Home

CO5BOLD User Manual

1. rhd shortrad dir 102 in rhd shortrad module F90 rhd short characteristics radiation direction loop 02 Category performance enhancement OpenMP Determine position of PARALLEL statement relative to outer loop in rhd shortrad dirX Both settings give the same results but might show a different performance on a specific machine Values o 0 default PARALLEL statement inside of outer loop o 1 PARALLEL statement outside of outer loop 3 8 Optimization Compiler Switches 33 3 8 rhd shortrad lambda 101 in rhd shortrad module F90 rhd short characteristics radiation lambda loop 01 Category feature activation Handling of extra arrays to allow partially implicit Lambda iteration Values o 0 default Only fully implicit Lambda iteration allowed or fully explicit treatment o 1 Also partially implicit Lambda iteration allowed rhd shortrad formal t01 in rhd shortrad module F90 rhd short characteristics radiation formal timing 01 Category additional output Produce timing information for routine which gives the formal solution of the radiation transport equation with the help of short characteristics It can be used together with OpenMP and should cause no noticeable performance loss Values o undefined default no timing information o defined call subroutines to measure elapsed time rhd shortrad step tO01 in rhd shortrad module F90 rhd short characteristics radiation step timing 01 Categ
2. In this case a file rhd done see Sect 5 4 is produced A messages like FEAR GAARA ok I A ACI I 1 1 2k 21 21 25 21 2 2k 21 21 A ACA A AI 21 21 211 kkk kk kkk k k kk Severe error SHORTRAD Time step below absolute limit Error index 100 Interrupt computation FEA AAR IC I I I IOI kkk kkk kk kk kkk kkk kkk k kkk ok kkk k k k kkk kkk k k k k marks an exit with an error and without rhd done file A message about the final model like ACTION Write final model lt lt lt lt lt lt lt lt Model file rhd end opened on channel 12 is followed by some timing information like e g Timing statistics rate x factor 1000000 x 10000 Process Samples Total time Mean time sec sec RHD code 1 410 830017 410 830017 uio output routines 2T 37 469997 1 387778 HYD bound_3Dcenter 2000 3 540000 0 001770 Hydrodynamics routines 2000 48944 660156 24 472330 HYD 1 2000 16350 610352 8 175305 HYD 2 2000 16052 459961 8 026230 HYD 3 2000 16520 798828 8 260400 Viscosity routines 3D 2000 25446 400391 12 723200 VIS make box modelvis 2000 0 010000 0 000005 VIS copy box modelvis 2000 1750 960083 0 875480 VIS delete box modelvis 2000 0 000000 0 000000 Radiation transport routines 2000 243729 000000 121 864502 SHC step 2020 244064 515625 120 824020 SHC step dtime init EOS 16024 49312 738281 3 077430 SHC step dtime explicit 16024 174542 343750 10 892558 SHC formal 16024 170085 437500 10 614418 SHC formal init 16024 35591 378906 2 221
3. Make the UIO IDL routines visible to IDL somehow There are several ways to do that Three are described below NOTE Before using IDL the environment variables for the CO5BOLD paths should have been set Use setarcdeppaths sh or ksh csh for this purpose Im portant is UIOPATH which specifies where to find the IDL routines for the UIO handling We recommend to create an initialisation file e g named start pro which should be called after starting IDL e g start This is necessary to define relevant paths of IDL subroutines and to provide the UIO package Thus the file should contain the following Add user IDL directory to search path addpath expand path UIOPATH id1 addpath addpath texpand path HOME HYDRO IDL rhdpro if strtrim addpath 2 ne then path addpath path delvar addpath Initialize uio routines uio init progrm by hand Alternatively one might set the IDL path variable accordingly like export IDL_PATH UIOPATH id1 for example in the bashrc file before starting IDL Or you just make a symbolic link from the UIO IDL routines at their original location to a sub directory of the main IDL directory which should be in the IDL path anyway The initial single call of uio init is necessary in any case 7 2 CO5BOLD Data in IDL Important to know is that all operations can be performed in the command line of IDL This allows an interactive
4. o constant open boundary with constant extrapolation of all values no gravity no radiation o closed closedtop closed wall can handle gravity open for outward radiation o closedbottom closed wall handles gravity radiation in diffusion approximation o periodic periodic boundaries for hydrodynamics and radiation o transmitting transmitting boundary for hydro and outward radiation Any of these values can be specified But in fact not all of them are recognized by all modules Therefore some parameters are for test purposes e g shock calculations only In simulations of a solar like star with the MSrad radiation transport module the side bound aries have to be periodic In simulations of a red supergiant all boundaries including the sides will typically be transmitting As an alternative closed boundaries can be chosen in this case e character top bound The boundary condition at the top of the model is given by for instance character top bound f A80 b 80 n top boundary conditions transmitting Possible values are o reflective closed wall no gravity no radiation o constant open boundary with constant extrapolation of all values no gravity no radiation o closed closedtop closed wall can handle gravity open for outward radiation o periodic periodic boundaries for hydrodynamics and radiation o transmitting transmitting boundary for hydro and outward radiation In almost every simulation of stellar co
5. CK p oF Xp oT XT oT _ Olnp M m OlnT Ee EE OlnT I5 1 Vad Em I XP Xp Definition of adiabatic exponents 2 2 3 COSBOLD equation of state COSBOLD equation of state input p e COSBOLD equation of state output T Pi T S Cap Ge dey 5 p e de All required thermodynamic coefficients can be expressed in 3 2 30 2 2 4 Derived thermodynamic coefficients First the missing derivative 35 can be found from the relation which is obtained from the equality of the mixed derivatives in Eal4 written as 1 p ds de d S pre TR terms 11 12 20 21 of 22 23 8 2 EQUATIONS Then First adiabatic exponent This relation is obtained by combining Eq 4 with the identity _ Pp 2 dp 52 do E p 26 The adiabatic sound speed is then obtained as Third adiabatic exponent 28 This relation is obtained by combining EaH with the identity oT oT dT gt d d 29 El pe 5 and then using Eq 2 Adiabatic temperature gradient 30 since 5 se 55 _T3 1 31 lnp Omp Omp Ty Adiabatic energy changes or um c2 zv We define the coefficients c and c through the relation ds c dlnp c d In p 34 Entropy change at constant density 2 2 A collection of thermodynamic relations M Steffen AIP 9 This relation is obtained from the equality of the mi
6. The Fortran standard does not quarantee that unformatted i e also UIO files are readable on all machines But it is always possible to produce formatted UIO files on a machine which are readable on all others And with some fiddling with compile options or the call of machine specific subroutines provided by the compiler vendor it was up to now always possible to enable the access to binary UIO files of one common format IEEE big_endian on all machines and compilers tested Compaq alpha Cray Hitachi HP V2500 HP Itanium2 IBM Intel AMD with Linux OS and PGI or Intel compiler NEC SX 5 SGI and Sun Each file entry is a header data unit The header contains information to identify the entry and to specify the format and size of the following data block This block usually consists of a scalar or an array In some cases it is empty e g for labels or it contains more complex information for tables The first version of UIO routines was written in FORTRANTT It still exists However further development was done with the Fortran90 versions Therefore the use of the FORTRANT7 routines is not recommended anymore The current Fortran version of the UIO routines is a set of Fortran90 modules To allow a communication between Fortran and IDL programs an IDL version of the UIO routines has been written The correspondence between Fortran and IDL routines is rather close But in detail there are differences Currently IDL Version 6 0 is us
7. cO range 0 1 1 0 typically 0 5 off 0 0 0 5 The default is 0 9 Nevertheless since the Roe solver is constructed to handle shocks and rapid changes in density and energy this check is usually not needed It can be switched off by setting c maxeichange 0 0 90 5 CONTROL AND DATA FILES e real c radcourant The radiation transport routines are subject so time step restrictions too And in typical scenarios its the radiative timescale are the shortest one and poses the tightest restriction Contrary to the hydrodynamics routines the timescale relevant for the stability of the radi ation transport scheme can only be estimated using the characteristic timescale of a small sinusoidal temperature disturbance with a wavelength of the grid size in a homogeneous background and grey radiative energy exchange The radiative Courant factor can be set e g with real c radcourant f E15 8 b 4 n RAD Courant factor u 1 amp cO range 0 0 lt C_radCourant typically 1 0 2 5 If the estimate of the timescale would be correct a value of 2 0 would cause the temperature fluctuation on the shortest scale to flip its sign setting the absolute stability limit A value of 1 0 would lead to a damping of theses fluctuations within one time step But in practice even higher values for example 2 5 show a reasonable behavior This might be due to the effect that the shortest radiative timescale only occurs at single points or in 2D laye
8. o ross_m05 opta o ross m O opta o ross m20 opta o sunurl opta o sunur2 opta o t5000g44mm20 o t5000g47mm20 o t6300g40mm20 o t6500g44mm20 opta opta opta opta o zzcetilg opta o zzcetil opta e character opapath The opacity file does not have to be in the working directory Instead its path can be specified e g with character opapath f A80 b 80 n path of opacity file amp cO astro b bf for opa dat home a bf for opa dat 5 3 Parameter File rhd par 77 5 3 7 Hydrodynamics Control character hdscheme With this parameter the type of the hydrodynamics scheme can be specified as in character hdscheme f A80 b 80 n Hydrodynamics scheme amp cO Roe approximate Riemann solver of Roe type amp ci RoeMagKin Roe solver kinetic magnetic field transport amp c2 None skip hydrodynamics step entirely Roe Possible values are o None The hydrodynamics step is skipped entirely for test purposes Note that in this case some initializations necessary for the generation of the mean file are omitted too o Roe default The standard Riemann solver of Roe type is activated This value will in almost every case be chosen o RoeMagKin The standard Roe solver is extended to transport passively a magnetic field This is a test implementation to check if the general magnetic field handling works character reconstruction This parameter determines the order and agg
