| Version Release Date |
Add-ons |
| 4.1alpha 18- Jan-2016 User Guide |
Check page 2 of the User Guide |
| 3.1 25-Jan-2013 User Guide |
- Support of PBS/Torque besides LoadL as a Distributed Resource Manager - New maximum length of -- a user-defined name (name of a model, variable, factor, coordinate, macro, operator, …) old: 20, new: 64 -- a record of a user-defined file (edf, mdf, … file) old: 160, new: 360 -- a resolved directory path w/o file name and the resolved SimEnv workspace old: 109, new: 256 - General configuration file <model>.cfg: -- New sub-keyword out_builtin_vars for controlling storage of built-in model variables to model output files -- New sub-keywords nodata_value_byte, nodata_value_short, nodata_value_int, nodata_value_float, and nodata_value_double to define corresponding nodata values for model and post-processor output. -- Changed representativeness of nodata values for float / real*4 and double / real*8 data types. - New approach for optional result descriptors, defining result name, description, and unit - NetCDF model and post-processor output files: -- New NetCDF attribute description file <model>.ndf: This file allows for management of global and local attributes -- Default attributes changed - Structure of post-processor output files <model>.inf.[ ieee | ascii ] modified - Bug fixes |
| 3.01 10-Jul-2012 User Guide |
- Experiment description file .edf: Experiment types renamed - New experiment type: Variance based global sensitivity analysis – GSA_VB: Decompose variance at model output into linear or total effects of individual factors by applying Sobol’ method. - New experiment type: Bayesian calibration – BAY_BC: Reduce uncertainty about factor values by deriving a representative sample from factor prior distributions by taking into account measurement values from the system available for model – system intercomparison. - Experiment type GSA_EE (formerly global sensitivity): The p-level grid is now constructed from equidistant quantiles of assigned distributions for each factor. Non-dimensionality of sensitivity measures allows for comparison across different output functions - Experiment description file <model>.edf for experiment types with probabilistic / random sampling: - - New sub-keywords sample_exclude and sample_include for defining intervals to use for probabilistic factor sampling - - New probabilistic / random sampling method quasi based on Sobol’ low-discrepancy sequences - - Sampling method random renamed to pseudo, sampling method Latin hypercube renamed to stratified - - New sub-keyword seed with values [ fixed | random ] to initialize the random number generator - GAMS model interface: New sub-keyword script in <model>.gdf for performing a shell script before running each GAMS model single run - Post-processor: - - General configuration file <model>.cfg: New sub-keyword display_values for displaying additionally the first values of each result during experiment post-processing - - New advanced operators distr_par, mask_file, unmask_file, and usage - - Experiment type DFD (formerly behaviour): Experiment specific operator “behav” renamed to “dfd” - - Post-processing operators for experiment type UNC_MC are now also available for all the other experiment types - Bug fixes |
| 2.11 12-Jan-2010 User Guide |
- Regular interface for Matlab models including simenv_slice - Interface function simenv_slice now available for Python and Java models - New sub-keyword slices in <model>.cfg for applying interface function simenv_slice in user models - New sub-keywords include_runs and exclude_runs in <model>.cfg to flexibly define the set of single runs to be performed during an experiment - Behavioural analysis (<model>.edf): control checking of the sample defined in an external file (specific comb file ) by additional keys strict and nonstrict - New logfile policy for GAMS models - Sub-keyword time_limit in <model>.gdf for GAMS models again active - Bug fixes |
| 2.02 29-Jul-2009 User Guide |
- Model interface for Java - SimEnv now available under Linux for the Intel Fortran compiler ifort - New sub-keyword "factors_in_output" for <model>.cfg - Bug fixes |
| 2.01 07-Apr-2008 User Guide |
- SimEnv is now also available under Linux - SimEnvVis V0.7 with new visualization techniques, among others graphical table, scatterplot matrix and icon-based visualization - Support of experiments on multicore processor machines - Environment variable SE_HOME relocated old: /usr/local/simenv/bin new: /usr/local/simenv - New directory structure of SimEnv - [ simenv | <model> ].jcf_[ dis | par | seq ] renamed to [ simenv | <model> ]_[ dis | par | seq ]_[ aix | linux ].jcf - [ simenv | <model> ].opt_opt renamed to [ simenv | <model> ]_opt_options.txt - Bug fixes |
