mpi_alm_tools*

This module implements MPI parallelization of the alm2map and map2alm routines. It is not compiled by default during installation, but rather intended for users who need massive parallelization in their own programming. Typical applications are Monte Carlo simulations and Markov chain type analyses.

The routines can be called in two modes, either simple or advanced. The former mimics the interface of the standard routines, but with an additional MPI handle as a first argument, and is intended for applications which requires only one or a few transforms. The latter interface provides both more flexibility (in particular the option of pre-computation of the Legendre polynomials) and a simpler interface when multiple transforms are required. This interface is particularly well suited for Monte Carlo simulations and Markov chain type analyses.

Location in HEALPix directory tree: src/f90/mod/mpi_alm_tools.f90 

EXAMPLE:

• Simple one-line interfaces:
  • mpi_map2alm_simple
  • mpi_alm2map_simple

• Three-step advanced interfaces:
  1. Initialization:
     mpi_initialize_alm_tools
  2. Execution of spherical harmonics transforms
     • mpi_map2alm (root processor)
     • mpi_alm2map (root processor)
     • mpi_map2alm_slave (slave processor)
     • mpi_alm2map_slave (slave processor)
  3. Finalizing:
     mpi_cleanup_alm_tools