Home My Page Projects Code Snippets Project Openings FullSWOF_2D
Summary Activity Forums Tracker Tasks Docs News SCM Files Listes Sympa


Activity: Start Date: End Date:

Time Activity By
14:35:39 Source Code scm commit: The order 2 is parallelized but the results are fairly different from the sequential case. This is due to the contribution of the vector hsa but I don't know why. Frist order, at least int he case of "radial dam dry" the parallel code works perfectly giving the same results that in the sequential code. Performance, on my laptop, does seem to be small. (r451)marco mancini
12:19:25 Source Code scm commit: added 0.5 in the center for radial_dam_dry initialisation (r450)marco mancini
14:14:23 Source Code scm commit: parallelised lecture of huv.txt in intialisation (r449)marco mancini
13:31:22 Source Code scm commit: Added forgotten MPI communications in order1.cpp. Now the test-case of the column produce de same results with 1 or 4 procs (r448)marco mancini
11:04:45 Source Code scm commit: Analyser : FileMpiClass completed: now it is very user friendly. Animations added. (r447)marco mancini
16:45:50 Source Code scm commit: Analyser : set in classes ..... result with 4 proc differs from that with 1 (r446)marco mancini
14:48:33 Source Code scm commit: Changed again header in output to make it easier to retrive info for procs .... Added simple analyser to plot parallel (r445)marco mancini
14:03:54 Source Code scm commit: Many changes : - outputs are moved from Parallel to Output derives classes - initialization of huv is moved from Parallel to initialization_hub classes - Parallel class is passed const (r444)marco mancini
07:29:12 Source Code scm commit: svn propset svn:ignore build . (r443)marco mancini
12:17:45 Source Code scm commit: Ingnoring OTHERS svn no-relevant fils (r442)marco mancini
11:54:27 Source Code scm commit: -Parallel class is now derived from boost::mpi::cartesian_communicator -New parametres are introduced (mpi_proc_x, mpi_proc_y) to define mpi dimensions. If not given default is calculated. -Parallel class is more simple. -The parallel class is not more longer "globlal" but it must be passed to schemas. -C binds MPI are still used, in particular for communications but in -the future they should be replaced by boost::mpi communications (r441)marco mancini
09:58:00 Source Code scm commit: Intermediate version before trying to use boost::cartesian_communicator (r440)marco mancini
15:02:33 Source Code scm commit: Parallel : upto the end it run. Now the results (r439)marco mancini