| 1.24 31-May-2006 User Guide |
- Model interface for ASCII files: Get in <model>.run an ASCII file with all factors and their adjusted values and transform ASCII model output directly into SimEnv model output. - Matlab and Mathematica model interfaces: Adjust the factors of a Matlab / Mathematica model and run the model by calling one dot script in <model>.run. Apply the model interface for ASCII files to put model output to SimEnv - Renamed dot scripts: py_model_ini renamed to simenv_ini_py gams_model_ini renamed to simenv_ini_gams gams_model_run renamed to simenv_run_gams gams_model_end renamed to simenv_end_gams - the following scripts are now dot scripts: simenv_ini_py, simenv_ini_gams, simenv_end_gams - Bug fixes |
| 1.23 09-Nov-2005 User Guide |
- New experiment type Global Sensitivity Analysis: Get a ranking of factors with respect to their global sensisivity to a model output by qualitative measures. - Stopping rule for Monte Carlo experiment: Define a result stopping function in <model>.edf to be used in a stopping rule to exit the experiment before the stated number of single runs is reached. - Set and export the environment variable NETCDFHOME in your file .profile if you link interfaced Fortran/C/C++ models by $SE_HOME/simenv_mod_<lng>.lnk Value for AIX: /usr/local/lib - Post-processor operator get_data now also available for NetCDF data files - For conformity reasons with the literature the term “target” is replaced by “factor” througout the whole system. In particular, replace the keyword “target” by “factor” and the sub-keyword “adjusts” by “sample” in the experiment description files <model>.edf. - According to the above new terminology argument items in the first argument of the post-processor operators behav, sens_*, lin_*, sym_* are renamed - Factor sample file <model>.edf_adj, generated during experiment preparation renamed to <model>.smp - Bug fixes |
| 1.22 23-Jun-2005 User Guide |
- Experiment preparation and performance: - - Unification and simplification of target adjustments. Target adjustment values adj_val are sampled to <model>.edf_adj as specified in <model>.edf and are modified within simenv_get_* by the target default value in a unique manner according to the target type. Special rules for Monte Carlo and local sensitivity analysis are unnecessary furtheron. - - Introduction of an additional adjustment type = relative: adjusted_target value = (1.+adj_val)*target_def_value - Experiment performance controlled by LoadL: - - Experiment single runs now can be distributed to all processors of a job class, exploiting in the course of the experiment the free proces-sors of the class in an optimal manner. - Wildcard operands for the experiment post-processor: - - Apply successively a result expression to all model output variables and/or experiment targets by using wildcard operands. - SimEnv runs now in the Bourne shell sh: Update the first line in .run and .rst accordingly - The whole operating system environment necessary for SimEnv is set within SimEnv. The dot script simenv.env is now dispensable. - - Exception (include in $HOME/.profile): export SE_HOME=/usr/local/simenv/bin (default) export PATH=$SE_HOME:$PATH (optional) - Operating system environment variables now can be used when specifying directories in user-defined files and as arguments in post-processing operators - Files renamed: - - in $SE_HOME: simenv_mod_inc.[ f | c ] to simenv_mod_[ f | c ].inc simenv_opr_inc.[ f | c ] to simenv_opr_[ f | c ].inc <model>_inc.[ f | c | py | sh ] to <model>_[ f | c | py | sh ].inc - - in the current workspace: simenv.[ jcf_par | jcf_seq | opt_opt ] to <model>.* - Bug fixes |
| 1.21 29-Mar-2005 User Guide |
- Semi-automated model interface: - - Generate automatically during experiment preparation an up-to-date model source code include file for getting experiment targets - Composed post-processing operators: - - <model>.odf now enables definition of composed operators for experiment post-processing. - <model>.odf: - - Changed declaration of argument numbers - <model>.cfg: - - Sub-keyword "out_size_threshold" replaced by "out_separation" - - Sub-keyword "distributed" replaced by "structure" - - Sub-keyword "out_ieee_blocksize" removed - <model>.[cfg|edf|gdf|mac|mdf|odf] and coordinate transformation file: - - Keyword "[cfg|edf|gdf|mac|mdf|odf|trf]" renamed to "general" - Fortran/C/C++ model interface and operator interface: - - Include files simenv_[mod|opr]_[f|c].[icl|h] renamed - GAMS model interface: - - In <model>.run insert now $SE_HOME/gams_model_run as a dot script (see Glossary) instead of calling it as a shell script - Operating system environment variables under Unix: - - Dot script simenv.env can be used to set all environment variables necessary for SimEnv or - - Prefix PATH by /usr/opt/ssh2/bin:/usr/sbin: - - PYTHONPATH re-adjusted - Bug fixes |
