomputers at National Energy Research Scientific Computing (NERSC) Center.

    > Top of the page

    Relevant links

  • Software using S2HAT library:
    • pureS2HAT -- a set of C-routines allowing for the 'pure multipole' calculations.
    • computeXlsAll -- a routine to reduce sets of alm coefficients precomputed and distributed by the S2HAT transforms (multi map, polarized, spin-weighted, etc) to a set of all plausible cross-spectra.
  • Some other working spherical harmonic transform packages and libraries:
  • Check ANR-MIDAS'09 webpage to learn more about our CMB data analysis software development effort.

  • > Top of the page > Home

    CNRS Centre National de la Recherche Scientifique
    Laboratoire Astroparticule et Cosmologie
    Equipe de Traitement des Données et Simulations (ADAMIS)
    APC Contact: Radek Stompor (radek(at)apc(dot)univ(dash)paris7(dot)fr)

    © 2007 rs