################### # Release PRISSMA 1.2 # ################### Web -> http://www.cerfacs.fr/prissma/ Flowchart -> http://www.cerfacs.fr/prissma/index.php?body=flowchart SVN Installing : -> http://www.cerfacs.fr/prissma/index.php?body=installing external -> svn checkout http://www.cerfacs.fr/prissmasvn/tags/PRISSMA_V1.2 INSTALL_DIR internal CERFACS -> svn checkout svn+ssh://login@dogon/home/cfd1/domasium/svn/DOMASIUM/tags/PRISSMA_V1.2 INSTALL_DIR Available hostypes : -> BLUEGENE, CORAIL, DIDEROT, EGEE, EMILION, IMHOTEP, JADE, KALI, MAC, MADNESS, OCTOPUS, PCLINUX Test cases external -> svn checkout http://www.cerfacs.fr/prissmasvn/branches/TEST_CASES internal CERFACS -> svn checkout svn+ssh://emac@dogon/home/cfd1/domasium/svn/DOMASIUM/branches/TEST_CASES PRISSMA solves the Radiative Transfer Equation (RTE) in gaseous participating media without scattering. Spatial finite volume schemes: -> DMFS (Diamond Mean Flux Scheme), EXP (Exponential Scheme); Joseph et al. International Journal of Thermal Sciences, 2005. Discrete ordinate method with angular quadrature: -> Sn, LC11 Spectral Models: -> CK, FSK (Homogeneous), FSCK, TFSK, TFSCK, WSGG, GRAY -> Database : SNB (SNB-CK, SNB-FSCK,..) or NBKM (NBKM-CK, NBKM-FSCK,...) Memory usage for 1 000 000 cells: PREPROCESSING -> 2.1 Gb PRISSMA -> CK : 495 Mb -> WSGG : 485 Mb -> TFSCK : 853 Mb, table = 381 Mb Important modifications : #------------------- - Screen output cleared. An approximation of the maximum memory usage printed at the beginning of the calculation (ie. from a value defined in dom_constants.HH) - Update for precompilation with palm. PL_COMM_EXEC is not a parameter and can not be defined in the module mod_pmm.F. Now the variable COMM_PARA is filled in pmm_begin.F. Makefile has been updated for coupling. If the coupling library has been compiled the objects .o which use the coupling pre-compilation key are cleaned to avoid compatibility troubles. - Parallel calculation over frequencies and directions. Theoretic speed-up: - CK model : 8 904 (24 directions); 35 616 (96 directions) - Global model with 15 gases : 360 ; 1440 - The convergence test for the reflexion for global models is done outside the loop on the quadrature points which accelerate calculation with reflexion. The solution for a case with purely reflecting walls is slightly more diffusive. The explanation is that some quadrature points needed more iterations and have a small weight on the integrated quantities. This modification does not work for the SNB-CK model and is not used. - The test criterion of the reflexion is done on H. This vector is smaller than G which allows to reduce the MPI communications in the spatial scheme. The integration of the parallel calculation over quadrature point/frequencies and directions is now down just after scattering in a new subroutine postprocessing.F for Lb, G, Qr and Qp. - MPI communications (send/receive) removed and replaced by MPI_ALLREDUCE at the end of the calculation: pmm_returnvectors.F, slave_return.F and MPIREDUCE in band_integ.F/band_integ_snb.F are removed. Reduction of the memory usage with the MPI changes some vectors has been removed. - New spectral Model : FSCK from Liu et al. 2004. This model introduces Planck weighting for non uniform media. It is possible to specify the reference state and the Planck temperature used for weighting. These parameters are optional and are set on the same line of the spectral model in prissma.choices. The first parameter is an integer, possible values are : 0 : reference state is given in the file input_fsck.dat 1 : by the averaged quantities on the total volume 2 : by the maximal temperature 3 : by the maximal H2O/CO2 molar fraction The second temperature is the Planck temperature. If Tp=0.d0 the model uses the gas temperature Tp=Tg. Ex : FSCK 2 2000.d0 If not parameters are given, by default the model use the averaged quantities on the total volume for reference state (1) and Tp=Tg (0) (ie. FSCK alone => FSCK 1 0d.0) Calculation of weights at boundaries for FSCK: then optional in band_integ.F (not activated). - New narrow band database : NBKM from CETHIL. Optional line 10 in prissma.choices. If line 10 is blank SNB is chosen by default. - Keywords for spectral models has been renamed : - CK replaces SNB-CK; - FSK (hypothesis of homogeneous media) replaces FSCK (old version 1.0) called SNB-FSCK in Poitou et al.; Proceedings of Eurotherm83, 2009; - TFSCK replaces TAB (ie. tabulated FSK); - FSCK is a new one model from Liu et al. Proceedings of Eurotherm83, 2004; - TFSCK is the tabulated version of FSCK. - Clipping in TFSCK/TFSK: A Clipping was introduced to set k=0 if T<300K or X_i<1E-5 in TFSK and TFSCK. - Using keyword DERI on line 5 of prissma.choices instead of the spatial scheme allows to calculate kappa derivatives. By default this derivative is calculated with temperature for other quantities the index of celldata in slave.F when calling deri_kappa or deri_kappa_snb must be modified in slave.F. The result is stored in the incident radiation field G.out. This option should be used with the CK model. It is cautious to calculate kappa derivatives with global models as they were not built to validate the derivative of kappa and it should be tested. - Probes lists : It is possible to store radiative heat flux at some location using the new tool Track_cells. Once you have produce Track.in, you have to specify the vector by filling the third, the fourth and the fifth column in it. When you run PRISSMA, the code will read Track.in to extract incident radiative flux in the normal direction, opposed to the normal direction and the net radiative flux at the point specified. Output of PRISSMA will be written in Qr_probe.out How to : 1 - Run track_cell and fill it with the appropriate parameters: first line locate the INFILES directory where the PRISSMA mesh is. In following lines put x y z coordinates of point to track. You can add as many lines as you want. 2 - Rerun track_cell. This will find cell in PRISSMA mesh and create a Track.in in INFILES directory. 3 - Fill the 3rd, 4th and 5th column in Track.in with the coordinates of the normal. 4 - you can run PRISSMA, this will create a Qr_probe.out in OUTFILES directory where you will find : - Incident Radiative flux in the normal direction, - Incident Radiative flux opposed to the normal direction, - Net Radiative flux at the point. TOOLS : #------ - PREDATAS: Add initialisation, Fix patch vector to have the right number of boundary faces - DOM2ASCII corrected to avoid divide by 0 during interpolation. - INCONVERT: bug corrected. - OUTCONVERT: Kp removed, calculated by visu in postprocessing - TABFSK: it has been modified to tabulate either FSK or FSCK model. - Update c routine for CORAIL endianess. - createdomcase was modified to copy datas in the local directory instead of linking them. New tools (T. Pedot) : #---------------- - Balance flux : Calculates the radiant energy transmitted trough each patch, makes the sum and compares it to the radiative source terms in the domain (which should be equal). - Quadtracing : Traces the spatial quadrature for probes. Writes a Track.out for visu tools. - Track_cells : Finds the closest node to specified coordinates and creates a Track.in. Incident radiation flux can be taken in PRISSMA.