9. openrd openur pptrnt qmaadd gmadel rd rdfifo rdhdex rdhead rdlabl rdtab Skipda slhdex st2dim Struct rd tab0 tabc tabm tabr tabw uclose uopen vnanrm Wwf2rf Wr wrfifo wrhdme wrhead wrlabl wrtab Determine the default output format for numbers Compose dimension string Extract value of keyword from table Extract one term from the input line Transform a list of items into its components Determine file contents list of all entries with its positions Get information about environment Initialization procedure for input output routines Merge the input term keyword value gt keyword value Merge a list of terms keywords and their values into a line table Make list with possible conversion types From column number or entry name find table entry number Open file for reading read header Open file for writing write header Print term table in pretty form Transform a string into a string with quotation marks if necessary Parse string inline and remove quotation marks if necessary Reading scalar and array data of all types Read file header Read header of variable and extract keywords Read header Read label Read table of integer real and or character data from file Skip data block Search header of variables given by list and extract keywords Parse dimension string Read uio file and put data into anonymous structure Create empty table structure Change a
10. eos mm20 l eos Standard EOS file for metal poor star M H 2 0 with ex tended range in internal energy and density towards lower values The older file eos mn20 eos did not reach far enough eos gamma140 eos EOS table for simple gas with constant 1 4 In this case all quantities could be faster computed than by interpolation in a table Nevertheless for compatibility reasons to be able to use the existing EOS Fortran routines the table is provided eos gamma166 eos EOS table for simple gas with constant T 5 3 cpheos mm00 eos Copenhagen EOS for their solar composition e character eospath The equation of state file does not have to be in the working directory Instead its path can be specified e g with character eospath f A80 b 80 n path of EOS file amp cO astro b bf for eos dat home a_bf for eos dat 76 5 3 6 Opacities e character opafile The opacity file can be specified with e g 5 CONTROL AND DATA FILES character opafile f A80 b 80 n opacity file name amp cO g2va opta big grey opta Y ci empty gt no radiation transport phoenix_opal_grey opta So far there exist already a couple of files o davmf opta o f5v opta o g2va opta o g2v lowhe opta o g2v m20 opta o g2v opta o hmin pOO opta o opal lowhe opta o opal m05 opta o opal m10 opta o opal m20 opta o phoenix dust grey opta phoenix dust ob4 opta phoenix opal grey opta
11. ful z itime ful z time format A 14 16 E15 8 amp Now do the data handling demo print Mean density avg ful z rho amp i i 1 amp endif amp endwhile All necessary counter information is stored in the structure listdata Note that you can specify an entire group of files with e g modelident _ modelident _ modelident 3 or modelident 129 310 217 63 5 Control and Data Files Table 12 shows a list of all files necessary to run CO5BOLD Figure 6 gives similar information but is not quite up to date Executing the makefile produces an executable rhd exe Its name can of course be changed afterwards The names of the three control files rhd par rhd stop and rhd cont and of the status file rhd done cannot be changed without modification of the source code The names of EOS opacity and CO5BOLD data files can be chosen freely in the parameter file rhd par Table 12 only contains dummy names File Sect I O Type Description rhd exe executable main program rhd par I control data UIO central control file rhd stop I control file to force controlled stop of simulation rhd cont I control file to force continuation after stop data eos I data UIO tabulated equation of state data opta I data tabulated opacities rhd sta I data UIO start model e g end file of last run rhd done O status exit status written if run was successful rhd end O data UIO end
12. p STADIR PARFILE rhd par 103 104 6 RUNNING A SIMULATION Copy start file if test s BAKDIR INFILE then cp p BAKDIR INFILE rhd sta else cp p STADIR INFILE rhd sta cp p rhd sta BAKDIR INFILE fi Copy executable cp p RHDEXE rhd exe Execute RHD rhd exe gt rhd out RHD_EXIT if RHD EXIT 0 s rhd done then Exit status not zero error may have occurred during execution of rhd exe if RHD EXIT 0 then echo Non zero exit status RHD EXIT occurred during execution of RHD else echo No rhd done file found assume error during execution of RHD fi echo Execution of job chain terminated Modify nrhd cmd set termination character awk N 4 print next N A4 print next IN A4 print next DONE T printf 4 77s Nn 0 DONE T next print rhdi cmd cat gt nrhd cmd Terminate chain by simulating ACT eoc ACT eoc else Modify command file add in appropriate column to indicate proper execution eum awk 3 t A print next A4 print next IN A4 print next DONE T print gt cmdline printf 77s n 0 DONE T next print rhdi cmd cat gt nrhd cmd fi sel Move data into backup directory cp p rhd par 4 BAKDIR PARFILE mv rhd out BAKDIR 0UTFILE out mv rhd p
13. 2 Running CO5BOLD on a Machine with Batch System Last modification 2002 12 13 gt set xv BASEDIR users bf dat job j1 STADIR BASEDIR sta Start directory contains par rhdi cmd rhd1 job WRKDIR BASEDIR wrk Work directory BAKDIR BASEDIR bak Backup and output directory RHDEXE STADIR rhd exe RHD program to be executed export NCPUS 4 echo NCPUS export OMP_NUM_THREADS 1 Jump into work directory cd WRKDIR Loop execute RHD code possibly several times in one job for IRUN in 1 do echo echo CO5BOLD Run IRUN date Clear up work directory set te rm set e FERRERA RARA ROH REPAIR R RERE IRR RERE I EH B IE IG E Get old command file select actual command line and read variables from command line us cp p STADIR rhdi cmd rhdi cmd awk i t 1 print next INx A4 print next IN A4 print next DONE T print gt cmdline printf 77s n 0 DONE T next print rhdi cmd cat gt dummy_nrhd cmd read INFILE OUTFILE PARFILE ACT lt cmdline mv dummy nrhd cmd OUTFILE _nrhd cmd echo INFILE echo OUTFILE echo PARFILE echo ACT THEETHEIETHEIETE EH I B E B E BI E EE BERE ERA RR ERA T IB RERE ETHER HE EHE TH BIER if ACT dum then PER NE RRI NA RRI RR HERR HE PTE RE TEAR EE ECE RHD execution Copy parameter file cp
14. 23 80 B5 57 0 ET E 73 cell centering NEC NEC SX 5 OMP NUM THREADS OMP SCHEDULE opacities OpenMP activation in configure script chunk size OMP NUM THREADS OMP SCHEDULE on Cray VXI on Hitachi on SGI on Sun output full data sets mean data rhd done rhd out PGI radiation transport README rhd cont rhd done rhd exe rhd out rhd par boundary conditions character bottom_bound 113 character description 70 character dustschene 81 character eosfile 5 character eospath 75 character file id 0 character grav mode 70 character hdschene 77 character history 70 character infile_start 92 character opafile 76 character opapath 76 character outconv_end character outconv_fine character outconv_full character outconv_mean character outfile_end character outfile_fine 95 character outfile_ful1 92 character outfile_mean character outform_end character outform_fine 96 character outform_full 93 character outform_mean character radpressure character radraybase 83 character radraystar character radschene 81 character reconstruction 77 character side_bouna 72 character top bound 72 dust 80 effective temperature equation of state fileform uio 69 gravity header 69 hydrodynamics input output integer dtimestep out fine integer endtimestep 87 integer n_hydcellsperchunk integer
15. 86 5 CONTROL AND DATA FILES o 0 0 1 0 Restricted step size o 1 0 No restriction standard step size o gt 1 0 Extra large steps This value has to be chosen carefully to get optimal performance Is the step size too small the convergence is safe but too slow A too large step size inhibits convergence and leads to a decrease in the time step which results in a bad performance too real c radtvisdtau Using the LHDrad module the limit in delta optical depth rho kappa dx below which the radiative temperature viscosity temperature smoothing is to be applied can be set with e g real c_radtvisdtau f E15 8 b 4 amp n Optical depth limit for temperature viscosity u 1 Qui The introduction of this temperature diffusion is a somewhat desperate and inelegant attempt to improve the behavior of the Greens function hot cells should be cooled cool cells should be heated This diffusion is necessary for not well resolved models It is switched off with c_radtvisdtau lt 0 0 real c_radtvis Using the LHDrad module the amount of the radiative temperature viscosity tempera ture smoothing can be specified e g with real c_radtvis f E15 8 b 4 n Temperature viscosity u 1 1 6 For well resolved models it should be switched off with c radtvis lt 0 0 But often its use is necessary 5 3 11 Process Time Management In this group several parameters can be set which control the start of the time coun
16. HLR Stuttgart An environment variable has to be set to F RECLUNIT BYTE before execution of a program to enable UIO to compute proper record lengths The cross compiler on sunsrv or crossi is called with sxf90 Thus the environment variable F90 COMPILER sx90 has to be set before running the configure script No optimized version of CO5BOLD has been achieved yet Some maybe useful switches are e sx5 generate instructions for SX 5 e C vopt normal optimization in vector mode e Wf M noflunf M noinv M noinexact M setall suppress some exceptions e P openmp parallelization with OpenMP e Ep call cpp preprocessor e pi exp inlining of a list of routines see Sect e dw floatO no special environment variable use internally and in files the 4 Byte big endian format The compiler flags in the configure script are F90FLAGS C hopt sx5 dw floatO Wf L nostdout L fmtlist L inclist L mrgmsg L transform M noflunf M noinv M noinexact M setall pi exp rhd shortrad operator exp rhd shortrad dtauop F90MODULES F9OTIME DMSrad raytas 1 Hhttp nash bch msu edu docs ifc FortranReleaseNotes htm 3 8 Optimization Compiler Switches 43 3 8 12 SGI Origin CO5BOLD has been compiled and tested on up to 8 processors on the SGI 2000 machine at TAC in Copenhagen and the SGI 3800 machine at the NSC in Linkoping More recently the code was used on the UKAFF machines see Sect 3 8 13 and the compu
17. SHORTrad module requested assume default F90 SHORTRAD 1 No dust module requested assume default F90 DUST O No MHD module requested assume default F90 MHD O No explicit machine requested assume default F90 MACHINE local List of control environment variables F90 COMPILER 3 2 Compilation Procedure for CO5BOLD 15 F90 PREFLAGS F90 POSTFLAGS F90 PARALLEL scalar F90_DEBUG F90 LHDRAD F90 MSRAD F90 SHORTRAD F90 DUST F90 MHD F90 MACHINE local MACHINE 1686 F90 BASEPATH home bf for i O oO p od Linux system with 1686 architecture PGI compiler version 3 3 2 pgf90 byteswapio fast Mvect sse Mcache align Minfo inline Write compiler name and flags into file compiler flags info Makefile already exists It is appended to Makefile old New Makefile WIlIUbbOn 5eseem oo o 9 do E maba Anh qaoqoe A new Makefile is produced An existing one is appended to Makefile old Addition ally the file compiler flags info is written which contains the compiler call in Fortran format 7 Check the output of the configure script and the header of the new Makefile You get an overview over the relevant environment variables that control the configure script see Sect 3 6 with env grep F90_ Obs at the beginning there might be none 8 Look into the header and if necessary the rest of the configure script or into Sect 3 6 to find out how to change the environment variables to control the script prop
18. Sections and 3 8 then the procedure should be fairly easy The general compilation procedure is now If a directory for the current machine exists in the tar ball and the configure script is there tar zxvf for tar gz cd for hd rhd YOUR MACHINE configure make If the directory for the current machine has to be created tar zxvf for tar gz cd for hd rhd mkdir YOUR MACHINE cd YOUR MACHINE In s conf configure configure make The compilation process is explained in more detail in Sect The configure script is described in its header and in Sect The directory structure is shown in Tab 1 All Fortran files are listed in Tables 3 and 3 2 Compilation Procedure for CO5BOLD The installation procedure has changed significantly since the last release now there is a configure script see Sect that creates the complete temporary makefile which can be used to compile CO5BOLD and produce the executable rhd exe Installation procedure 1 Choose create a proper base directory This will usually be HOME Then the master directory will typically be HOME for this is the default created by the tar file Some prefer to rename it to HOME HYDRO 2 Put all source files and the configure script there This will be done typically by expanding the gzipped tar file for tar gz e g with tar zxvf for tar gz or by copying all files from an existing installation On a restricted UNIX you might be forced to use g
19. UIO DATA FORMAT 4 3 4 Recommendations for Standard File Structure The very first entry in an UIO file is always the fileform uio entry containing information about the file format and conversion type Afterwards entries can follow in any order But it is perhaps a good idea to start the file with three special entries file id description history as in fileform uio form formatted convert ieee 4 character file id f A80 b 80 n File identification uio demofile character description d 1 2 f A80 p 1 b 80 n File description This is a file to demonstrate the recommended start entries for all UIO files character history d 1 1 f A80 p 1 b 80 n File history UIO demo file 22 Dec 1997 14 15 15 A recommended format for sets of multi dimensional arrays e g hydrodynamics x axis y axis z axis density velocities energy is shown in Sect 5 1 4 4 Files amp Directories amp Paths All UIO routines are located in sub directories of a common directory called e g uio which also contains a old Readme file The subdirectories and their contents are bin Shell scripts uiolook uiocat uioinfo 90 Fortran90 source codes object files executables idl IDL routines man man1 manual pages for shell scripts uiolook uiocat uioinfo tex old description files in IATEX the most recent version is part of this document To use the UNIX scripts and the makefile you need a global system variable UIOPA
20. a smooth restart Before the restart the rhd stop file has to be deleted The simulation can be continued by just initiating a new run If the file rhd cont exists at the beginning of a simulation the code tries to resume an interrupted computation The initial model will not be taken from the start model file infile_start but from the final model outfile_end The data for the full and the mean file is not written into new files but will be appended to the existing ones In this way a simulation can be interrupted and continued in a fairly safe way It is possible to analyze the final model and to changes values in the parameter file Keep in mind that after a restart with rhd cont the specifications about the length of the job e g the number of time steps will be counted from the restart point and not from the beginning of the original simulation To interrupt a job with rhd stop can be very handy The continuation with rhd cont and the old parameter file is not to be preferred over an ordinary restart with a new parameter file If a run was successful i e it was completed because one of the regular termination conditions was fulfilled e g the requested number of time steps was performed the exit status file rhd done is produced Currently it contains the date and time of its generation The existence of this file can be checked within a script to determine if the simulation was successful and should be continued Note the existence of an
21. c radmaxeichange real c radthintimefac In the LHDrad module only the radiative time step restriction due to energy changes can be relaxed further in the optically thin by specifying e g real c radthintimefac f E15 8 b 4 amp n time scale reduction in optically thin u 1 amp cO range 0 1 1 0 typically 0 5 0 60 A value 0 0 or real c radtvisdtau lt 0 0 switches off this relaxation 5 3 Parameter File rhd par 91 e real c viscourant The tensor viscosity routines have their own time step restriction The recommended typical viscous time step can be set e g with real c viscourant f E15 8 b 4 n viscous Courant factor u 1 amp c0O range 0 0 lt C visCourant typically 0 5 1 0 better 0 25 0 5 As the corresponding viscous timescale is typically longer than the radiative one and even the Courant timescale from the Roe hydrodynamics routines this factor is often irrelevant The absolute upper stability limit is located at c viscourant 2 0 Values around 0 5 to 1 0 are more typical In some extreme cases in simulations of the solar chromosphere it has turned out that an even lower value 0 2 is necessary to prevent some spikes in the neighborhood of strong colliding shocks e real c viscourantmax The absolute upper limit for the viscous time scale can be set with real c viscourantmax f E15 8 b 4 n maximum viscous Courant factor u 1 cO range C visCourant lt C visCourantmax typica
22. can stay unchanged for a group of similar simulations Only the parameters with an explicit time dimension should be checked in all cases they scale with characteristic timescales and depend particularly on gravity e real dtime start The initial time step recommendation of a simulation is usually taken from the start model file It can be overwritten e g with real dtime start f E15 8 b 4 n Start time step u s 1 0E 03 A value lt 0 0 means that the original value from the start model is used real dtime_min In some rare cases it might be useful to specify explicitly the minimum time step with e g real dtime_min f E15 8 b 4 n Minimum time step u s amp c0 dtime min 0 0 gt no restriction 1 0 This value is used even if restrictions from the Courant condition try to enforce a smaller value A fixed time step can be prescribed by setting dtime min dtime max to some positive value A value dtime min lt 0 0 means that this time step restriction is completely ignored which is the case that should usually be chosen real dtime max It is possible to explicitly specify the maximum time step too e g with real dtime max f E15 8 b 4 n Maximum time step u s amp c0 dtime max 0 0 gt no restriction 1 0E 05 A fixed time step can be prescribed by setting dtime_min dtime_max to some positive value A value dtime_max lt 0 0 means that this time step restriction is completely ignored which is the case
23. case distinction with masks weights 0 0 or 1 0 slope reduction of source function 5 case distinction with if then else construct modified slope reduction of source function o 6 default case distinction with masks weights 0 0 or 1 0 modified slope reduction of source function o 8 test version rhd shortrad operator 102 in rhd shortrad module F90 rhd short characteristics radiation operator loop 02 Category performance enhancement Select the way the short characteristics operator is accessed Values o 0 default The routine with the short characteristics operator is called within a loop and should be inlined o 1 The program fragment with the short characteristics operator is included No inlining necessary 32 3 PROGRAM FILES INSTALLATION COMPILATION e rhd shortrad dtauop 101 inrhd shortrad module F90 rhd short characteristics radiation delta tau operator loop 01 Category performance enhancement Choose type of short characteristics tau coupling operator Values o 1 case distinction with if then else construct default if rhd shortrad operator 101 1 3 5 o 2 case distinction with masks weights 0 0 or 1 0 default if rhd_shortrad_operator_101 2 4 6 rhd_shortrad_dtauop_102 inrhd shortrad module F90 rhd short characteristics radiation delta tau operator loop 02 Category performance enhancement Select the way the operator for the t
24. character box_id f A80 b 80 n Block identification integer dimension d 1 2 1 3 f I7 p 6 b 4 real time f E13 6 b 4 n time u s integer itime f I11 b 4 n time step number u 1 real xci d 63 63 63 63 63 63 f E13 6 p 4 b 4 amp 64 5 CONTROL AND DATA FILES n x1 coordinates of cell centers u cm ds 0 0 0 1 0 1 real xc2 d 63 63 63 63 63 63 f E13 6 p 4 b 4 amp n x2 coordinates of cell centers u cm ds 0 1 0 0 0 1 real xc3 d 63 63 63 63 63 63 f E13 6 p 4 b 4 amp n x3 coordinates of cell centers u cm ds 0 1 0 1 0 0 real xb1 d 63 64 63 63 63 63 f E13 6 p 4 b 4 amp n x1 coordinates of cell boundaries u cm ds 0 1 0 1 0 1 real xb2 d 63 63 63 64 63 63 f E13 6 p 4 b 4 amp n x2 coordinates of cell boundaries u cm ds 0 1 0 1 0 1 real xb3 d 63 63 63 63 63 64 f E13 6 p 4 b 4 amp n x3 coordinates of cell boundaries u cm ds 0 1 0 1 0 1 real rho d 63 63 63 63 63 63 f E13 6 p 4 b 4 n Density u g cm 3 real ei d 63 63 63 63 63 63 f E13 6 p 4 n Internal energy u erg g real vi d 63 63 63 63 63 63 f E13 6 p 4 n Velocity 1 u cm s real v2 d 63 63 63 63 63 63 f E13 6 p 4 n Velocity 2 u cm s real v3 d 63 63 63 63 63 63 f E13 6 p 4 n Velocity 3 u cm s label endbox label enddataset date 02 01 2002 16 17 43 322 4 b 4 b 4 b 4 b The UIO format is described in some detail in Sect Each entry has a type e g label real characte
25. e g with character radraybase f A80 b 80 n Base axis system amp cO unity random randomgroup random Allowed values are o unity default During all time steps and radiative sub steps the direction of the rays stays the same o random At each time step and radiative sub step a new base axis system is chosen at random o randomgroup At each new time step a new base axis system is chosen at random It is kept for all radiative sub steps Because typically only a relatively small number of rays is chosen per time step with radraystar it is advisable to vary the directions of the rays by choosing radraybase random or randomgroup to cover the entire sphere at least over a longer time e character radraystar Using the modules LHDrad or SHORTrad the list of ray directions i e the number of rays and their coordinates relative to the base axis system can be specified with e g character radraystar f A80 b 80 n List of relative ray directions amp c0O x1 1 x2 1 x3 1 oktaeder 3 tetraeder 4 cube 4 amp ci ikosaeder 6 dodekaeder 10 oktaeder Examples for allowed values are o x1 N 1 one single ray along x1 axis not enough to specify fluxes in all directions o x2 N 1 one single ray along x2 axis not enough to specify fluxes in all directions o x3 N 1 one single ray along x3 axis not enough to specify fluxes in all directions o oktaeder N 3 default octahedron 84 5 CONTR
26. export RHDBPATH RHDBSRCPATH RHDBPATH export RHDBSRCPATH HDW HDWPATH FORTRANDISK hd hdw export HDWPATH HDWSRCPATH HDWPATH export HDWSRCPATH gt DUST DUSTPATH FORTRANDISK hd dust export DUSTPATH DUSTSRCPATH DUSTPATH export DUSTSRCPATH 4 MHD MHDPATH FORTRANDISK hd mhd export MHDPATH MHDSRCPATH MHDPATH export MHDSRCPATH mean MEANPATH FORTRANDISK hd mean export MEANPATH MEANSRCPATH MEANPATH export MEANSRCPATH Architecture dependent directories for object file and executables TIMEEXEPATH TIMEPATH MAC export TIMEEXEPATH CONEXEPATH CONPATH MAC export CONEXEPATH UIOEXEPATH UIOSRCPATH MAC export UIOEXEPATH STREXEPATH STRPATH MAC export STREXEPATH MATEXEPATH MATSRCPATH MAC export MATEXEPATH GASEXEPATH GASPATH MAC 5 export GASEXEPATH EOSEXEPATH EOSSRCPATH MAC export EOSEXEPATH OPTAEXEPATH OPTAPATH MAC export OPTAEXEPATH HSTEXEPATH HSTPATH MAC export HSTEXEPATH RADEXEPATH RADPATH MAC export RADEXEPATH RHDEXEPATH RHDPATH MAC export RHDEXEPATH RHDBEXEPATH RHDBPATH MAC export RHDBEXEPATH HDWEXEPATH HDWPATH MAC export HDWEXEPATH DUSTEXEPATH DUSTPATH MAC export DUSTEXEPATH MHDEXEPATH MHDPATH MAC export MHDEXEPATH MEANEXEPATH MEANPATH MAC export MEANEXEPATH This script can be executed with HOME bin setarcdeppaths sh This line can be put e g into the bashrc file Some line
27. flag ierr of the routine uio dataset rd such errors can be avoided This way it is unnecessary to know exactly of how many time steps the model file consists IDL gt uio_openrd nc modelfile outstr ierr IDL gt i 0 IDL gt repeat begin amp IDL gt ful uio dataset rd modelfile channel nc ierr ierr outstr err msg amp IDL gt if ierr eq 0 then begin amp IDL gt hier Daten bearbeiten oder in anderer Variable speichern IDL gt i i 1 amp IDL gt endif else begin amp IDL gt print IDL gt Reached EOF amp IDL gt endelse IDL gt endrep until ierr ne 0 or EOF nc IDL gt uio_closrd nc To read a number of entries from a list of files in sequence the routine uio_datasetlist_rd pro see Sect 4 7 3 is appropriate 7 2 3 Loading the Equation of State IDL gt eosfile eos dat par eosfile IDL gt tabinter_rdcoeff eosfile eos The table for the equation of state is provided in the structure EOS NOTE Always check file name and path 7 2 4 Loading the Opacity Table IDL gt opafile opa dat par opafile IDL gt dfopta opafile The opacity table will be stored as common block OPTA_COMMON NOTE Always check file name and path 7 2 5 Computation of Deduced Quantities After having read the model data FUL and the tables for the equation of state EOS and opacity OPTA COMMON see 7 2 3 7 2 4 more quantities can be calculated IDL gt eosbox ful eos eos opa
28. for AIX Version 8 1 1 Library Particularly useful hints Porting Programs from the Crays to the sP The compiler to be used for OpenMP runs is called with x1f90_r which binds the object files to the thread safe components IBM documentation For scalar code x1f90 might be sufficient Important switches are e qsuffix f f90 qsuffix cpp F90 To tell the compiler about the existence of Fortran90 and to enable the preprocessing for files with suffix F90 e qextname flush etime Append underscore to both routine names The non standard routine call_system used by the uio_mac_sun_module is recognized automatically e WF Drhd r03 The compiler does not send switches starting with D to the preprocessor but tries to interpret it itself unlike all othe compilers Therefore each switch e g Drhd_r03 has to escaped with WF as e g WF Drhd_r03 http www 306 ibm com software awdtools fortran xlfortran library littp publib boulder ibm com infocenter comphelp index jsp http os cc biu ac il documentation http os cc biu ac il documentation xlf 8 1 1 html http pcf nersc gov computers SP craytoSP htm 3 8 MFlaps Arne aon I WET Fer mS E HYD reut Optimization Compiler Switches 39 Performance test on Hitachi SRBODO at HLRS e 128x128x192 3 time steps A de narm y Ls E Cc y r Lt M enar m 300 k 2 if f d
29. information average stratification mean fluxes surface intensities at regular intervals see Sect 5 2 can be specified with e g character outfile mean f A80 b 80 n Output file name rhd mean Leaving it empty means that no file of this type is written character outform end The format see Sect 4 3 1 of the final model files can be chosen e g with character outform end f A80 b 80 n utput file format amp cO formatted unformatted unformatted Allowed values are o unformatted default fast compact possibly machine dependent output strongly recommended o formatted slow machine independent output big files 5 3 Parameter File rhd par 93 e character outconv end The conversion type see Sect 4 3 1 of the final model files can be specified e g with character outconv end f A80 b 80 n Output file conversion amp c0 ieee_4 ieee_8 crayxmp_8 native ieee 4 The allowed values depend on the machine Leaving this field empty means that the default is chosen that is build into the local UIO module If the type ieee 4 is supported which is always the case so far it should be chosen e character outform full The format see Sect 4 3 1 of the full model files can be chosen e g with character outform full f A80 b 80 n 0utput file format amp cO formatted unformatted unformatted Allowed values are o unformatted default fast compact possibly machine dependent output str
30. model rhd full O data UIO sequence of 2D or 3D snapshots large rhd mean O data UIO derived data mean flux intensity rhd out O data text human readable text output Table 12 List of all control and data files of CO5BOLD 5 1 Model Files rhd sta rhd end rhd full If the UIO scripts Sect 4 6 are properly installed you can view the contents more precisely the headers of the data entries of an UIO file with uiolook filename e g uiolook st35gm04n05 03 end gives the output slightly edited fileform uio form unformatted convert ieee_4 version 0 1 2000 11 26 amp date 02 01 2002 16 17 26 036 system craSHi machine craSHi osrelease 10 0 0 6 amp osversion UoK 4 hardware CRAY SV1 language Fortran90 program RHD character file id f A8 b 8 n File identification character description d 1 1 f A24 p 1 b 24 n File description character history d 1 20 f A80 p 1 b 80 n File history character version f A80 b 80 n Program version label dataset n RHD model date 02 01 2002 16 17 26 043 character dataset_id f A10 b 10 n Type of box hierarchy real modeltime f E13 6 b 4 n time u s integer modelitime f I11 b 4 n time step number u 1 real dtime f E13 6 b 4 n time step u s real time_out_full_last f E13 6 b 4 n Time of last output of full model u s real time_out_mean_last f E13 6 b 4 n Time of last output of averaged data amp u s label box date 02 01 2002 16 17 26 049
31. n_radband B5 integer n raditer 82 integer n radmaxiter 83 integer n radminiter 82 integer n_radphi 84 integer n_radsubray 84 integer n_radtaurefine integer n_radtheta 84 integer n_radthickpoint integer n_viscellsperchunk integer plustimestep 88 integer starttimestep luminosity 114 opacities 76 process management radiation transport reading in IDL real abux 94 real abuy 94 real c radkappasmooth 94 real c coredrag 75 real c courant 89 real c courantmax 89 real c dustOX BI real c_hptopfactor 74 real c_maxeichange 89 real c_pchange 74 real cputime real cputime_remainlimit real c_radcourant 90 real c_radcourantmax 90 real c_radhtautop 74 real c_radimplicitmu 85 real c_raditereps 85 real c_raditerstep 85 real c_radmaxei change 90 real c_radthintimefac 90 real c_radtintminfac 95 real c_radtsmooth 95 real c_radtvis 86 real c_radtvisdtau S6 real c_schange 73 real c_slopered 77 real c_tchange 74 real c_tsurf 74 real c_visartificial 79 real c_visbouna 78 real c_viscourant real c_viscourantmax 1 real c_visdrag 78 real c_visneul 94 real c_visneu2 94 real c visprturb 79 real c_vissmagorinsky 79 real c_vistensordiag 79 real c_vistensordiv 80 real c_vistensoroff 80 real dtime_incmax S9 real dtime_max 88 real dtime_min 88 real dtime_min_stop real
32. of the guantity The suffix xmean indicates a simple average The suffix xmean2 indicates the root mean sguare average note the simple average is not subtracted Some entries e g ferb xmean have a hidden b in their name have one element more e g 121 instead of 120 than most of the others and are characterized by the ds keyword see Table 8 These quantities are located at the cell boundaries in contrast to the usual cell centered quantities Clearly there are also two sets of axes e g xc3 and xb3 corresponding to the cell or boundary centered quantities Note The total energy flux can be written as sum feb total feipkgvrb feipbtfekbtfegbtfevbtferb The flux fepb is already part of feipb 5 3 Parameter File rhd par The parameter file rhd par also has the UIO format But it will be usually Fortran formatted ASCII It contains a list of parameter entries which are collected in groups to make it easier to find an entry Otherwise the order is arbitrary except for the very first fileform uio entry If there are more than one entry with the same name the first occurence will be used by COSBOLD But the doubling of entries is strongly discouraged because it will almost certainly lead to confusion at some time In addition the IDL routine to read the parameter file will fail with an error message Additional entries can be added if the names differ from the standard ones described below Theses entries will be i
33. only how nicely I4TEX can display formulae The 3D hydrodynamics equations including source terms due to gravity are the mass con servation equation Op Opvl Opv2 p T 0 1 Ot Ox1 0x2 0x3 1 the momentum eguation pul pul vl P 9 pul v2 pul v3 pgl pv2 pv2v1 pv2v2 P pv2 v3 pg2 2 Ot pu3 ont pu3 v1 2 pu3 v2 0x3 pv303 P p 93 and the energy eguation Opeik Olpeikt P vl peik P v2 i peik P v3 t dxl i 912 23 p gl v1 g2 v2 g3 v3 3 In addition there are equations for the 3D tensor viscosity and the non local radiation transport 2 2 A collection of thermodynamic relations M Steffen AIP 2 2 1 Basic thermodynamic equations Differential relations p de Tds dp 4 p where e is the internal energy 1 dh Tds dp 5 p where the specific enthalpy h is defined as h e p p 6 This implies 5 77 Y 5 r 9 e T 2 2 A collection of thermodynamic relations M Steffen AIP 2 2 2 Definition of often used thermodynamic coefficients ar ar ais P NOT NOT amT de Os v 5 T i Gl t2 Definitions of further thermodynamic coefficients Olnp XT OMT Olnp E K Como UL 525 ar XT OlnT Xp Definition of specific heats a Os Q EN Se o Xp It can be shown that 2 D 2 p p Cy aa T Kp Cp Cy Q
34. only there the switch xvector no is required This is done automatically in the Makefile inline Optimization routines that should be inlined see Sect DMSrad raytas 2 Optimization choose non default version of loop in SUBROUTINE raytas in file MSrad3D F90 See Sect Drhd shortrad formal 101 1 Optimization split loop for exp dtau computation into two loops See Sect Drhd shortrad dirl 101 21 Optimization Transpose arrays and use routine rhd shortrad dir3 for rays in x1 direction See Sect Drhd hyd entropyfix p01 1 Optimization version with masks weights See Sect 8 7 xarch native64 Produces 64 bit code optimized for local machine 13http www sun com forte fortran 47 4 UIO Data Format 4 1 Quickstart Introduction to UIO The UIO Universal Input Output routines are a set of routines in Fortran90 and IDL to manage I O of scalars arrays and a certain table type Files can be formatted or unformatted The formatted ASCII text data representation is machine independent and appropriate for human reading for short files The binary representation uses the Fortran unformatted read and write routines provides much faster I O gives smaller files and the IEEE format is a quasi standard among many platforms compilers On all machines the native binary representation can be chosen On some platforms additional conversion types are offered IEEE on most machines CRAY format an CRAYs
35. or unformatted conv ieee 4 or native crayxmp 8 call uio openwr ncout file form form conv conv Header and data block are written together with one command as e g in call uio wr ncout time time name time unit s call uio wr ncout rho 1 10 rho name density unit g cm 3 There are two different routines to close a file after reading or writing A file opened for writing is closed by call uio closwr ncout To open a file for reading only the file name has to specified File form and conversion type are determined automatically 56 4 UIO DATA FORMAT file test txt call uio openrd ncin file In contrast to the writing of an entry by one routine call the reading is performed in two separate sub steps for the header and the data part After the reading of the header e g with use uio siz module use uio nam module integer ntt character let termt 2 nttmx call uio rdhd ncin termt ntt the identifier type and dimension if any of the entry is contained in the character array termt with ntt entries and special actions may be taken The data part may be skipped with uio skipda ncin termt ntt or it can be read with call uio rd ncin termt ntt time ident If the entry is an array it may be necessary to allocate memory call uio exkeyw termt ntt dimna dimstr call uio st2dim dimstr ilow iup ndim ndim allocate rho ilow 1 iup 1 call uio
36. radiative energy transfer with a long characteristics or a short characteristics ray scheme and possibly additional 3D turbulent diffusion in individual sub steps The 3D hydrodynamics step is fur ther simplified with directional splitting The 1D sub steps are performed with a Roe solver accounting for an external gravity field and an arbitrary equation of state from a table The radiation transport is computed with either one of three modules e MSrad module It uses long characteristics The lateral boundaries have to be periodic Top and bottom can be closed or open solar module e LHDrad module It uses long characteristics and is restricted to an equidistant grid and open boundaries at all surfaces old supergiant module e SHORTrad module It uses short characteristics and is restricted to an equidistant grid and open boundaries at all surfaces new supergiant module There are preliminary versions of modules for the formation and advection of dust and the transport of magnetic fields available COSBOLD is written in Fortran90 The parallelization is done with OpenMP directives To get a brief overview you might want to look into the Quickstart Sections How to Compile CO5BOLD Sect 3 1 Introduction to UIO Sect 4 1 How to Make a Proper Parameter File Sect 5 3 1 How to Run CO5BOLD Sect 6 1 6 2 EQUATIONS 2 Equations 2 1 Basic Equations So far this section demonstrates
37. rhd end file only indicates that CO5BOLD managed to exit gracefully due to an error or in a regular way 5 5 Text Output rhd out 97 5 5 Text Output rhd out During execution expecially during the initialization phase CO5BOLD writes lots of informa tion to standard output After the header its with a block Compiler call e g pgf90 byteswapio fast Mvect sse Mcache align Minfo inline Minline rhd hyd a vg rhd hyd upwind rhd hyd pred0 rhd hyd predm rhd hyd predp rhd hyd alpha rhd h yd constanteg rhd hyd minmodeg rhd hyd minmod rhd hyd vanleereq rhd hyd vanleer rhd hyd superbeeeg rhd hyd superbee rhd hyd ppeq rhd hyd pp rhd hyd hdflux rhd _hyd_entropyfix Minline rhd rad3d raylhd rhd rad3d solve rhd rad3d solveeq rhd Shortrad operator rhd shortrad dtauop Drhd hyd roeid 101 0 Drhd r02 Drhd rO 3 DMSrad raytasi Drhd hyd entropyfix p01 1 Drhd roeid step tO1 Drhd roeid f lux tO1 Drhd vis tO1 Drhd bound tO1 Drhd shortrad step tO1 Drhd shortrad fo rmal tO1 Drhd shortrad lambda tO1 These lines were produced by the configure script see Sect 3 6 and written into the file compiler flags info which is accessed from rhd F90 via include during compilation Various modules now have a routine XXX switchinfo that prints the values of the compiler switches used during the compilation of that particular module The output can look e g like Compiler switches rhd hyd module oo IDF 0 rh
38. value o 0 1 slow adjustment o lt 0 0 not allowed 74 5 CONTROL AND DATA FILES real c pchange The inoutflow boundary condition not only controls entropy and velocity but also the pressure in the bottom layers It is locally adjusted towards the global average to damp out possible instabilities The adjustment rate can be specified e g with real c pchange f E15 8 b 4 amp n Rate of pressure change for open lower boundary u 1 120 real c_tchange In the case of a transmitting upper or outer boundary the temperature of the material streaming into the model is adjusted with a rate given e g by real c_tchange f E15 8 b 4 amp n Rate of temperature change for open upper boundary u 1 0 3 real c_tsurf In the case of a transmitting upper or outer boundary the temperature of the material streaming into the model is adjusted towards a temperature teff c_tsurf This temper ature can be specified as fraction of the effective temperature e g with real c_tsurf f E15 8 b 4 n Temperature factor for open upper boundary u 1 0 62 The value depends on where the outer boundary is located relative to the photosphere If the boundary lies at a point where the solar photospheric minimum temperature is located it can be fairly small If the boundary is far away from the photosphere of a red supergiant the value can be even smaller On the other hand if the boundary lies somewhere within the solar chromosphere even values abov
39. when the transmitting boundary condition is chosen The value can be set e g with 5 3 Parameter File rhd par 79 real c_visbound f E15 8 b 4 amp n Boundary drag viscosity parameter u 1 0 001 This extra drag force is usually not necessary and should be switched off with c visbound 0 0 5 3 8 Tensor Viscosity Control In many test problems it is not necessary to activate the 2D 3D tensor viscosity But when strong slow shock fronts are aligned with the grid the Roe solver runs into problems and at least some additional 2D or 3D viscosity is necessary And even if the Roe solver can handle sharp shocks by its own the radiation transport algorithm might cause trouble because of the enormous opacity variations across a shock front Here the tensor viscosity is useful too e real c_vissmagorinsky A turbulent viscosity of Smagorinsky type can be activated e g with real c_vissmagorinsky f E15 8 b 4 amp n Turbulent eddy viscosity parameter Smagorinsky type u 1 1 2 In many cases values around 0 5 are sufficient to stabilize the code Larger values 1 2 in the example above are only necessary for some nasty under resolved supergiant models Setting c vissmagorinsky c visartificial 0 0 skips the tensor viscosity step entirely e real c visartificial A standard artificial viscosity can be activated e g with real c visartificial f E15 8 b 4 amp n Artificial viscosity tensor parameter 1 1 2 In many c
40. 0 e e 39 3 UKAFF machine grand small model ok 44 4 UKAFF machine grand large modell 45 5 UKAFF machine ukaff large model 2e 45 6 Program schemel xv unas al Box Rok hoe Rm RR EUR oe hok ROLE AAA 102 List of Tables 1 List of source directories lll lll ses 18 2 List of old source directories 22e 18 EMT 21 4 List of all low level modules 22e 22 5 List of all old modules 23 c CL 49 Hg TTL II ITE 49 PLC E 50 9 DIO Fortran90 files 53 10 Contents of uio base module f90 e 53 11 Contents of uio mac module 54 12 CO5BOLD control and data files lll 63 13 Radiation transport parameters 4 2 2 22 2s 82 1 Introduction CO5BOLD nickname COBOLD is the short form of COnservative COde for the COmputa tion of COmpressible COnvection in a BOx of L Dimensions with 1 2 3 It is used to model solar and stellar surface convection For solar type stars only a small fraction of the stellar surface layers are included in the computational domain In the case of red supergiants the computational box contains the entire star CO5BOLD solves the coupled non linear eguations of compressible hydrodynamics in an ex ternal gravity field together with non local freguency dependent radiation transport Operator splitting is applied to solve the eguations of hydrodynamics including gravity the
41. 000 0 000000 Main 2 3 ray direction 3 0 707105 0 000000 0 707109 Main 3 3 ray direction 3 0 707105 0 000000 0 707109 Time step ratio dtime dtime rad 1 750409E 01 dtime rad drhoei limit this all 6 667E 02 2 510E 03 6 276E 02 0 000E 00 n_subdtime 2 minmax T 1 116348E 03 1 630658E 05 Main 1 3 ray direction 1 1 000000 0 000000 0 000000 Main 2 3 ray direction 3 0 000000 0 707105 0 707109 Main 3 3 ray direction 3 0 000000 0 707105 0 707109 Time step ratio dtime dtime rad 1 836970E 01 dtime_ rad drhoei limit_this all 6 353E 02 3 188E 03 7 971E 02 6 276E 02 5 5 Text Output rhd out n subdtime 8 minmax T 1 118540E 03 1 630164E 05 Main 1 3 ray direction 2 0 707105 0 707109 0 000000 Main 2 3 ray direction 2 0 707105 0 707109 0 000000 Main 3 3 ray direction 3 0 000000 0 000000 1 000000 Time step ratio dtime dtime rad 1 791384E 01 dtime rad drhoei limit this all 6 515E 02 3 894E 03 8 884E 02 4 814E 03 End of rhd_shortrad_step A simulation ends with e g Time step number itime 49050 time 2 6823742Et08 t job 1 273407E 06 2 End Computation SES LS SS gt Exit information Requested number of time steps done Exit status 0
42. 000E 01 0 200000E 01 0 300000E 01 0 400000E 01 0 500000E 01 0 600000E 01 0 700000E 01 0 800000E 01 0 900000E 01 0 100000E 02 0 110000E 02 0 120000E 02 0 130000E 02 0 140000E 02 0 150000E 02 0 160000E 02 0 170000E 02 0 180000E 02 0 190000E 02 0 200000E 02 table f77table d 1 5 1 7 f 1X b 1 n test table to test the table routines integer inti f 15 b 4 n Integer 1 Spalte real reali f F5 1 b 4 n Real 2 Spalte character chari f A16 b 16 n Char 3 Spalte real real2 f E13 6 b 4 n Real 4 Spalte integer int2 f 15 b 4 n Integer 5 Spalte inti reali chart real2 int2 1 2 0a 0 100000E 02 1 2 4 0 ab 0 200000E 02 2 3 6 0 abc 0 300000E 02 3 4 8 0 abcd 0 400000E 02 4 5 10 0 abcde 0 500000E 02 5 6 12 0 abcdef 0 600000E 02 6 7 14 0 abcdefg 0 700000E 02 7 4 3 Structure of UIO Files 4 3 1 Data Representation ASCII or Binary While opening a file for writing the file format formatted or unformatted and the conversion type native e g ieee 4 have to be specified The formatted ASCII data representation allows I O independent of platform or compiler It is useful for parameter files which can be read and edited by hand for the direct inspection of data the transfer between very different systems or for the import of data which exist e g in a table format From the specified conversion type only the default output format for numbers e g E13 6 for 4 byte reals is determ
43. 01 rhd hyd gravcorr p01 rhd hyd roeid 101 28 rhd hyd upwind p01 rhd r01 rhd r02 rhd r03 rhd rad3d dir t01 rhd rad3d fromray 101 rhd rad3d r02 rhd rad3d solve tO1 rhd rad3d step tO1 rhd rad3d toray 101 rhd roeld flux 101 rhd roeld flux t01 rhd roeld slope 101 rhd roeid step t01 rhd shortrad diri 101 3 112 rhd shortrad dir 102 rhd shortrad dtauop 101 2 rhd shortrad dtauop 102 rhd shortrad formal 101 32 rhd shortrad formal t01 rhd shortrad operator 101 31 rhd shortrad operator 102 rhd shortrad step t01 rhd vis density p01 rhd vis t01 timing c factor configure script control variables F90 BASEPATH F90 COMPILER F90 DEBUG F90 DUST F90 LHDRAD F90 MACHINE F90 MHD F90 MSRAD 25 F90 PARALLEL F90 POSTFLAGS D4 F90 PREFLAGS F90 SHORTRAD 25 craSHi Cray B5 E B7 111 data format dust entropy fix environment variables EOS equation of state fluxes formatted Fortran Hewlett Packard Hitachi HP hydrodynamics 6 hydrodynamics routines IBM IL 47 105 ET UIO routines uio data 60 uio datasetlist rd uio dataset rd 6T uio init INDEX inlining alpha Cray VX1 IBM Intel compiler on Linux PGI compiler on Linux SGI Sun Intel Itanium LHDrad Linux little endian 35 40 41 49 Loadleveler magnetic fields makefile configure script UIO rad BJ ET
44. 1 1 during compilation this is done by the configure script if the environment variable F90_DUST is set to 1 see the description of the variable in Sect 2 put arrays specifying the initial conditions of the additional density into the start model as real quc001 real quc002 3 select a proper model describing dust or molecule formation in the parameter file with character dustscheme e rhd box bmag01 in rhd box module F90 and rhd F90 rhd box b magnetic 01 Category feature activation COSBOLD can handle magnetic field arrays if this compiler switch is set Values o default no handling of magnetic field arrays o handling of magnetic field arrays is activated To actually account for magnetic fiels in a simulation it is necessary to 1 set the switch Drhd box bmag01 1 during compilation this is done by the configure script if the environment variable F90 MHD is set to 1 see Sect 3 6 2 put arrays specifying the initial conditions of the boundary centered magnetic field arrays into the start model as real bbi real bb2 real bb3 3 select an hydrodynamics scheme that is able to handle magnetic fields in the parameter file with character hdscheme Hydrodynamics Roe solver e rhd roeid slope 101 in rhd hyd module F90 rhd roe 1 dimension slope loop 01 Category feature activation When this compiler switch is set a new extra stabilization mechanism can be activated If on
45. 129 SHC formal dirloop 16024 133010 687500 8 300717 SHC formal exp 28014 9014 280273 0 321778 99 100 5 CONTROL AND DATA FILES SHC formal exp expl2t 28014 3950 409912 0 141016 SHC formal dir3 36083 63222 242188 1 752134 SHC time dir3 36083 19676 009766 0 545299 SHC formal limitei 16024 1306 419922 0 081529 SHC step dtime final 16024 12730 129883 0 794441 SHC formal dir2 11959 22205 000000 1 856761 SHC time dir2 11959 5731 149902 0 479233 SHC formal diri 4044 10674 429688 2 639572 SHC time diri 4044 1893 760010 0 468289 SHC step dtime final output 181 8620 870117 47 629116 Radiation trans output only 20 335 929993 16 796499 In this example the value for the overall time rhd code 410 830017 sec is not useful because of an overflow in the counter However it is evident that the radiation transport consumes most of the time 243729 000000 sec followd by the hydrodynamics routines 48944 660156 sec and the tensor viscosity routines 25446 400391 sec Some of these values are split further 101 6 Running a Simulation 6 1 Quickstart How to Run CO5BOLD The generation of a start file and the modification of the parameter file is a somewhat complex process In short the following steps have to be performed 1 2 Produce an executable rhd exe see Sect 3 1 and put it into your working directory Choose a start model e g the final model of an earlier simulation or a model produced
46. 20 f E13 6 p 4 b 4 amp n Magnetic field 3 amp u G ei xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 n Internal energy amp u erg g rhoei xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Internal energy amp u erg cm 3 rhoek xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Kinetic energy amp u erg cm 3 rhoeg xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Gravitational energy amp u erg cm 3 t xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 n Temperature u K t xmean4 d 1 1 1 1 1 120 f E13 6 p 4 b 4 n Temperature amp u K t xmeankapparho d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Temperature amp u K p xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 n Pressure amp u dyn cm 2 s xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 n Entropy amp u erg K g rhos_xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Entropy amp u erg K cm 3 gammal xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n 1st Adiabatic Coefficient amp u 1 gamma3_xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n 3rd Adiabatic Coefficient amp u 1 delta xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Expansion coefficient amp u 1 kapparho_xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Absorption Coefficient amp u 1 cm quc001 xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Number density of CO amp u 1 cm 3 rreal rhovb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp real real n Mass flux amp u
47. 3 6 The calls to the timing routines that would be executed in parallel are removed by not setting the appropriate compiler macros see Sect 3 7 In addition the switch rhd shortrad dir 102 see 3 7 might be set according to experience about performance enhancements The user has to find optimum values for the parameters n hydcellsperchunk for the Roe solver module see Sect 5 3 7 and n viscellsperchunk for the tensor viscosity module see Sect to optimize the size of the chunk given to one thread per time So far only for the SHOR Trad module the enironment variable OMP SCHEDULE can be set before running CO5BOLD to control its OpenMP scheduling behaviour Important parallel loops in the SHORTrad module have a SCHEDULE RUNTIME modifier that allows this external control The old default is achieved by not defining the variable or by setting export OMP SCHEDULE STATIC 1 On some machines e g Intel Xeon with Linux and PGI compiler a dynamic scheduling activated with export OMP SCHEDULE DYNAMIC 1 is advantageous The size of the individual chunks might be set to larger values than 1 in the examples above The optimal value has to be found empirically A good starting point is number of grid points in 1D Number of treads which gives for a model with 171 grid points on a 4 processor machine export OMP NUM THREADS 4 export OMP SCHEDULE STATIC 43 The behaviour of the other modules is not affected The number o
48. A7 AIN 47 A wae a ee E a E E ee a 47 4 3 Structure of UIO Files 2 ee 48 4 3 1 Data Representation ASCII or Binary a 48 432 Data File Structural o soe soa ah llle 49 4 3 3 ables 6 som odo Romo e oh om REG eR a e d xm m ue e Ro 50 ETT 52 44 Files amp Directories dz Paths 2 a a 52 Ano Hortran9O s a 4 8 x back ous ee a Gesehen Bulk e doors ede ure dna Paus 53 AE AAN 53 4 5 2 Use of UIO Modules im Fortran90 e 54 A ake ta AA Bai Lio Men ARA Sa Dp Ten tin even IN Batan data c 54 a Se ete Ainan ete sangan ee ee KS 55 Bete dee iy eat in Be Me qc Xx 56 4 6 1 Installation of UIO UNIX Scripts e 56 4 6 2 Quick Examination of Files uiolook a 57 4 6 3 Transformation of Files uiocat CONTENTS 4 6 4 Information about Conversion Types uioinfo 4 7 IDL UIO Routines 4 71 Initialization of UIO Routines under IDL 4 7 2 Reading Data with uio_data pro lee 4 7 3 Reading Data with uio dataset rd pro or uio datasetlist rd pro 5 Control and Data Files 5 1 Model Files rhd sta rhd end rhd fulll o o o e T been NA a Ree BARA ae dE e CP 5 2 2 Contents of Individual rhd mean File Entry WIEN Pee we bbe EGS SES ERS en TN PIE aan 5 3 2 Header 5 3 3 Fundamental Model Parameters 0 0000008 ee 5 3 4 Boundary Conditions 0 5 3 5 Equation of Stat
49. Array 1 1 200 XB1 FLOAT Array 141 1 1 XB2 FLOAT Arrayl1 141 1 XB3 FLOAT Array 1 1 201 RHO FLOAT Array 140 140 200 EI FLOAT Array 140 140 200 V1 FLOAT Array 140 140 200 V2 FLOAT Array 140 140 200 V3 FLOAT Array 140 140 200 Spatial axes XC1 XC2 XC3 XB1 XB2 XB3 The indices stand for 1 x 2 y 3 z C for the grid cell centre B for the boundaries Most of the quantities are defined for the cell centres In case of doubt this can be found out by checking the array dimensions The substructure FUL EOS contains important quantities like the temperature T gas pressure P entropy S Structure lt 826a03c gt 6 tags length 109760000 refs 2 P FLOAT Array 140 140 200 DPDRHO FLOAT Array 140 140 200 DPDEI FLOAT Array 140 140 200 T FLOAT Array 140 140 200 DTDEI FLOAT Array 140 140 200 S FLOAT Array 140 140 200 The substructure FUL OPA only contains the opacity KAPPA 7 4 More IDL routines 109 Structure lt 826a5b4 gt 1 tags length 15680000 refs 2 KAPPA FLOAT Array 140 140 200 7 4 More IDL routines In the directory IDL rhdpro a lot of useful routines can be found which can be used for further processing and visualisation of COSBOLD data For the visualisation of 2 D models or 2 D data slices in general we recommend plotfield pro With combox pro unfortunately not commented yet further quantities can be calculated Furthermore we currently develop a widget base
50. Box can be set with this parameter To avoid numbers that do not fit into a 4 Byte real the luminosity per volume has to be specified as e g in real luminositypervolume f E15 8 b 4 n Luminosity per core volume amp u erg cm 3 s 4 5E 02 Reference volume is 4 3103 If this parameter is set to a value of 0 0 or below the entropy of the material within the core defined by as all cells within radius r0 grav is adjusted instead e real s inflow The entropy of the material streaming through an open boundary of type inoutflow into the model can be specified e g with real s_inflow f E15 8 b 4 n Entropy of core material amp u erg K g 3 25E 09 In the case of a central potential the entropy in a sphere with radius r0 grav is adjusted towards this entropy value In both geometry supergiant as well as solar this value is very important as it finally but indirectly determines the luminosity and effective temperature of the star A value of 0 0 default or below disables this energy input e real c schange The entropy s inflow of the material in the bottom layer solar case inoutflow boundary condition or the central region of the model supergiant case is not just set to the specified but adjusted towards it The adjustment rate can be controlled with e g real c schange f E15 8 b 4 amp n Rate of entropy change for open lower boundary u 1 0 3 Guide values are o 1 0 fast adjustment o 0 3 typical
51. CO5BOLD User Manual Bernd Freytag Matthias Steffen Sven Wedemeyer B hm Hans G nter Ludwig May 3 2004 2 CONTENTS Contents 5 6 PRINTER 6 2 2 A collection of thermodynamic relations M Steffen AIP 6 4o Won p wo oo Bee de CRT See eta ra 6 ore 7 Ox hoe Hey Hi eb Red beer de X Bere eG 7 ef bine a POR wo ook BRE AS T 2 2 5 Ideal gas with constant specific heats polytropic gas 11 13 bobo Gute dede d aii aout Gare oy ded os had 13 PPP 13 pe ae Ghana capa ene Geel ee dase Go ee 16 Dh Te tae En Tee ete 16 2 5 Fortran biles s e i 2o x woe a A Ee oe ee Rpp de ae oa eos 21 3 6 Configure Script s s sa so o UR hob EOS Y 3 oO na 24 DEED 25 rrr 33 3 8 1 General OpenMP settings n 34 3 8 2 General Inlining 2222s 34 3 8 3 Cray OVI s soa ok Row Rmo om m hU Re RO d d GR ee de pos 35 pe eM Aid Ge a eos GG Ey ed ene Go Gn ee 35 3 8 5 Hewlett Packard V2500 2222s 36 3 8 0 Hewlett Packard Itanium 2 22e 36 3 8 7 Hitachi SR8000 ok 37 9 95 9 IBM ok ke cs kk o Rx ox 9 ee ba oe an an ana 38 Ba SA san an aan ERA AAN NA es Gal RA an a GT e 40 seig Ge ae ae Fhe Dena an Ge Bee ee Al 3S1 NEC SX s a oboe an a oA dom Sk we GER BOR EA we Boh RO mos Bod 42 3 8 12 SGI Oria 43 3 8 13 SGI Origin 2000 3800 at UKAFF lt 43 Pu Lue RM ee ees ect ee a ee E eee EE 44 4 UIO Data Format
52. D or the UIO UNIX scripts the respective configure script takes care of this step Many compilers generate module information files with suffixes like M mod or kmo To clean up information files with other suffixes they have to be included in the cleaning step Calling examples make make UIO make UIO F9OFLAGS g make clean 4 5 Fortran90 55 make cleanall make remove make removeall A section of a typical makefile using the UIO routines may be Compiler options F90C f90 F90FLAGS Libraries UIOMAC uio mac sun module Dependencies of exe files on object files and libraries test exe test o F90C F90FLAGS o 0 UIOPATH f90 uio base module o UIOPATH f90 test o UIOPATH f90 UIO test f90 F90C c F90FLAGS M UIOPATH 90 test f90 Dependencies on used modules UTOPATH 90 UIO cd UIOPATH f90 make UIO F9OFLAGS F90FLAGS 4 5 4 Sample Calls of Fortran UIO Routines The needed modules have to be declared by a use statement like use uio_bulk_module In the initial phase of the program the UIO routine package has to initialized by exactly one call of the uio_init routine with the name of the program as optional parameter call uio init progrm uiotst The internal list of logical I O unit numbers may be changed with calls of uio chunit and uio chconv A file can be opened for writing with file test txt form formatted
53. M ller Atmosphere is chosen as the following real world example for a file with a table in UIO format fileform uio form formatted convert ieee 4 version 0 0 1996 10 29 amp 4 3 Structure of UIO Files date 20 Feb 1997 18 40 45 system Sun0S machine saturn osrelease 4 1 3 osversion 3 hardware sun4m language IDL 4 0 program by hand character description d 0 3 f A80 p 1 b 80 d 13 Nov 1996 18 29 48 Holweger Mueller Atmosphere Hartmut Holweger amp Edith Mueller 1974 Solar Physics 39 19 30 table II empirical solar temperature stratification to fit solar spectral lines and limb darkening character history d 0 3 f A80 p 1 b 80 d 13 Nov 1996 18 29 52 Holweger Mueller Atmosphere from 1974 uio form 13 Nov 1996 18 29 52 conversion type added 20 Feb 1997 18 21 01 xi gt vmicro 20 Feb 1997 18 23 43 real teff f F6 1 b 4 n effective temperature u K texa T eff 5780 0 table atmosphere d 1 7 1 29 f X b 1 n Holweger Mueller Atmosphere amp cO Hartmut Holweger amp Edith Mueller 1974 Solar Physics 39 19 30 table II c1 Teff Sun 5780K real tauross f E9 2 b 4 n optical depth Rosseland u 1 real tau5000 f F7 3 b 4 n optical depth lambda5000 u 1 t 1og10 real t f F7 0 b 4 n temperature u K real pgas f F6 3 b 4 n gas pressure u dyn cm 2 real pel f F6 3 b 4 n electron pressure u dyn cm 2 real vmicro f F4 2 b 4 n microturbulence u km s real q f F8 5 b 4 n Ho
54. O package in IDL comes as a list of routines with names quite similar to the Fortran90 version Instead of using global variables as in Fortran90 there are now common blocks in Include files 4 7 IDL UIO Routines 59 3 2k 2k ak ak ak dad ale od ale ole a ok ok oko ok ale e ad ade ale oko aka od ale ole kakak kakak ok kok Ee kok aka ak a F kakak ak ak ak ad ak okak kakak aka kok ok ak a ad ak ak ok ak ak Routines functions uio pro r y adkey1 adkey2 adkey3 chconv chpos chunit closrd closwr cpentr d data most important user routine comfortable useful Add one keyword to term table keyword value no link character Add one keyword to term table with keyword value Add one keyword to term table with keyword value or default Actualize list of conversion types for all channels Give current file position or jump to specified position Initialize store and actualize a list of free and occupied unit numbers Close file after reading Close file after writing Copy entry from one file to another Read data from uio file s in quasi direct access mode Handle uio file s in quasi direct access mode dataset rd Read uio file and put data into anonymous structure uio datasetlist rd Read data from list of files and put it into an structure deform dim2st exkeyw exitrm exmtrm filcon getenv init meitrm memtrm mkcvls nc2nt
55. OL AND DATA FILES n time amp u s time step number amp 1 x1 coordinates of Bog nou gd cell u cm amp of 2x2 coordinates cell u cm amp B U 4 2x3 coordinates of cell u cm amp of x1 coordinates cell u cm amp of n x2 coordinates cell u cm amp 4 of n x3 coordinates cell u cm amp p 4 b 4 amp n Density amp u g cm 3 amp n Velocity u cm s amp n Velocity u cm s amp n Velocity u cm s p 4 b 4 amp n Velocity u cm s p 4 b 4 amp n Velocity u cm s p 4 b 4 amp n Velocity u cm s 6 p 4 b 4 amp n Mass Flux x1 u g cm 2 s 6 p 4 b 4 amp n Mass Flux x2 u g cm 2 s 6 p 4 b 4 amp n Mass Flux x3 u g cm 2 s p 4 b 4 amp n Magnetic field 1 amp u G p 4 b 4 amp n Magnetic field 2 amp u G p 4 b 4 amp n Magnetic field 3 amp u G x1 x2 x1 x2 xd amp amp amp centers amp centers amp centers amp boundaries amp boundaries amp boundaries amp 5 2 real real real real real real real real real real real real real real real real real real File with Additional Data rhd mean 67 bci xmean2 d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Magnetic field 1 amp u G bc2 xmean2 d 1 1 1 1 1 120 f E13 6 p 4 b 4 amp n Magnetic field 2 amp u G bc3 xmean2 d 1 1 1 1 1 1
56. OL AND DATA FILES o tetraeder N 4 tetrahedron o cube N 4 o ikosaeder N 6 icosahedron o dodekaeder N 10 dodecahedron o list 01 list 01 3 Choose ray systems from a list oktahedrons tetrahedrons If character radraybase isset to unity the rays will only be aligned to the axes or diag onals and thus avoid the time consuming interpolation step of the short characteristics method Several other choices are possible which are meant for test purposes only Choosing one of the five Platonic solids Ops German Greek names only so far means that the 3 to 10 rays are equally distributed over the solid angle from the center to each corner of the respective solid integer n radtheta Using the MSrad module the ray directions have to specified in a different way The number of ray sets in theta direction can be chosen with e g integer n radtheta f 14 b 4 amp n NTHETA Number of ray sets in theta direction c0 2 2 integer n radphi Using the MSrad module the number of ray sets in phi direction can be set e g with integer n radphi f I4 b 4 amp n NPHI Number of ray sets in phi direction c0 2 2 integer n radsubray Using the MSrad module the number of rays per cell with the same direction can be specified e g with integer n radsubray f I4 b 4 n KPHI Number of rays per cell c0 2 2 integer n radthickpoint With the MSrad module the lower part of the model can be computed in diffusion approxi
57. TH JEOSEXEPATH uiolol 90 exe Unix shell scripts Figure 1 Old directory scheme 17 rad hdrad SRADPATH MAC jRADEXEPATH 18 3 PROGRAM FILES INSTALLATION COMPILATION Paths Abb Description 4HOME for con f90 CON constants and units HOME for dust 90 DUST source terms due to dust or molecules HOME for eos f90 EOS equation of state HOME for hd mhd MHD MHD routines HOME for hd rhd RHD main rhd routines hydro Bernd s radiation transport HOME for hd rhdb RHDB basic rhd routines HOME for mat str STR string handling HOME for opa opta OPTA opacities HOME for rad hdrad RAD Matthias radiation transport HOME for uio f 90 UIO I O routines HOME for time f90 TIME timing routines Table 1 List of source directories with path and file name abbreviation and a short description Paths Abb Description HOME for mat str STR string handling HOME for uio 90 UIO I O routines HOME for eos f90 EOS equation of state HOME for rad hdrad RAD opacities Matthias radiation transport HOME for hd rhd RHD main rhd routines Table 2 For historical reasons list of old source directories with path and file name abbreviation and a short description setarcdeppaths sh or setarcdeppaths csh These variables are now ignored by the configure script that produces the Makefile to generate the CO5BOLD executable rhd exe Howeve
58. TH pointing to this directory The path to the shell scripts and to the man pages should be added to your shell path variables e g in one of the login scripts C shell cshrc 4 110 setenv UIOPATH HOME uio setenv PATH PATH UIOPATH bin setenv MANPATH MANPATH UIOPATH man dp dee Korn shell kshrc yio UIOPATH HOME uio export UIOPATH PATH PATH UIOPATH bin export PATH MANPATH MANPATH ULOPATH man export MANPATH xk 4 5 Fortran90 53 File contents uio base module f90 Collection of basic modules uio mac module f90 Possibly machine dependent routines Standard version all machines uio mac ieee module f90 Machine dependent routines IEEE format uio mac sun module f90 Machine dependent routines Sun SGI Linux HP uio mac intel module f90 Machine dependent routines Linux Intel little endian uio mac crayts module f90 Machine dependent routines CRAY uio mac crayxmp module f90 Machine dependent routines CRAY uio mac decalpha module f90 Machine dependent routines alpha uio mac hitachi module f90 Machine dependent routines Hitachi uio bulk module f90 Main part of UIO routines uio filedef module f90 Standard file descriptors and labels uio table module f90 Table manipulation routines uio var module f90 definition and handling of UIO flexible variable uio varfile module f90 definition and handling of file structure of UIO flexible var
