Changes between Version 11 and Version 12 of doc/tec/changelog_2018

Timestamp:

Sep 15, 2010 11:59:05 AM (14 years ago)

Author:

raasch

Comment:

--

doc/tec/changelog_2018

@@ -433 +433 @@
 |----------------
 {{{#!td style="vertical-align:top;width: 50px"
+02/03/10
+}}}
+{{{#!td style="vertical-align:top;width: 50px"
+SR
+}}}
+{{{#!td style="vertical-align:top;width: 75px"
+r496
+}}}
+{{{#!td style="vertical-align:top"
+3.7a
+}}}
+{{{#!td style="vertical-align:top"
+C
+}}}
+{{{#!td style="vertical-align:top"
+Particle data for restart runs are only written if environment variable {{{write_binary}}} = ''.T.'', i.e. particle data are now handled in the same way as the other restart data. (palm)
+}}}
+|----------------
+{{{#!td style="vertical-align:top;width: 50px"
+01/03/10
+}}}
+{{{#!td style="vertical-align:top;width: 50px"
+SR
+}}}
+{{{#!td style="vertical-align:top;width: 75px"
+r494
+}}}
+{{{#!td style="vertical-align:top"
+3.7a
+}}}
+{{{#!td style="vertical-align:top"
+N
+}}}
+{{{#!td style="vertical-align:top"
+Output in netCDF4-format.
+
+netCDF4 allows for parallel I/O. Also, variables are not limited in size any more (i.e. arrays can be larger than 2GB). The new d3par-parameter [wiki:/doc/app/d3par#netcdf_data_format netcdf_data_format] can be used to select among the different netCDF file formats (1: classic format, 2: 64-bit-offset format (this is now the default), 3: netCDF4/HDF5 format, 4: netCDF4 format with NF90_CLASSIC_MODEL bit set). Setting {{{netcdf_data_format}}} > 2 only affects 2d- and 3d-output, i.e. there is no parallel output for profiles, time series, spectra, and particles. Parallel output of particle data may be realized in the future.
+
+In case of parallel output, all PEs write into one file, which means that postprocessing the data with {{{combine_plot_fields}}} (merging data from single files - one per PE - into one file) is not neccessary. In this case, flag files are automatically set, telling {{{combine_plot_fields}}} that nothing is to be done.
+
+Using netCDF4 format requires a netCDF4/HDF5 library. Respective settings for {{{-I}}}, {{{-L}}}, and {{{-l}}} compiler/loader options for the SGI-ICE at HLRN can be found in the new example configuration file [source:palm/trunk/SCRIPTS/.mrun.config.imuk_ice2_netcdf4 .mrun.config.imuk_ice2_netcdf4]
+
+'''Attention:'''
+So far, netCDF parallel I/O has only been tested a little. Further performance optimizations may be required, especially in case of large data sets (e.g. data chunking, or lustre file system settings with {{{lfs setstripe}}}). Please report bugs and performance problems immediately!
+
+{{{combine_plot_fields}}} outputs cpu-time needed for file merging.
+
+Modules to be loaded for compilation ('''mbuild''') or job execution ('''mrun''') can/should now be given in the configuration file ({{{.mrun.config}}}) using variable {{{%modules}}}. Example:
+{{{
+   %modules     ifort/11.0.069:netcdf       lcsgih parallel
+}}}
+This method replaces the (undocumented) {{{%mpilib}}}-variable.
+
+'''WARNING:'''
+All fixed settings of modules in the scripts '''mbuild''', '''mrun''', and '''subjob''' have been removed! Please set the {{{%modules}}} variable appropriately and remove the {{{%mpilib}}} variable in your configuration file.
+
+Compilation time can be reduced be using parallel make. The number of allowed simultaneous compiler calls (threads) is set in the configuration file using variable {{{%mopt}}}, e.g.
+{{{
+   %mopts  -j:4   lcsgih parallel
+}}}
+allows a maximum of 4 routines to be compiled simultaneously (provided that the makefile does not show any dependencies between these routines).
+}}}
+|----------------
+{{{#!td style="vertical-align:top;width: 50px"
 ...
 }}}