| 1.15 11-Jan-2005 User Guide |
- SimEnv under Linux: - - SimEnv is now available under SUSE-Linux on Intel-based platforms and compatibles - Result post-processing: - - New operators get_data and regrid - File <model>.mdf: - - Sub-keyword "var_extents" renamed to "index_extents" - Scripts <model>.run and <model>.rst: - - Include always as first line - - - #! /bin/ksh (for Unix) - - - #! /bin/bash (for Linux) - New SimEnv-internal environment variable SE_OS - Bug fixes |
| 1.14 05-Oct-2004 User Guide |
- Status of an active simulation experiment: - - Can now be acquired from any simulation experiment - User-defined operator interface: - - Now also available for C/C++ user-defined operators - - Interface function names have been changed - Model and user-defined opertor interfaces: - - Additional used symbols are documented for Fortran and C/C++ models / user-defined operators - Visual model evaluation: - - Support of block-structured grids - - Visualization technique "Parallel Coordinates" now implemented - SimEnv file wrap-up: - -Service simenv.cln deletes now files exactly according the settings in <model>.cfg - Bug fixes |
| 1.13 03-Aug-2004 User Guide |
- Experiment performance: Unused targets, undefined model output - - Targets that are not interfaced to the model (by simenv_get) and total or partial unwritten model output to SimEnv data structures (by simenv_put) are logged to <model>.mlog with a warning - Experiment performance: additional log-file <model>.nlog - - Model native terminal output is now redirected to <model>.nlog - - <model>.mlog and <model>.nlog are now organized that all information for one single run is lumped together. - Experiment performance: GAMS models - - An additional native single run related log-file <model>.nlog is stored in the sub-directory run<run_char> - Bug fixes |
| 1.12 23-Jul-2004 User Guide |
- Mandatory operating system environment variable PYTHONPATH: - - Specify it according to page 7 in the User Guide - New sub-keyword "distributed" in <model>.cfg: - - enable model output to data structures for distributed models - Post-processing: - - Modified structure of results enables now optional specification of a result description and unit - Post-processing: - - Optional direct result-by-result visualization - Post-processing: - - New arguments for operators classify, hgr, hgr_e, hgr_l - - Operator stat renamed to stat_full, new operator stat_red - - New operator move_avg for multi-dimensional moving averages - Bug fixes |
| 1.11 16-Jun-2004 User Guide |
- New experiment type Optimization: - - Minimize a cost function that is composed from model output and consequently depends on experiment targets over a bounded target space by a simulated annealing method. - Major improvements in visual result evaluation - Introduction of a result coordinate transformation file to incorporate external results from the operator get_experiment into post-processing for the current model - <model>.gdf: new sub-keyword "time_limit" for keyword "gdf" - Bug fixes |
| 1.04 14-Apr-2004 User Guide |
- Post-processor local sensitivity analysis: - - Sensitivity, linearity, and symmetry operators implemented - <model>.cfg: new sub-keyword visualization for keyword postproc - Bug fixes |
| 1.03 26-Mar-2004 User Guide |
- New sub-keyword "options" for GAMS model description file <model>.gdf - Changes in structure of result output file <model>.res<res_char>.[ieee|ascii] - - This file is now complemented by a structure description file <model>.inf<res_char>.[ieee|ascii] - Description of model interface for C++ models - Bug fixes |
| 1.02 01-Mar-2004 User Guide |
- Model runs can now be performed in any directory - New operating system environment variable SE_WD - Modified interface for GAMS models and file <model>.gdf - New sub-keyword message_level for <model>.cfg - Bug fixes |
| 1.01 03-Feb-2004 User Guide |
- First major release |
Gefördert durch:
Mitglied der:
Mitglied der: