PRISSMA

Frequently Asked Questions

    What's new in version 1.2?

    Bug corrected in version 1.0?

    • Bug corrected for the calculation of radiative flux. Qr must be reduced over all processor before doing scattering because scattering uses domain decomposition. It is done in spatial_scheme.F

    What's new in version 1.0?

    • The calculation is done using data at nodes instead of cells. This modification allows to decrease significantly the spectral time calculation with global models. New routines are created for gathering/scattering values. These routines use domain decomposition for parallel calculations. All the tools have been updated to work with data at nodes. If the spectral model is SNBcK the gather is done directly on the data (T,Xi,p) before calculating the absorption coefficient. It allows to not increase the time calculation of the geometrical part with gathering.
      • Modification to calculate to results at node with scattering. The calculation of Qw was done in outprocessing using Qr at cells. Now the calculation of Qw is done inside of slave.F before scattering.
    • A new routine is used for WSGG to gather the WSGG weight on faces to calculate correctly Lo. This modification should improve the WSGG results close to the boundaries.
    • Some useless vectors where removed specially on the master. The new version with data at nodes and memory reductions save about 40% memory for the master. Parameter for exponential scheme are now reeded by slaves and only if exp scheme is used. Removed storage of Le at each face/cell in spatial scheme. The vector H is calculated only on the boundary faces. Calculation of Lotot is removed.
    • New routine is add to find the index of the boundary cell/face using a dichotomy in the spatial scheme. It seems that the geometrical time is slightly reduced.
    • Some routines have been rewritten and simplified : slave.F, band_integ.F. Spatial_scheme.F replaces scheme_dmfs.F and scheme_exp.F. Simplifications in code using modules to avoid argument passing.
    • Preprocessing compilation key are used to include PALM instructions for coupling unit.
    • The Makefile is updated to correct subroutines modules dependencies.
    • A script avbp2prissma is added : it allows to write the initial solution from an AVBP solution (AVBP is required to compile the tool AVBP2PRISSMA).
    • There was a bug in the tool solinit : the total number of boundary cells was wrong, it is correctly counted now in initbc.