The RAMSES Simulation Server - Terrestrial Systems Ecology
Contents
1. 16 S EITERAT RE OST 16 1 RAMSES Simulation Server RASS 1 1 Developer s Guide 1 Library Architecture base architecture RASS is described n Thoeny 1994 Table 1 Current RASS MW dependency matrix generated by HierSRC mod Note that the DM is combined into one entry and the imports of SysDep are omitted 0 ImportMWHish 26 MatBase 32 JumpTab 1 SimIntegrate 44 MWFunctions 33 MWDocProcs 2 SimDeltaCalc 40 MWRunTimeSys 10 MWVars 3 SimGraphUtils 41 MWRTCHandlers 9 MWTypes 4 SimEvents 42 MWMonitoring 18 MWErrors 5 SimObjects 43 MWDefaults 34 MWDocUtils 7 SimBase 29 MWObjects 35 6 SimMaster 11 MWSimLib 38 DM Lib 25 MWSGUBase 22 MWSimLibAuxO 12 RASSNotYet 24 Matrices 21 MWSimLBase 13 SysDep 28 MatShape 30 MWEvtBase 14 Portab 27 MatCopy 31 MWHiling jth June 12 2003 Page 1 RASS 1 1 Developer s Guide 2 Module descriptions 2 1 Portability Layer 2 1 1 SYSDEP SysDep encapsulates the interface to the local System It s implementation module has to be reimplemented on every new RASS host The interface is the following file IO CONST SDNullDevice SDStdInput SDStdOutput SDStdError VAR SDEol CHAR readOnly variable SDDirSeperator CHAR readOnly variable SDM2PathEnvVarName ARRAY 0 10 OF CHAR readOnly variable TYPE SDFile INTEGER SDFileResult SDDone SDFileNotFound SDTooManyFiles2. 0300 00 CO CO 010 NEN NEE NUNG NE NE N 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 j C R3 C 00 F3 R3 C CO CO CO R2 D LT RASS NYI contains now NotYetImplemented CreateWindow in DMWindows Since the batch DM was designed for non interactive usage it is unable to create a window but it redirects the text written to a window to the standard output file jth June 12 2003 Page 8 RASS 1 1 Users Guide EPC Version 5 Trouble shooting Symptom Possible solutions The linker doesn t find libMWLib a or The libraries are not in the linker search path Eventually libMWLib_x a ask the local system administrator to put it into an accessible place If the libraries are not in the linker search path i e local to your home directory use the environment variable LD LIBRARY PATH to tell the linker where it can find the library The linker complains about an out of Use ranlib to bring the symbol table up to date date symbol table of libMWLib a The executable simulation program The file names are case sensitive on Unix systems You doesn t find my data files may have to rename some files If your data files aren t in the start up directory specify the search path with the environment variable M2PATH A MDP file consists only of one line on You should transfer all your MDPs using F
3. FISCHLIN A 1996 Calculating temperature dependence over long time periods A comparison of methods LISCHKE H L FFLER T J FISCHLIN A 1996 Aggregation of Individual Trees and Patches in Forest Succession Models Capturing Variability with Height Structured Random Dispersions FISCHLIN A BUCHTER MATILE L AMMON K HEPPERLE E LEIFELD J amp FUHRER J 2003 Bestandesaufnahme zum Thema Senken in der Schweiz Verfasst im Auftrag des BUWAL KELLER D 2003 Introduction to the Dialog Machine 2 ed Price B editor of 2 ed Erh ltlich bei Download from http www ito umnw ethz ch SysEcol Reports html SYSTEMS ECOLOGY ETHZ INSTITUTE OF TERRESTRIAL ECOLOGY GRABENSTRASSE 3 CH 8952 SCHLIEREN ZURICH SWITZERLAND
4. 0000 000 kg patch m Grob 0 000 50000000 kg patch gj Fig 3 Typical screen here from the case study ForClim of a RAMSES post analysis session on the simulationists personal computer This session supports the interactive analysis of the uploaded simulation results by processing a stash file which RASS previously produced on the simulation server We tested PA with ten ForClim simulation runs which lasted under RASS t 145s and produced a 944 stash file For each run were documented 95 variables at 600 simulation time points Transfer of the stash file to the local personal computer required 26s such that together with the transfer of ForClim to the RASS server tc 76s Loading of the PA session required ca 4s and setting up of the workspace Fig 3 including the prepa ration of all runs within PA additional 12s yielding t4 l6s The graphical representation and tabulation of six variables from a single simulation run required ca 6s Fig 3 whereas the simultaneous inspection of the same variable from all ten runs required 27s These times increased by ca 14 96 to 7s and 31s respectively when the stash file was directly accessed via a LAN from the mass storage device of the simulation server Thus the average time to inspect one simulation run amounted to ca 30s via the LAN to only 27s The combined use of RASS and PA compares favourably with the 85s needed for a single ForClim simulation wit
5. read written are returned as the function value SDGetFileSize returns the size of file in bytes SDGetFilePos returns the current position in the file SDSetFilePos sets the current position to pos dynamic memory i e Heap PROCEDURE SDAlloc VAR a ADDRESS byteCount LONGCARD PROCEDURE SDDealloc VAR a ADDRESS PROCEDURE SDValidPtr a ADDRESS BOOLEAN SDAlloc allocates a heap block of size byteCount The heap block is aligned to Portab MaxAlign SDDealloc deallocates a previous allocated heap block SDValidPtr returns TRUE if a might be a valid address jth June 12 2003 Page 3 RASS 1 1 Developer s Guide Time CONST SDJan 1 SDFeb 2 SDMar 3 4 SDMai 5 SDJun 6 SDJul 7 SDAug 8 SDSep 2 10 SDNov 11 SDDec 12 SDSun 1 SDMon 2 SDTue 4 SDThu 5 SDFri 6 SDSat 7 hour is always within range 0 TYPE SDTime SDDate RECORD year month day hour minute second 100 dayOfWeek INTEGER END RECORD PROCEDURE SDGetTime VAR time SDTime PROCEDURE SDTimeToDate time SDTime VAR date SDDate PROCEDURE SDDateToTime date SDDate VAR time SDTime SysDep encapsulates the time in the opaque type SDTime SDGetTime returns the current system time in the parameter time SDTimeToDate SDDateToTime are routines to convert between SDTime and SDDate Environment PROCEDURE SDGetArgCount CARDINAL PROCEDURE SDGetArgument argNum
6. Ecosystem Models Case Studies of Statistical Down scaling in the Alps NEMECEK T FISCHLIN A DERRON J amp ROTH 1993 Distance and Direction of Trivial Flights of Aphids in a Potato Field PERRUCHOUD D amp FISCHLIN A 1994 The Response of the Carbon Cycle in Undisturbed Forest Ecosystems to Climate Change A Review of Plant Soil Models TH NY J 1994 Practical considerations on portable Modula 2 code TH NY J FISCHLIN A amp GYALISTRAS D 1994 Introducing RASS The RAMSES Simulation Server Out of print 22 23 24 25 26 27 28 29 30 Diese Berichte k nnen in gedruckter Form auch bei folgender Adresse zum Selbstkostenpreis bezogen werden Order any of the listed reports against printing costs and minimal handling charge from the following address GYALISTRAS D amp FISCHLIN A 1996 Derivation of climate change scenarios for mountainous ecosystems A GCM based method and the case study of Valais Switzerland L FFLER T J 1996 How To Write Fast Programs L FFLER T J FISCHLIN A LISCHKE H amp ULRICH M 1996 Benchmark Experiments on Workstations FISCHLIN A LISCHKE amp BUGMANN 1995 The Fate of Forests In a Changing Climate Model Validation and Simulation Results From the Alps LISCHKE H L FFLER T J FISCHLIN A 1996 Calculating temperature dependence over long time periods Derivation of methods LISCHKE H L FFLER T J
7. REFERENCES Anonymous 1992 EPC Modula 2 User s Reference Manual Second Edition January 1992 Edinburgh Portable Compilers Ltd Bugmann H 1994 On the Ecology of Mountainous Forests in a Changing Climate A Simulation Study Ph D thesis No 10638 Swiss Federal Institute of Technology Z rich ETHZ Bugmann H and Fischlin A 1994 Comparing the behaviour of mountainous forest succession models in a changing climate In Beniston M ed Mountain Environments in Changing Climates Routdlege London amp New York 206 221 invited refereed Cellier F E and Fischlin A 1982 Computer assisted modelling of ill defined systems In Proceeding of the 5th European Meeting on Cybernetics and Systems Research University of Vienna Austria April 8 11 1980 McGraw Hill Washington N Y 417 429 Fischlin A Gyalistras D Roth O Ulrich M Th ny J Nemecek T Bugmann H and Thommen F 1994 ModelWorks 2 2 An interactive simulations environment for work stations and personal computers Second completely revized edition Internal Report No 14 Systems Ecology Group ETH Zurich Fischlin A and Schaufelberger W 1987 Arbeitsplatzrechner im technisch naturwissenschaftlichen Hochschulunterricht Bulletin SEV VSE 78 Januar 15 21 Fischlin A 1991 Interactive Modeling and Simulation of Environmental Systems on Workstations In M ller D P F ed Analysis of Dynamic Systems in Medicine Bio
8. a MS DOS PC SDPathEnvName contains the name of the environment variable which identifies the search path for M2 files It is used by the Lookup procedure of DMFiles It is set to PATH on the Mac implementation denotes the PATH section of the User Profile and to M2PATH on the SUN implementation SDFile is the type of the file variables which can be used by the procedures from SysDep which operates on files Every file procedure of SysDep generates a result of the type SDFileResult The result of the last file operation can be requested by the function SDLastFileResult SDLookup opens a file with name fileName and access mode readOnly new and binary It returns the file identification in the VAR parameter file If readOnly is TRUE the file access is opened read only If readOnly is FALSE the file has read and write access If new is TRUE a new file will be created If a file named fileName already exists it will be truncated to length zero If new is FALSE the file must exists to complete the operation successfully SDDelete deletes the file fileName SDRename renames the file oldName to newName and SDClose closes the file which is identified by file SDFlush flushes the file buffer of the operating system SDEOF returns TRUE if the access pointer has reached the end of file SDTruncateFile truncates the file at the current file position SDFileRead Write reads writes the content of buffer with size The number of bytes actually
9. be maintained such that a program which relies on them shares the same behaviour using the Batch DM or the interactive DM The transition from RAMSES to RASS could be improved Especially the conversation of text files MDPs and data files is not yet supported by the RASS tools Only the renaming MOD2mod and DEF2def is provided DMSystem does report the ComputerSystem as unknown This is annoying since it is therefore not possible to see where a stash file is created To keep DMSystem portable it should be able to get a ComputerSystem identification string from SysDep 5 Literature Fischlin A 1986 Simplifying the usage and the programming of modern working stations with Modula 2 The Dialog Machine Internal report Project Centre IDA CELTIA Swiss Federal Institute of Technology Z rich ETHZ Z rich Switzerland Fischlin A et al 1994 ModelWorks 2 2 An Interactive Simulation Environment for Personal Computers and Workstations Internal Report 7 14 Systems Ecology ETHZ Fischlin A Mansour M A Rimvall M amp Schaufelberger W 1987 Simulation and computer aided control system design in engineering education In I Troch K P amp Breitenecker F ed Simulation of Control Systems 459 Pergamon Press Oxford a o p pp 51 60 Fischlin A amp Schaufelberger W 1987 Arbeitsplatzrechner im technisch naturwissenschaftlichen Hochschulunterricht Bulletin SEV VSE 78 15 21 Keller D 19
10. interactive components However due to the open system architecture such as the DM interface it is possible to add interactive elements Fig 2 Since most of them are potential branches in the execution thread special care has to be taken The Batch DM handles the core user interface components of the DM as follows Components without default actions are not imple mentable in the Batch DM they lead to a program abort However in most cases they can be easily replaced by the programmer with another DM function Dialogues DM dialogues entities have default values Therefore the dialogues lead to a predictable postcondition Menus The menu structure is maintained internally but no menus are shown on a screen However execution of menu commands is possible under program control Windows No windows are provided however the text written on them is directed to a standard output file 3 2 3 Data and Result Files RASS is based on the same DM library interface as the standard MW Both expect and produce files in textual form Therefore the input and output files are fully interchangeable 3 2 4 Current Implementation A simulation experiment to be solved by RASS is given by and the corresponding input data files only these have to be transferred and convertedaccording to the conventions of the specific simulation host The transfer and translation can be accomplished e g by the means of FTP File Transfer Pr
11. is a formally well defined language with some exceptions In particular the type transfer functions and the LONG type language extensions caused portability problems Therefore RASS and the Batch DM were written in a portable subset of Modula 2 There exists no standard library for Modula 2 However in order to port RASS to a different machine only the modules SysDep Portab and DMMathLib have to be re implemented SysDep and Portab consist of 593 lines of source code in the EPC Modula 2 imple mentation Since the RASS binary to text conversion for numbers is dependent of the IEEE floating point standard itruns only on machines which follow this standard 4 3 Planned Improvements We plan to implement a RASS shell with communi cation facility This would allow a user transparent transfer of MDPs and or data files Not every powerful host provides a Modula 2 compiler It is planned to apply a Modula 2 to C transla tor to RASS and MDPs in order to run simulations on these hosts Since most C compilers generate optimised code an additional gain in performance could be expected 5 CONCLUSIONS RASS is capable of solving correctly and efficiently any simulation experiment given it is defined in form of a RAMSES MDP Model Definition Program This is possible regardless of the original design for interactive usage We could demonstrate that the Dialog Machine provides a solid basis to write interactive programs such as
12. order to prevent this we replace GetExistingFile with DMFiles Lookup in the import list FROM DMFiles IMPORT Lookup TextFile GetReal SkipGap ReadChars Close Response and inside the procedure ReadAndSetParameters Lookup parFile Sensitivity DAT FALSE Now we can test our changes on RAMSES and transfer the Sensitivity MOD again to the RASS host There we have to rerun the shell script MOD2mod We don t have to rerun RASSMakeMake since the structure of the MDP hasn t changed Rerun make which builds us finally the new executable simulation program of Sensitivity jth June 12 2003 Page 7 RASS 1 1 Users Guide EPC Version An other run of Sensitivity gives us the following output which happens to be exactly the same as on RAMSES T 1 thoeng Sensitivityli2d Sensitivity 1 pmax 1 226 Ks 000 x 1 pmax 1 226 Ks 000 x 1 pmax 1 225 000 xO 2 pmax 226 Ks 000 x 2 pmax 000 2 pmax 000 d pmax 1500 00 3 pmax 1500 00 3 pmax 1500 00 1 pmax 13 000 1 pmax 13 000 1 pmax 13 000 2 pmax 245 000 2 pmax 545 000 2 545 000 3 3 3 1 1 1 2 2 2 3 3 3 E D co J cn cm eo F3 gt pmax 1500 00 pmax 1500 00 1500 00 13 000 13 000 13 000 245 000 pmax 545 000 545 000 1500 00 1500 00 pmax 1500 00 hoeny Sensitivitylizs 4
13. report the position 0 0 Some WriteLongXYZ procedures are not implemented yet jth June 12 2003 Page 7 RASS 1 1 Developer s Guide 2 2 11 DMWINDOWS amp DM2DGRAPHS All DMWindows and DM2DGraphs procedures are not implemented Later implementations should maintain the window data structures internally to support programs which rely on the internal state of these structures 2 2 12 DMEDITFIELDS amp DMENTRYFORMS All procedures are not implemented yet This gives the default behaviour 2 2 13 DMMENUS The menus data structures are maintained internally but the menus are not shown on the screen This was necessary because many clients request the internal state of a menu i e using MenuExists CommandExists IsCommandChecked to maintain the program state 2 3 MW Modules The interface is exactly the same as that of the interactive Version 2 2 For a description see the according MW literature Fischlin er al 1994 The basis was the MW 2 2 library All interactive components have been commented out As a consequence of this procedure the curve attributes are not defined and can not be written to the stash files The default attributes are written instead 1 e autoDefCStr autoDefSStr This allows to open the stash files from within the PostAnalysis session The only change in the program control is the procedure RunSimEnvironment If an experiment is installed it calls SimExperiment instead of calling RunDialogMachine If n
14. the same as on the interactive RAMSES As a result you will only get a stash file if at least one monitoriable variable is activated for filing or if you switch the filing on by calling SimBase SetDocumentRunAlwaysMode jth June 12 2003 Page 3 RASS 1 1 Users Guide EPC Version 4 Tutorial This tutorial guides you through a sample RAMSES to RASS transition The transfer using FTP is not described here The chosen sample model is Sensitivity File name Sensitivity MOD as described in Fischlin er al 1994 It demonstrates a parameter sensitivity analysis of a Michaelis Menten algae growth model It creates an additional window where it writes the chosen parameter sets At the begin Sensitivity reads its model parameters from the data file Sensitivity DAT 4 1 Developing Sensitivity MDP The model is already developed and produces the following output when ran from within RAMSES On a Macintosh after some window rearrangement 1 File Edit Settings Windows Solve Techniken 2 HE Parameter Sets SSS QiDO Jc UN 2 2 Q wu 0 JJ c c 6 h CO LJ LJ LI CO CO CO CO CO CO CO CO CO CO CO CO CO CO CO CO CO WON h WN UN h N OQ WN QN www ssssmymuyuuugggggs ogg Ww www ggg sgg gu Www ssysguymgyuugggggs sgg Www sss ggggggggWso ggg DOWN MN 00 CO FO FO WW 0 O Q0 09 00 CO F
15. you are familiar with ModelWorks Fischlin er al 1994 It is highly recommended to read also Fischlin 1991 and Thoeny et al 1994 You should also be able to work with Unix workstations RASS allows to run interactive ModelWorks Model Definition Programs MDPs in batch oriented environments like Unix workstations The generated results of a batch simulation should be the same as on the interactive RAMSES host but you have to consider the following aspects 1 The calculated results don t have to be numerical identical This is due to the different floating point hardware on different hosts Especially simulations with very small step sizes may produce detectable differences It is impossible to say on which host the quality of the results is better 2 RASS ModelWorks is build on top of the Batch DM Thoeny er al 1994 No user interactions are possible ARASS program will follow the default execution thread it simulates default answers on every interactive DM component i e pressing the enter key in every modal dialogue 2 Distribution and installation RASS consists of the libraries ibMWLib a libMWLib_x a and the shell scripts MOD2mod DEF 2def listdepf and RASSMakeMake Y ou have to install the libraries in your linker search path and the shell scripts in the command search path of your shell It is recommended to install the RASS library in usr local lib and the shell scripts in usr local bin Please ask your system admin
16. 4th completely revised ed Departement Informatik ETH Z rich Switzerland Zeigler B P 1976 Theory of modelling and simulation Wiley New York a o Zeigler B P 1979 Multilevel multiformalism modeling an ecosystem example In Theoretical Systems Ecology Academic Press New York 17 54 RASS 1 1 User s Guide EPC Version J rg Th ny T INTRODUCTION 2 DISTRIBUTION AND INSTALLATION eesnsnenssssnsnenssssssnenensasnnsnensnsnssnenenenennensnenensnssusnene 3 RUNNING MDESWITHRASS na na aaene Sal Developitig WHIPS ae raue 3 2 Transferring MDPs to the RASS 3 3 Building executable simulation programs seen 3 4 Running executable simulation programs essere tette 4 1 Developing Sensitivity MDR 4 2 Transferring the MDP Sensitivity to the RASS host sss 4 3 Building the executable simulation program Sensitivity ss 4 4 Running the executable simulation program Sensitivity ssse S TROUBLE SHOOTING a EB see 6 LITERATURE 10 1 RAMSES Simulation Server RASS 1 1 Users Guide EPC Version 1 Introduction This user s guide describes how to install and use RASS on Unix workstations using the EPC Modula 2 compiler This text assumes that
17. 89 Introduction to the Dialog Machine Bericht Nr 5 Projektzentrum IDA ETH Z rich 37 pp Thoeny J Fischlin A amp Gyalistras D 1994 RASS Towards bridging the gap between interactive and off line simulations In J Halin W K a R R ed CISS First Joint Conference of International Simulation Societies Zuerich Switzerland The Society for Computer Simulation International P O Box 17900 San Diego Cal 92177 USA p pp 99 103 jth June 12 2003 Page 16 BERICHTE DER FACHGRUPPE SYSTEM KOLOGIE SYSTEMS ECOLOGY REPORTS ETH Z RICH Nr No 1 A N 12 13 14 15 16 17 18 19 20 21 FISCHLIN A BLANKE T GYALISTRAS D BALTENSWEILER M NEMECEK T ROTH O amp ULRICH M 1991 erw und korr Aufl 1993 Unterrichtsprogramm Weltmodell2 FISCHLIN A amp ULRICH M 1990 Unterrichtsprogramm Stabilit t FISCHLIN A amp ULRICH 1990 Unterrichtsprogramm Drosophila ROTH O 1990 Maisreife das Konzept der physiologischen Zeit FISCHLIN A ROTH O BLANKE T BUGMANN H GYALISTRAS D amp THOMMEN F 1990 Fallstudie interdisziplin re Modellierung eines terrestrischen Okosystems unter Einfluss des Treibhauseffektes FISCHLIN A 1990 On Daisyworlds The Reconstruction of a Model on the Gaia Hypothesis GYALISTRAS D 1990 Implementing a One Dimensional Energy Balance Climatic Model on a Microcomputer out of print FISCHLIN A amp R
18. BALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all cd DMLibDev OptHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all cd DMLibDev CoreHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all cd AuxLibDev make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all cd MWLibDev Base MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all cd MWLibDev High MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME all clean cd SysDep make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME clean DMLibDev OptHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME clean cd DMLibDev CoreHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME S RASSHOME clean cd AuxLibDev make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME clean cd MWLibDev Base MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME clean cd MWLibDev High MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME clean rm f o a depend cd SysDep make M2GLOBALFLAGS S M2GLOBALFLAGS RASSHOME RASSHOME depend cd DMLibDev OptHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME S RASSHOME depend cd DMLibDev CoreHigh MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME depend cd AuxLibDev make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME S RASSHOME depend cd MWLibDev Base MOD make M2G
19. CARDINAL VAR argument ARRAY OF CHAR PROCEDURE SDGetEnvVar varName ARRAY OF CHAR VAR varValue ARRAY OF CHAR PROCEDURE SDGetEnvVarListElem varName ARRAY OF CHAR index INTEGER VAR elemValue ARRAY OF CHAR Program arguments are only supported on systems which supports them SDGetArgCount returns the number of arguments A specific argument can be retrieved by SDGetArgument The value of an environment variable can be accessed by SDGetEnvVar If the environment variable with name varName doesn t exists a string with length zero is returned On the Mac implementation SDGetEnvVar returns the content of the section varName of the User Profile Since an environment variable can contain a list i e the PATH and the divider of list elements is system dependent a procedure SDGetEnvVarListElem is provided jth June 12 2003 Page 4 RASS 1 1 Developer s Guide Termination PROCEDURE SDSetTermProc p PROC VAR done BOOLEAN PROCEDURE SDExit status INTEGER PROCEDURE SDExitMsg status INTEGER module procedure reason ARRAY OF CHAR SDSetTermProc allows to add cleanup procedures On termination of a program they are called in the reverse order of their installation SDExit terminates a program with the status code status SDExitMsg writes also a termination message to SDStdError Interprocessing TYPE SDLocalMsgID INTEGER SDLocalMsgProc PROCEDURE ADDRESS PROCEDURE SDSetLocalMsgHand
20. Conference of International Simulation Societies Z rich Switzerland The Society for Computer Simulation International Box 17900 San Diego Cal 92177 USA 99 103 Adressen der Autoren Addresses of the authors J Th ny Dr A Fischlin Gyalistras System kologie ETH Z rich Institut f r Terrestrische Okologie Grabenstrasse 3 CH 8952 Schlieren Z rich SWITZERLAND e mail sysecol ito umnw ethz ch 1993 System kologie ETH Z rich Introducing RASS The RAMSES Simulation Server Jiirg Thony with contributions from A Fischlin and D Gyalistras RASS forms a new component of the RAMSES modelling and simulation concept It allows to automatically translate interactive RAMSES model definition programs and to execute them in a batch mode on a high performance computer This report describes RASS and its implementation RASS 1 1 It is a collection of three documents 1 RASS Towards bridging the gap between interactive and off line simulations 2 RASS 1 1 User s Guide 3 RASS 1 1 Developer s Guide The first document is an introduction to the general ideas and concepts behind RASS and was presented in 1994 at the CISS First Joint Conference of International Simulation Societies Zuerich The user s guide describes how to install and use RASS on Unix workstations It also contains a tutorial guiding the users through a sample transition from RAMSES to RASS The developer s guide is a descrip
21. ES and RASS Modula 2 lib OS HW All RAMSES software is implemented on top of the DM using the language Modula 2 The DM is a library that allows to program interactive applications indepen dently from the underlying graphical user interface All interactive DM commands are designed as a program state transition with a known precondition anda defined post condition The DM allows the transition only if the pre condition is true and the postcondition can be fulfilled otherwise the interactive DM modally demands a user intervention usually with a default answer to the request Since RASS is not interactive a new DM implemen tation Batch DM was necessary The Batch DM has the same programmatic interface but has no visible user in terface An optional terminal like I O is provided The various DM components are maintained internally such that the state transitions are the same as in the interactive DM The Batch DM is built solely on top of the two modules SysDep and Portab The only exception is the DMMathLib SysDep and Portab hide the local Modula 2 library and the OS hardware To prevent performance loss we implemented the DMMathLib with direct calls to the machine dependent math library instead of encapsulating these functions in SysDep 3 2 2 Running MDPs with RASS Essential is that any MDP has the same execution thread regardless whether executed interactively by MW or by RASS A RAMSES MDP has very few
22. LOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME depend cd MWLibDev High MOD make M2GLOBALFLAGS M2GLOBALFLAGS RASSHOME RASSHOME depend The rules all clean and depend call the corresponding Makefiles rules in the RASS hierarchy jth June 12 2003 Page 12 RASS 1 1 Developer s Guide echo n gt __ ALIASES rm f libMWLib a sh c mkdir tmplib cd tmplib for f in lib x a lib x a do ar x rm f N ar cur libMWLib a rm f done cd rmdir tmplib ranlib libMWLib a mwlib x echo n gt __ ALIASES rm f libMWLib x a sh c mkdir tmplib cd tmplib for f in lib x a lib x a do ar x f if f __ ALIASES then cat s ALIASES gt gt ALIASES fi rm f N ar cur libMWLib x a rm f done cd rmdir tmplib ar ur libMWLib x a ALIASES ranlib libMWLib x a The rules mwlib and mwlib x create the RASS MW object library resp definition library which contain all libraries of RASS jth June 12 2003 Page 13 RASS 1 1 Developer s Guide 3 3 Tools The rule depend of the RASS Makefiles call a shell script named listdepf bin csh foreach file mod set obj basename file mod o em2 twy M2FLAGS file sed is found in d sed s is found in Sobj 1 N unig end listdepf uses the EPC Modula 2 compiler to get a list of dependencies for all mod fi
23. MWDocUtils MWObjects MWSimLibAuxl MWSimLibAl MWSimLibAO MWErrors MWRTCHandlers MWSimLibAl MWRunTimeSys MWRunTimeS MWEvtBase MWRunTimeSys MWTypes MWFiling MWRTCHandlers MWRTCHand MWSGUBase MWVars MWMonitoring MWMonitori MWHigh SimBase SimGraphUtils SimObjects SimGraphUtils SimGraphU SimDeltaCalc SimIntegrate SimEvents SimIntegrate SimInte SimMaster SimDeltaCalc SimDeltaC Table 2 All RASS libraries with the corresponding modules and the aliases The library files are named libXXX a and libXXX x b where XXX is the name of the library The libXXX a files contain all object files and the files all definition modules and the __ ALIASES file All RASS libraries are combined into ibMWLib a and libMWLib These two files should be installed in usr local lib This location is in the d search path and can be accessed by anybody on our SUN Network Note that ranlib has to be called every time the libraries are copied jth June 12 2003 Page 10 RASS 1 1 Developer s Guide 3 2 Makefiles The Makefiles are structured in the same fashion as the RASS directories The main Makefile is in the root directory of the RASS development environment It is not possible to use any Makefile in the sub directories directly they are to be called from within the main Makefile All Makefiles contain three main rules 1 all this is the first rule and therefore also the standard rule It has the library of the directory where the Makefile
24. Make in the directory where your MDP resides It will generate a file named Makefile The Unix tool make uses this file to perform the compilation and linking for you Simply call make in the directory where your MDP and the generated Makefile reside Note that RASSMakeMake assumes that the first alphabetically module without corresponding definition module is the main module It is possible that you get some compiler error messages The reason is usually that Modula 2 is not defined exactly the same on various compiler implementations Please read the report Practical considerations on writing portable Modula 2 code Thoeny 1994 to get more information about writing portable Modula 2 code jth June 12 2003 Page 2 RASS 1 1 Users Guide EPC Version If you change the source code you have to call make again in order to bring your executable simulation program up do date You can use make with three optional parameters all clean and depend make all is the same as make without parameter make clean will remove all object files the result of the compilation process make depend will add the inter module dependencies to the Makefile This is useful if you plan to change your definition modules on the RASS host 3 4 Running executable simulation programs The executable simulation program has the same name as the main module of your MDP It can be started by typing its name If you have stored your data files in another location as t
25. O FO FO ENN NN NN NNN nn nn 21 22 23 24 25 26 2 jth June 12 2003 Page 4 RASS 1 1 Users Guide EPC Version 4 2 Transferring the MDP Sensitivity to the RASS host Create a directory named Sensitivity in your home directory on the RASS host mkdir Sensitivity Transfer Sensitivity MOD and Sensitivity DAT from your RAMSES host to the RASS host into this directory using FTP On the RASS host change to the directory Sensitivity check for the presence of the MDP and data file and call the shell script MOD2mod The output on the terminal should look like baikal thoenugDthoenu143 fed Sensitivity baikal userzy thoeng baikal thoenugISenzi tivi tij 150 Sensitivi tij DHT Sensi tivi tu baikal thoenugISenszi tivi tu191 HOD z2mad baikal thoenugISenszi tivi tu 1952 115 Senzitivity DAT Sensitivity mad baikal thoengI Sensitivi tu153 J You don t have to call DEF2def because the Sensitivity MDP contains no definition modules jth June 12 2003 Page 5 RASS 1 1 Users Guide EPC Version 4 3 Building the executable simulation program Sensitivity The shell script RASSMakeMake generates the Makefile which you can subsequently use to build Sensitivity In the directory Sensitivity use the commands RASSMakeMake make The output on the terminal should look like thoeny Sensi tivi ti 137 AASSMakeMake RASSMakeMake 1 0 jth 6 94 Generating Makefile The default make rule
26. OTH GYALISTRAS D ULRICH M UND NEMECEK 1990 ModelWorks An Interactive Simulation Environment for Personal Computers and Workstations out of print for new edition see title 14 FISCHLIN A 1990 Interactive Modeling and Simulation of Environmental Systems on Workstations ROTH O DERRON J FISCHLIN A NEMECEK T amp ULRICH M 1992 Implementation and Parameter Adaptation of a Potato Crop Simulation Model Combined with a Soil Water Subsystem NEMECEK T FISCHLIN A ROTH amp DERRON J 1993 Quantifying Behaviour Sequences of Winged Aphids on Potato Plants for Virus Epidemic Models FISCHLIN A 1991 Modellierung und Computersimulationen in den Umweltnaturwissen schaften FISCHLIN A amp BUGMANN H 1992 Think Globally Act Locally A Small Country Case Study in Reducing Net CO Emissions by Carbon Fixation Policies FISCHLIN GYALISTRAS D ROTH O ULRICH M TH NY J NEMECER T BUGMANN amp THOMMEN F 1994 ModelWorks 2 2 An Interactive Simulation Environment for Personal Computers and Workstations FISCHLIN A BUGMANN amp GYALISTRAS D 1992 Sensitivity of a Forest Ecosystem Model to Climate Parametrization Schemes FISCHLIN A amp BUGMANN H 1993 Comparing the Behaviour of Mountainous Forest Succession Models in a Changing Climate GYALISTRAS D STORCH H v FISCHLIN A BENISTON M 1994 Linking GCM Simulated Climatic Changes to
27. RASS implements the simulation session in a different way RAMSES Interactive modeling amp simulation MW RASS Batch simulation Interactive Post Analysis Fig 1 State transitions while developing and solving models using RASS within RAMSES A typical modeling simulation cycle Fig 1 involves the following steps 1 Develop a model interactively in the RAMSES simulation and modeling session and run local simulations interactively 2 Produce model behaviour with RASS in a batch mode 3 Analyse the simulation results interactively in the PA session 3 2 Design and Implementation RASS was designed to fulfil the following main requirements RASS has to achieve a high portability in two ways First it should be easily portable Second any MDP including its corresponding input output data files should be exchangeable between the batch and interactive simu lation environment without any changes The simulation results of RASS have to be reliable i e the batch processing must not hamper the validity of the simulation results The transition from the interactive RAMSES to the batch oriented RASS should be smooth and require a minimum of user interactions 3 2 1 RASS Architecture RASS resides on top of several software layers which enhance portability and software maintenance Fig 2 RAMSES MDP RASS MW Batch DM SysDep Portab amp Math Modula 2 lib OS HW Fig 2 Architecture of RAMS
28. SDNoWriteAccess SDNoReadAccess SDFileNameWrong SDOtherResult PROCEDURE SDLookup VAR file SDFile fileName ARRAY OF CHAR readOnly new binary BOOLEAN PROCEDURE SDDelete fileName ARRAY OF CHAR PROCEDURE SDRename oldName newName ARRAY OF CHAR PROCEDURE SDClose VAR file SDFile PROCEDURE SDFlush file SDFile PROCEDURE SDEOF file SDFile BOOLEAN PROCEDURE SDTruncateFile file SDFile PROCEDURE SDFileRead file SDFile buffer ADDRESS size LONGCARD LONGCARD PROCEDURE SDFileWrite file SDFile buffer ADDRESS size LONGCARD LONGCARD PROCEDURE SDGetFileSize file SDFile LONGCARD PROCEDURE SDGetFilePos file SDFile LONGCARD PROCEDURE SDSetFilePos file SDFile pos LONGCARD PROCEDURE SDLastFileResult SDFileResult Three standard files are defined SDStdInput SDStdOutput and SDStdError A undefined file is identified by SDNullDevice Every operation on the SDNullDevice file has no effect at all jth June 12 2003 Page 2 RASS 1 1 Developer s Guide An end of line in a text file is defined by SDEol on interface level SysDep will translate them to the local usage i e line feed on Unix carriage return on a Mac or carriage return line feed on 5 DOS PC The client of SysDep has to use SDEol in order to get the desired result SDDirSeperator exports the character which is used by the operating system to separate directory levels i e on a Mac on Unix or on
29. SYSTEM KOLOGIE ETHZ SYSTEMS ECOLOGY ETHZ Bericht Report Nr 21 Introducing RASS The RAMSES Simulation Server J Th ny A Fischlin D Gyalistras January 1995 Eidgen ssische Technische Hochschule Z rich ETHZ Swiss Federal Institute of Technology Zurich Departement f r Umweltnaturwissenschaften Department of Environmental Sciences Institut f r Terrestrische kologie Institute of Terrestrial Ecology The System Ecology Reports consist of preprints and technical reports Preprints are ar ticles which have been submitted to scientific journals and are hereby made available to interested readers before actual publication The technical reports allow for an exhaustive documentation of important research and development results Die Berichte der System kologie sind entweder Vorabdrucke oder technische Berichte Die Vorabdrucke sind Artikel welche bei einer wissenschaftlichen Zeitschrift zur Publi kation eingereicht worden sind zu einem m glichst fr hen Zeitpunkt sollen damit diese Arbeiten interessierten LeserInnen besser zug nglich gemacht werden Die technischen Berichte dokumentieren ersch pfend Forschungs und Entwicklungsresultate von allge meinem Interesse The first part of this report has been published Please cite as Th ny J Fischlin A amp Gyalistras D 1994 RASS Towards bridging the gap between interactive and off line simulations In J Halin W K a R R ed 55 First Joint
30. TP in text my Unix workstation ASCH mode I m using Fetch as the FTP Client on my RAMSES on Macintosh computers sets the type of the Macintosh computer and it doesn t allow MDPs to MoTx Fetch currently only allows the transfer me to transfer my RAMSES MDPs in of files with the type in text mode Use an other text mode FTP client e g Xferlt sfalken apple com It is also possible to set the preferences of the RAMSES shell Modeling session programming session MiniShell to export mode Thereafter the RAMSES shell will set the type of the touched MDPs to jth June 12 2003 Page 9 RASS 1 1 Users Guide EPC Version 6 Literature Fischlin A 1991 Interactive modeling and simulation of environmental systems on workstations In Richter D P F M a O eds Dynamic Systems in Medicine Biology and Ecology Berlin Springer 131 145 Fischlin A er al 1994 ModelWorks 2 2 An Interactive Simulation Environment for Personal Computers and Workstations Internal Report 14 Systems Ecology ETHZ Thoeny J 1994 Practical considerations on portable Modula 2 code Internal Report 20 Systems Ecology Group ETHZ Thoeny J Fischlin A amp Gyalistras D 1994 RASS Towards bridging the gap between interactive and off line simulations In J Halin W K a R R ed CISS First Joint Conference of International Simulation Societies Zuerich Switzerland The Society for Computer Simula
31. X TEXT_STOP echo n OBMS 5 echo echo echo all mainmod echo echo mainmod OBMS Makefile echo em2 M2LFLAGS o mainmod OBMS iSM2 ltermcap echo sed s X lt lt TEXT_STOP 1 rm f o a x x Q D D 3 Q csh c setenv M2FLAGS M2CFLAGS listdepf gt makedep echo DEPENDS a gt eddep echo gt gt eddep echo gt gt eddep echo d gt gt eddep echo r makedep gt gt eddep echo w gt gt eddep ed Makefile lt eddep rm f eddep makedep X X X X X X X X X X X DEPENDS dont remove this line make depends needs it TEXT_STOP gt Makefile tell the user what we have done echo echo echo The default make rule is all echo It has the main model mainmod as it s target echo echo The rule clean echo removes all object files echo echo The rule depend echo adds the inter module dependencies to the generated Makefile jth June 12 2003 Page 15 RASS 1 1 Developer s Guide 4 Future development of RASS The directory contains some prototypes of MDPs which run on distributed systems The library is necessary to use them installed on baikal ethz ch obtained from netlib2 cs utk edu The AuxLib should be completed such that at minimum all sample models of the RAMSES distribution are supported The internal windows data structures of DMWindow should
32. a RAMSES MDPs since they could be executed in a batch mode with only few insignificant restrictions Although the correctness of any MDP solved by RASS can not be proven at least for the case studies RASS produced correct results Thus the uploading from an interactively developed RAMSES MDP to a RASS batch simulation server is possible and can even be implemented in a smooth user transparent manner In the tested real world case studies we obtained substantial average gains in performance e g between 110096 and 7 200 for the simulationist s turnaround time Table 1 Therefore RASS allows to profit in two ways First it allows to freely engage in interactive development of complex simulation models on widely available PCs or work stations using the interactive RAMSES software Second at later stages 1 6 when complex well defined simulation experiments are of prime interest the RAMSES MDP can be easily trans ferred to more powerful machines e g a super computer running the RASS simulation server Since RASS is highly portable and can be imple mented easily on any host computer RASS provides a user friendly simulation environment It allows smooth transitions a from interactive work station based to simulation server based off line simulations b from the server back to simulationist s work station for an inter active exploration and analysis of the simulation results under the RAMSES post analysis PA session
33. and generates simu lation results which can be again explored interactively by the Post Analysis component of RAMSES First we present the architecture of RASS and discuss the concep tual problems which had to be solved Three case studies serve to demonstrate the obtainable performance gain and the quality of the simulation results Finally portability issues and future enhancements are discussed 2 MATERIAL The current implementation of RAMSES covers inter active MDP development and simulation by several components such as the standard DM Dialog Machine Fischlin and Schaufelberger 1987 Fischlin er al 1987 the standard MW ModelWorks and the PA session Post simulation Analysis RASS was implemented in Modula 2 using MacMETH Wirth et al 1992 on Macintosh computers and EPC Modula 2 Anonymous 1992 on SUN worksta tions It uses a new batch oriented implementation of the DM Batch Dialog Machine Three MDPs ForClim Diversity and NumInt were used as case studies All simulation experiments were performed either on an Apple Macintosh Quadra 700 for RAMSES or on a SUN S10 for RASS On the Macintosh we measured the time inside the MDPs with no other applications running during simulation On the SUN we measured the simulation experiment with the Unix time command during a minimal work load 3 RESULTS 3 1 Simulation with RASS A RASS task is conceptually a RAMSES simulation session Fischlin 1991 However
34. arch project a simulation software must support both interactive and batch simulations The RAMSES Research Aids for the Modelling and Simulation of Environmental Systems software Fischlin 1991 was designed to support interactive modular mo deling and simulation of ill defined systems with one or any combination of the classical model formalisms SQM SeQuential Machine DESS Differential Equation System Specification DEVS Discrete EVent System Specification Zeigler 1976 1979 Wymore 1984 In RAMSES any model implementation is made in form of a so called MDP Model Definition Program Fischlin et al 1994 A MDP represents an interactive program and therefore shares the common problems of all sophisticated interactive software These are the limited portability and the computational overhead of the user interface Thus the following questions arise Is it possible to reuse an interactive simulation program such as a RAMSES MDP for non interactive batch calculations without any changes on the source level Which problems need to be solved to correctly MDPs within a batch simulation environment How valid are the simulation results Which gain in performance can be achieved thanks to reduced graphics overhead and more powerful machines The here presented solution RASS RAMSES Simulation Server forms a new component of RAMSES RASS receives interactive MDPs in source form runs them off line on a simulation server
35. e local message handler zero from SysDep Therefore RASS could be easily extended by adding a handling of DM Events 2 2 4 DMCLOCK Only the procedures Now and Today are implemented 2 2 5 DMMATHLIB The DMMathLib is implemented with direct calls to the C Math library 2 2 6 DMMESSAGES The messages are written to the SDStdOutput Since the writing is done using a file variable it is easy to change it to another device Abort writes the three paragraph parameters to SDStdOutput and calls SDExit with exit status one 2 2 7 DMSTORAGE DMStorage is fully implemented but the levels are ignored 2 2 8 DMSTRINGS All procedures which are string resource operations are not implemented 2 2 9 DMSYSTEM The DMLevel is forced to the DMSystem startupLevel All ScreenSize and color related procedures return zero The ComputerSystem Keyboard and CPU are reported as unknown FPU is FALSE The SystemVersion and RomVersion are reported as unknown InitProcs are not supported The installed TermProcs are called only once at program termination 2 2 10 DMWINDOWIO write operations are redirected to SDStdOutput Since the writing is done using a file variable it is easy to change it to another device SetPos tries to achieve the desired position in X coordinates If the desired column is smaller than the current X coordinate then a new line is inserted All graphical output related procedures are not supported All mouse position procedures
36. gt Ids i 7 Modenvorks DAT Macintosh HD RASS Fort im Monitorable variables Simulation tars ES af ForClim model 2 4 Bern Ferch Monitorable variable names ldent US State ES Post simulation analysis 25 fidel Parameters S f f mvi Abies alba bim aba kd a ForClim E Abiotic environme Fort f mS mv Picea excelsa BiorPexce ne Siste ble fi mz 10 Acer pseudoplatan Apzeu fede Aaraneters t el f ms mv22 Quercus 1 ke Simulation results ELT Graph 48 simutation resutts m tf Modetvorks DAT Macintosh HD RASSY m ForClim model 2 4 Bern tho Monrtorabfe tar cab tes 4 o 8 m ForClim E Abiotic environme ho Hons torable tar cab tes 3 ForClim P Plant dynamics m 0 61 Abies alba Eiom aid Larix decidua Biom ox mu Picea excelsa Biom aus Pinus cembra Biom AS Pinus montana Biom oad aus Pinus silvestris Biom 4 Taxus baceata Biom A Acer campestre Biom 4 iri E Table 0 0 250 0 500 0 7500 10000 ime Time Aalba Fexoe Aj i 1140 0000000 1490180111 895571350 EEE 1160 0000000 48 95514407 0 00000000 0 Pexco 0 000 50000 000 karpatch 1180 0000000 157 66191101 000000000 o pseu 0 000 50000 000 kg patch 1200 0000000 0 00000000 0 00000000 2 Opete 0 000 0000 000 kg patch Fsilu 0 000
37. h the interactive MW on the simulationist s computer We concluded that the RAMSES PA not only allows to efficiently and flexibly analyse batch simulation results but that it may even be of interest for inspecting model behaviours interactively during a model development phase 4 DISCUSSION 4 1 Domain of RASS From the perspective of the simulationist the break even point for RASS depends on the available computer infrastructure The easier to use and the faster the connec tion between the interactive and the RASS host is the smaller simulation experiments may get to be still suitable The actual break even point is given by tt RAMSES RASS tc ts RASS gt ts RAMSES ts RASS lc A gain in performance occurs if tc lt ts RAMSES ts RASS A typical for a small model is less than 3 minutes The average performance gain for t is a factor 12 to 82 Given a model with t 3 minutes and a modest formance factor of tg 12 the break even point for a single simulation run is reached for ts ramses 196s Note if a simulation experiment is executed within a loop the smaller tc becomes and therefore the sooner the break point may be reached For ForClim this was found to be already the case if a structured simulation experiment consists at least of two runs See 3 3 2 Post Analysis 4 2 Portability In order to make Modula 2 code portable special care had to be taken Modula 2
38. he directory where you start your executable simulation program you have to specify the path in the environment variable M2PATH Consult the documentation of your favoured shell on how to set environment variables The syntax of M2PATH is path path where path stands for an optional repetition of path The pathes can be given relative to the start directory or to the root directory Examples using csh setenv M2PATH Datafiles setenv M2PATH Datafiles 1 Datafiles2 setenv M2PATH home users goofy RASS Models My Model Datafiles If you have used DM library routines which are not yet implemented in the Batch DM a file named RASS NYI will be created RASS NYI contains a list of all calls to those routines Each line contains one entry An entry can have two forms NotYetImplemented RoutineName in ModuleName FATAL NotYetImplemented RoutineName in ModuleName RASS NYI contains any number of entries of the first form If an entry of the second form exists the program was aborted inside the listed routine Note that the file RASS NYI will only be generated if you have called at least one of the routines which are not yet implemented After execution of a simulation program you should check the existence of RASS NYT If it exists examine its content In order to prevent the examination of an old RASS NYI it is recommended to remove or rename it prior to the next execution of a RASS program The filing of the MW library on RASS behaves
39. is all t has the main model Sensitivity as it s target The rule clean removes all abject files The rule depend adds the inter module dependencies ty generated Makefile thoenylSensitivi tij 138 make 2 sun 0 00 iMHLib c 0 Sensitivi Sensitivity mod EPC Maodula z Version 2 0 2 Copyright 1990 1991 1992 EPCL All Rights Reserved 212 lines compiled em sun 0 00 iMHLib o Sensitivity Sensitivity a iSh2 termcap thoeny Sensi tivi tij 138 jth June 12 2003 Page 6 RASS 1 1 Users Guide EPC Version 4 4 Running the executable simulation program Sensitivity Now you can run Sensitivity which will be finished very fast Then analyse the RASS NYI baikal thoeny Sensitivity 101 baikal thoenylSensitivityl102 leat RASS NYI FATAL NotYetimplemented GetExistingFile in OMFi les baikal thoenylSensitivityl103 B Whoops This transition went wrong The best thing at this point is to return to your RAMSES host and look at the MDP You will find GetExistingFile twice first in the import list FROM DMFiles IMPORT GetExistingFile TextFile GetReal SkipGap ReadChars Close Response and secondly inside the procedure ReadAndSetParameters GetExistingFile parFile Open parameter file Sensitivity DAT As you can see the call to GetExistingFile has no parameter for a default file name Therefore RASS was unable to continue and terminated your program In
40. istrator to install these files If the installation in usr local isn t possible you can also install RASS local to your home directory After copying the file ibMWLib a to the desired location you have to run ranlib PATH OF LIB IibMWLib a PATH_OF_LIB libMWLib_x a To allow the linker to find the library you have to specify the location using the LD LIBRARY PATH environment variable Please make sure that the RASS shell scripts are in the command search path of your shell consult the documentation of your favoured shell The EPC Modula 2 environment has to be installed on the host prior using RASS This has to be done by your system administrator jth June 12 2003 Page 1 RASS 1 1 Users Guide EPC Version 3 Running MDPs with RASS A simulation program is usually developed in the interactive RAMSES environment Then the Modula 2 source code and the input files are transferred to the RASS host There you can build an executable simulation program to finally run your simulation experiments 3 1 Developing MDPs Usually aMDP is developed on an interactive RAMSES host where you should also plan and test your simulation experiment Beside the default strategy of the Batch DM default answers to dialogues RASS will simulate exactly one user event for you If you have an experiment installed in your MDP RASS will execute it otherwise RASS will execute one simulation run Therefore you can test your simulation program on the in
41. ler msgID SDLocalMsgID handler SDLocalMsgProc VAR done BOOLEAN PROCEDURE SDGetLocalMsgHandler msgID SDLocalMsgID VAR handler SDLocalMsgProc PROCEDURE SDSendLocalMessage msgID SDLocalMsgID msg ADDRESS channels are not designed yet Idea a channel is a connection between two host consisting of two two way streams command stream text data stream binary data with translation facility A message handler is a procedure with an address parameter Many message handlers can be installed by SDSetLocalMsgHandler SDGetLocalMsgHandler retrieves the message handler with the identification msgID A message handler can be triggered to send the message msg by calling SDSendLocalMessage This mechanism can be used to implement the DialogMachineTask procedure e g calling GetNextEvent in the Mac implementation to enable the co operative multitasking jth June 12 2003 Page 5 RASS 1 1 Developer s Guide 2 1 2 PORTAB Portab exports a set of base types which differ among Modula 2 implementations This supports the programming of portable Modula 2 code Main clients of Portab are DMConversions and SysDep SHORTCARD SUN EPC Modula CARDINAL SUN EPC Modula CARDINAL SUN EPC Modula SHORTINT SUN EPC Modula INTEGER SUN EPC Modula CARD16 CARDINAL MacMETH CARD32 LONGCARD MacMETH CARDADR LONGCARD MacMETH INT16 INTEGER MacMETH INT32 LONGINT MacMETH MaxAlign INTEGER the biggest align
42. les in the current directory If RASS has to be ported to another Modula 2 compiler listdepf has to be rewritten The Modula 2 program GetImports mod also in the tools directory could be used whithin a new version of listdepf The shell scripts MOD2mod and DEF 2def all MOD files in the current directory to mod resp DEF to def RASSMakeMake is a tool which helps the user of RASS to generate a skeleton Makefile for his MDPs Only one MDP is supported per directory bin sh echo RASSMakeMake v1 1 jth 12 94 we don t overwrite an existing Makefile echo if f Makefile then echo The Makefile alredy exist Please remove or rename it first exit fi generate a list of the object files in OBMS and assume that the first module without a definition module is the main module and therefore the target OBMS mainmod basename 51 mod for f in mod do OBMS SOBMS basename f mod o if f basename f mod def then if mainmod then mainmod basename f mod fi fi done export OBMS export mainmod echo Generating Makefile put the content of the Makefile to stdout and redirect it into a file jth June 12 2003 Page 14 RASS 1 1 Developer s Guide called Makefile why not sed s X lt lt TEXT STOP X XM2CFLAGS sun D0 iMWLib XM2LFLAGS sun D0 1MWLib X X SUFFIXES mod X X mod o X em2 M2CFLAGS c o mod
43. logy and Ecology Informatik Fachberichte 275 Springer Berlin a o 131 145 Fischlin A Mansour M A Rimvall and Schaufelberger 1987 Simulation and computer aided control system design in engineering education In Troch I Kopacek P amp Breitenecker F eds Simulation of Control Systems Pergamon Press Oxford 51 0 Fischlin A and Ulrich M 1987 Interaktive Simulation schlecht definierter Systeme auf modernen Arbeitsplatzrechnern die Modula 2 Simulationssoftware ModelWorks In Proceedings of Simulation in Biologie und Medizin February 27 28 1987 Vieweg Braunschweig 1 8 Kreutzer W 1986 System simulation programming styles and languages Sydney a o Addison Wesley Vancso K Fischlin A and Schaufelberger 1987 Die Entwicklung interaktiver Modellierungs und Simulationssoftware mit Modula 2 In Halin J ed Simulationstechnik Informatik Fachberichte 150 Springer Berlin 239 249 Vancso Polacsek K 1990 Theory and practice of computer assisted simulation and modeling on professional workstations Ph D thesis No 9104 Swiss Federal Institute of Technology Z rich ETHZ Wymore A W 1984 Theory of Systems In Handbook of Software Engineering Van Nostrand Reinhold Company New York Wirth N Gutknecht J Heiz W Sch r H Seiler H Vetterli C and Fischlin A 1992 MacMETH A fast Modula 2 language system for the Apple Macintosh User Manual
44. ment for the base types Typically the definition module of Portab has to be redefined for every new RASS implementation I recommend to define the whole type set anew and to keep the other sets for documentary purpose as comment MaxAlign is the maximum alignment boundary of all the base types i e LONGINT LONGREAL MaxAlign is calculated by standard Modula 2 code and is not guaranteed to be correct Please check its actual value when you encounter strange behaviour with dynamic data i e wrong calculations 2 2 DM Modules For a description of the DM interface see the according DM literature Fischlin 1986 Fischlin er al 1987 Fischlin amp Schaufelberger 1987 Keller 1989 This documentation describes only the special Batch DM implementation details The concept of the Batch DM is described in Thoeny et al 1994 2 2 1 DMCONVERSIONS The module DMConversions is implemented using only standard Modula 2 code no calls to system software are made It is therefore fully portable However its efficiency could be improved 2 2 2 DMFILES DMFiles uses SysDep SDPathEnvVarName as the search path for Lookup The single path components are extracted using SysDep SDGetEnvVarListElem This provides a portability between various RASS platforms jth June 12 2003 Page 6 RASS 1 1 Developer s Guide 2 2 3 DMMASTER Most of the procedures and functions of DMMaster are not yet implemented The procedure DialogMachineTask calls th
45. mentioned hierarchy The directory tools contains shell scripts which are of use for the user of RASS The directory Models contains some sample MDPs which are included for testing purposes 3 1 Libraries The Unix tool ar is used to create the libraries Since ar is restricted to filenames with a maximum of 15 characters including and extension special care has to be taken for files with longer names All editing is done on the files with the original name length They are copied to files with names of 15 characters prior compilation or library archiving To enable the EPC compiler to find the real names a special file named __ ALIASES has to be included in an archive This file contains aliases for module names in the form ModuleName ShortName cfg Table 2 jth June 12 2003 Page 9 RASS 1 1 Developer s Guide Library Modules Aliases SysDep SysDep Portab RASSNotYet DMCore DMFiles DMConversions DMMaster DMStrings DMConversions DMConvers DMSystem DMWindowIO DMWindows DMClock DMStorage DMMessages DMMathLib DMOpt DM2DGraphs DMEntryForms DMEditFields DMMenus DMEditFields DMEditF DMEntryForms DMEntryF AuxLib TabFunc TFTypesA TFBase TFDocProcs TFEdit TFTypesAndVars TFTypesA TFFuncs TFMenus Matrices JumpTab Histograms StochStat MathProcs RandGen Lists MultiNormal Jacobi RandNormal ReadData MatBase MatCopy MatShape RandGenO MWBase MWDefaults MWFunctions MWSimLBase MWDocProcs MWSimLibAux0 MWSimLibA0 MWMonitoring MWSimLib
46. o experiment is installed it calls SimRun 2 4 AuxLib Modules The core of the AuxLib modules were ported and included in the RASS library The Matrix package is only available in a old version O Roth s version 1991 The new af version is right now not portable from MacMETH to an other Modula 2 compiler The body of the DisplayArray procedure of StochStat is commented out The reason is that the implementation tries to work with simulated open arrays defined as a POINTER TO 0 80001 OF REAL which leads to a page fault on protected memory systems jth June 12 2003 Page 8 RASS 1 1 Developer s Guide 3 RASS Development environment and techniques This chapter describes the RASS development environment on a Unix Workstation using the EPC Modula 2 compiler The RASS development environment is organised in a hierarchical tree of directories On the top level are SysDep AuxLibDev DMLibDev MWLibDev tools and Models DMLibDev is subdivided into CoreHigh and OptHigh All modules contained in CoreHigh are referenced by the MW library The modules in OptHigh are modules which are referenced by many MW simulation models Note that this doesn t reflect the interactive DM s kernel and optional module structure cfg Table 2 MWLibDev is subdivided into High MOD and Base MOD which reflects the organisation of the Mac development environment The RASS MW library contains all modules object and definition of the
47. omatic translation of interactive programs developed for a graphical user inter face with windows and menus in order to execute them in a batch mode to produce correct and reliable simulation results Three complex case studies from the field of ecology and engineering showed that performance gains of 1 100 7 200 overhead included can be obtained when RASS is run on a SUN 510 server relative to the time needed when running the MDP interactively on an average personal computer Since all transfers showed to be user friendly and smooth we concluded that RASS offers RAMSES simulationists an efficient and attractive alter native for solving interactive MDPs whenever off line simulations are needed or when it is a necessity because of the high computing requirements 1 INTRODUCTION The simulation of ill defined systems e g ecological systems challenge most existing simulation software by specific sophisticated requirements Cellier and Fischlin 1982 Kreutzer 1986 Fischlin and Ulrich 1987 Vancso et al 1987 Vancso 1990 At an early stage of a research project an interactive simulation environment is of paramount help for the modeling of ill defined systems However at later stages i e when simulation studies such as sensitivity analysis parameter identification or stability properties are of prime interest batch production of simulation results is needed In order to satisfy both needs during the entire course of a rese
48. otocol To provide a user friendly transition from interactive RAMSES to RASS we created a tool called RASSMakeMake It generates automatically a make script that produces the executable simulation program The output of a simulation can be explored locally or transferred back to the host where the interactive PA resides 3 3 Case Studies We selected three case studies to cover all three currently by RAMSES supported standard modeling formalisms SQM DEVS and DESS a ForClim Bugmann 1994 SOM that models the stochastic species succession of forests and is currently used to study the impact of climatic change on forests Bugmann and Fischlin 1994 ForClim s input data files contained machine specific characters which had to be re moved before the ForClim was able to process them under RASS We ran that experiment that generates the so called reference output which was not numerically identical but since ForClim s output depends on pseudo random numbers it is highly dependent on the precision of the floating point instructions However the results deviate only insignificantly from the expectations and were therefore interpreted as correct b Diversity Fischlin eral 1994 a DEVS simulat ing the reinvasion of species and diversity restoration on an island which has been hit by a volcanic eruption RASS returned exactly the same number of years for diversity restoration as the interactive version c NumInt a DESS
49. resists as it s target It generate also some basic library test programs 2 clean this rule deletes all object and library files 3 depend this rule adds dependency rules to the end of the Makefile depend csh c setenv M2FLAGS M2FLAGS listdepf gt makedep echo DEPENDS a gt eddep echo gt gt eddep echo gt gt eddep echo d gt gt eddep echo r makedep gt gt eddep echo w gt gt eddep ed Makefile lt eddep rm f eddep makedep DEPENDS make depend needs this line It is a csh script which uses listdepf see Tools to generate a list of dependency rules Then it uses ed to edit the Makefile i e to find the line with DEPENDS to delete the rest of the file and to add the dependencies Finally it removes the temporary files makedep and eddep which have been generated during this procedure Main Makefile M2GLOBALFLAGS sun D0 nobounds norange RASSHOMESS PWD Definition of the flags for all calls to the driver program of the EPC Modula 2 compiler The switch sun activates the compiler s SUN mode and tells the linker to use the SUN library Note that the switch sun works also with the EPC Modula 2 compiler on the IBM AIX machines The home of the development environment is set to the value of the environment variable PWD this is the string of the current working directory jth June 12 2003 Page 11 RASS 1 1 Developer s Guide all cd SysDep make M2GLO
50. teractive RAMSES host prior transferring it to the RASS host If you start your simulation respectively experiment and get the desired result by answering all eventual modal dialogues just by pressing the enter key you can expect to get the same behaviour on the RASS host 3 2 Transferring MDPs to the RASS host A developed MDP consists of one or several Modula 2 source files and optional data files They have to be transferred to the RASS host Because most hosts have a different definition of text files it is highly recommended to use FTP File Transfer Protocol to transfer files since it translates them according to the specification of the target host The EPC Modula 2 compiler expects source files with the extension mod and def All modules must have the name Modulename mod or Modulaname def for definition modules The naming is case sensitive If you transfer files from the RAMSES environment they are usually named using the extensions MOD and DEF The two shell scripts MOD2mod and DEF 2def performs this renaming for you To use these scripts change to the directory where your source files reside and call both scripts 3 3 Building executable simulation programs After you have transferred your source and data files to the RASS host you can build an executable simulation program This involves both the compilation of the sources and the linking with the MW library The shell script RASSMakeMake helps you with this task Run RASSMake
51. that compares the position of earth satellite computed with fixed step integration methods of various orders with an analytically determined expectation The purpose of NumInt is to explore the range of valid simulation results limited either by too big or too small step sizes rounding errors The RASS re sults were the same for lower order integration methods whereas those of the higher order integration methods dif fered for the smaller step sizes due to the rounding errors three MDPs were transferable without any changes The only exception was ForClim which required initially one iteration 3 3 1 Performance Measurements In order to compare a simulation cycle of an interac tive MDP between interactive RAMSES and RASS we started measurements only from the moment of an already developed RAMSES MDP and neglected constant terms if they were approximately the same on both hosts e g the time to build and link load the MDP Terms tt Turnaround time ts Time to execute a simulation experiment te Time to transfer MDPs and or data files to the simulation host and transfer the result files back to the PA host ta Time to set up the interactive analysis PA of the results tt RAMSES ts tt Rass ts lc tg Since tg lt lt tg we get tt RAMSES ts tt Rass ts MW RASS Mac SUN 1 tt ts ts te tt ts te ForClim 37800 3134 132 3266 Diversit
52. tion International P O Box 17900 San Diego Cal 92177 USA p pp 99 103 jth June 12 2003 Page 10 RASS 1 1 Developer s Guide J rg Th ny 1 LIBRARY 1 2 MODULE DESCRIPTIONS na an de keeutecsc 2 23 A TE T E 2 PSU CIPIT P PNE 2 21 2 AOLE DEE EEE etos t MN NM ma M eC EA 6 PAPA MIN NIGAM P EE 6 2 2 1 DMCOnVETSiONS ERE anaiai E E EE EE EEE EE TE AE EEEE KEE E 6 PAAA DI S EI 6 PASS NEUVES Eee 7 22 DIAG TOK cu E 7 2 25 DMM thLib uni eb ana 7 PALACE bU hu np T 7 2 2 7 DMStorase Er 7 2 2 5 US e e Ot pO c LIE Aue anne A HD 7 22 2 DPMS y Ste Mir 22 want tte e Mnt udi sense 7 2 2 IO DMW NJO W He ET RR 7 2 2 11 DMWindows amp DNDDGrpks een 8 2 2 12 DMEditFields amp 0 a lebens 8 2 243 NA CIWS een eat 8 2 3 MW Modules aan een 8 PANNO CS oh sei see ee 8 3 EPC RASS DEVELOPMENT ENVIRONMENT AND TECHNIQUES 9 TAT S st ass 9 32 Makenles tetti tiam t hosce tas adita did tp e Dum E i m 11 3 9 LOOMS t uso ee ee 14 4 FUTURE DEVELOPMENT OF 55
53. tion of the RASS modules and the development environment as required for further development 1 Research Aids for the Modelling and Simulation of Environmental Systems 55 Towards bridging the gap between interactive and off line simulation J rg Th ny Andreas Fischlin and Dimitrios Gyalistras Systems Ecology Institute of Terrestrial Ecology Swiss Federal Institute of Technology Z rich ETHZ CH 8952 Schlieren Z rich ABSTRACT Interactive model exploration is an important step in the process of model building of ill defined systems such as ecosystems a task which is well supported by RAMSES Research Aids for the Modelling and Simulation of Environmental Systems However interactive simulation is of minor interest at a later stage of research when large scale batch oriented simulation experiments are needed RASS the RAMSES Simulation Server typically located on a high performance computer translates com piles links and executes in a batch mode interactive RAMSES model definition programs MDP MDPs must be given in source form and formulated according to one or any combination of the following standard forma lisms SQM Sequential Machine DESS Differential Equation System Specification or DEVS Discrete EVent System Specification The simulation results generated and returned by RASS can then be explored interactively by means of the post analysis component of RAMSES RASS was found to allow for aut
54. y 638 13 120 133 70758 860 115 975 Table 1 Turnaround time in seconds of three case studies in the interactive RAMSES environment on a Macintosh Quadra 700 and the batch RASS environment on a SUN S10 The over all performance gain obtainable by the simu lationist was a factor 12 82 for tg anda factor 11 72 for tt Table 1 3 3 2 Post Analysis The RAMSES PA session allows to interactively ex plore simulation results previously written to a stash file by MW or RASS Fig 1 The PA supports an arbitrary number of stash files simulation runs per file and models per run limited only by the computer s available memory Based on MW PA mimics the model behaviours by reading the results from stash files instead of computing them on line Fig 3 Not only does it allow to compare stored results among several stash files but also with interactively simulated model behaviours 29 5 24 15 17 2 GE Model parameters T Shell PostAnalysis File Edit Settings Windows Solve Models Model names Ident WAYS E Parameter names Ment Unit A Post simulation analysis 7 Post simulation analysis ul State variables A State variable names Ident Unit A Data frames Post simulation analysis td Data frame elements Ment
Download Pdf Manuals
Related Search