59. _data mode close family mod2 uio_data mode allclose of UIO files 4 7 3 Reading Data with uio_dataset_rd pro or uio_datasetlist_rd pro For a detailed description of how to handle UIO files in IDL see Sect With the new IDL routines uio_struct_rd pro uio dataset rd pro and uio_datasetlist_rd pro files are not read entry by entry anymore but in larger blocks or data sets With uio_struct_rd all entries in a file are read and put into an IDL structure variable This is appropriate for the COSBOLD parameter file or for the UIO table file in Sect 4 3 3 e g par uio_struct_rd st35gm04n05_03 par atm uio struct rd holmu atm When groups of entries in an UIO file are properly marked with label dataset and label enddataset delimiters confer the example in Sect 5 1 each group can be accessed with uio dataset rd The first block can be read with ful uio dataset rd st35gm04n05 03 ful1 Or ful uio dataset rd st35gm04n05 03 full ndataset 0 Dataset number i 1 counting starts at zero can be read with ful uio dataset rd st35gm04n05 03 full ndataset i If a dataset with that number does not exist an empty structure is returned In this case when called with additional keywords like ful uio dataset rd st35gm04n05 03 full ndataset i outstr outstr ierr ierr an error message is returned in outstr and ierr is set to a value larger than 0 To read all entries in a list of files in sequence t
60. ar BAKDIR OUTFILE par mv rhd sta BAKDIR OUTFILE sta mv rhd end BAKDIR OUTFILE end mv rhd full BAKDIR OUTFILE full mv rhd mean BAKDIR OUTFILE mean chmod gotr BAKDIR PARFILE BAKDIR OUTFILE chmod go w BAKDIR PARFILE BAKDIR OUTFILE HEE EGE EEE EES IE EA AE EA EA fi 6 2 Running CO5BOLD on a Machine with Batch System 105 TEHETHEI HEHTETEHIERERETEHIENE RE TETERERERETEHENERETETERERETETERHENERETEHETERETETERENERETETMENERETETERENERETEHENE EIER RE RO RR Dispose modified command file for next job cp p nrhd cmd STADIR rhd1 cmd Exit loop if end of chain reached if test ACT eoc then break fi End of loop 2 2 SS SSR SS SE Ea done Submit next job if not end of chain reached if test ACT eoc then cd STADIR echo Resubmit job usr local bin qsub rhd1 job qsub rhdi job nice nohup rhdi job amp nohup rhd1 job amp else echo End of job chain fi FEE ERA AR RR Basic informations about the batch system of the ngstr m Sun clusteq are available A more detailed description of the script will come sometimes in the future e g from Sven s manual http www fysik4 fysik uu se sunkluster 106 7 DATA ANALYSIS WITH IDL 7 Data Analysis with IDL In this section some basic commands for handling CO5BOLD data in IDL are described See also Sect 71 Preparations
61. ases values around 0 5 are sufficient to stabilize the code Larger values 1 2 in the example above are only necessary for some nasty under resolved supergiant models Setting c vissmagorinsky c visartificial 0 0 skips the tensor viscosity step entirely e real c visprturb The Prandtl number for turbulent mixing can be specified e g with real c_visprturb f E15 8 b 4 amp n Turbulent Prandtl number u 1 8 0 Values between 1 0 and 10 0 appear reasonable Note that larger values lead to smaller amounts of turbulent mixing A value of 0 0 switches off the turbulent mixing terms but not the entire tensor viscosity e real c_vistensordiag The factor in the stress tensor in front of of the diagonal terms can be set with real c_vistensordiag f E15 8 b 4 amp n Diagonal factor for viscous stress tensor u 1 amp cO typically 1 0 1 0 This is not really parameter one would try to adjust The total amount of viscosity should be controlled with real c_vissmagorinsky and real c_visartificial But the param eter can be used to tentatively switch off the diagonal terms completely or to change its importance compared to the other terms 80 5 CONTROL AND DATA FILES e real c vistensoroff The factor in the stress tensor in front of of the off diagonal terms can be set with e g real c_vistensoroff f E15 8 b 4 amp n 0ff diagonal factor for viscous stress tensor u 1 amp c0 typically 0 5 0 5 This is not really
62. au coupling short characteristics module is accessed Values o 0 default The routine with the tau coupling operator is called within a loop and should be inlined o 1 The program fragment with the tau coupling operator is included No inlining necessary rhd shortrad formal 101 in rhd shortrad module F90 rhd short characteristics radiation formal loop 01 Category performance enhancement Select version of loop splitting for exp dtau computation Values o 0 default dtauhalf exp mdtauhalf expl2t mdtauhalf are computed in single loop o 1 dtauhalf exp mdtauhalf expl2t mdtauhalf are computed in separate loops This prevents the SUN1 machine Sunfire Solaris Forte 6 2 from doing some perfor mance degrading optimization rhd shortrad dirl 101 in rhd shortrad module F90 rhd short characteristics radiation direction 1 loop 01 Category performance enhancement Choose routine version for rays in x1 direction Values o 0 default Use routine with permuted indices for rays in x1 direction In this case the innermost loop index is the third array index The transposition of arrays is not needed but some machines e g SUN1 do not like this index arrangement o 1 Transpose arrays and use routine rhd shortrad dir3 for rays in x1 direction The extra step for the transposition of some arrays and the reverse procedure needs some time But now the routine with the optimum index ordering can be used
63. can be set e g with integer n raditer f I4 b 4 amp n Number of radiation transport iterations c0 10 8 5 3 Parameter File rhd par 83 For a simulation of a solar type star with comparatively long radiative time scales it will typically be set to 1 For starts with shorter radiative time scales values around 10 may be considered All three radiation transport modules understand this parameter e integer n radmaxiter The absolute maximum number of iterations can be specified e g with integer n radmaxiter f I4 b 4 amp n Maximum number of rad transport iterations c0 30 0 If more iterations are needed the computation for the current time step is stopped and resumed with a smaller one which means that the hydrodynamics and the tensor viscosity step have to be done again Usually n radmaxiter will either be set to a values somewhat larger than the recommended number of iterations n raditer or to 0 which disable the check for too many iterations completely This can be safely allowed in many cases and has the advantage that there is no need to save the initial model before calling the radiation transport module which saves some memory To disable the iteration of the radiation transport sub step set n radminiter n raditer n radmaxiter 1 All three radiation transport modules understand this parameter e character radraybase Using the modules LHDrad or SHORTrad the orientation of the base axis system can be selected
64. cit hydrodynamics communication is restricted to neighboring grid cells For those wondering the black curves do not lie between the red and green curve since more components than just radiation and hydrodynamics add up to the total time Furthermore the normalization of the execution times given by CO5BOLD is not exact 3 8 14 Sun SunFire CO5BOLD has been used on the SunFire machines firel fire2 and fire3 in Uppsala with compiler version Sun WorkShop 6 update 2 Fortran 95 6 2 2001 05 15 and later An older version was not able to compile CO5BOLD properly Information about Fortran and the Sun 3 8 Optimization Compiler Switches Machine grand 315x315x81 grid points 1 OBM band 3 time steps Per processor execution time increase 0 5 10 15 20 25 Number of processors Figure 4 UKAFF machine grand large model Machine ukaff 315x315x81 grid points 1 OBM band 3 time steps Per processor execution time increase 0 5 10 15 20 25 Number of processors Figure 5 UKAFF machine ukaff large model 45 46 3 PROGRAM FILES INSTALLATION COMPILATION compiler can be found on the Sun Forte pad under documentation Important switches are openmp Enable OpenMP fast xvector yes no General optimization On the Sun the fast option switches on more or less all optimization features of the compiler That works reasonable well However during the compilation of gasinter routines f90 and
65. ct is activated the slope can be reduced by averaging with the results from a MinMod reconstruction by setting c_slopered to a positive non zero value This value can be set e g with 78 5 CONTROL AND DATA FILES real c_slopered f E15 8 b 4 amp n Slope reduction parameter in case of strong density contrast u 1 amp c0 0 00 off default 0 02 reasonable value 0 10 large value 0 02 Typical choices are o 0 0 Slope reduction switched off Original reconstruction is used o 0 02 Moderate slope reduction in case of large density jumps o 0 10 More pronounced slope reduction in case of strong density contrast integer n hydcellsperchunk In every directional sub step neighboring 1D columns are independent from each other They can be grouped and computed in chunks of arbitrary size The approximate number of grid cells per chunk can be specified e g with integer n hydcellsperchunk f I9 b 4 amp n Number of cells per hydro chunk amp c0O 0 gt one 2D slice at a time amp c0 1 gt minimum chunk size inefficient amp c0 2500 reasonable value amp c0 1000000000 maximum chunk size inefficient and memory intensive 20000 The exact number is determined at run time to get approximately equal sizes of the individual chunks The choice of this parameter does not affect the result of the computation but the memory usage and performance Smaller and more chunks may result in an opti
66. ctivate this radiation transport module e F90 SHORTRAD SHORTrad radiation transport o 0 do not activate compile and link this module default o 1 activate this radiation transport module e F90 DUST DUST module o 0 do not activate compile and link this module default o 1 activate this source step module If this variable is set to 1 the compiler is called with Drhd box guc01 1 see Sect e F90 MHD MHD module o 0 do not activate compile and link this module default o 1 activate this magnetohydrodynamics module e F90 MACHINE Explicit machine specification This isusually not necessary use local or instead o local machine o sun4u Sun o See the header of the configure script for an up to date list o local local machine default o dummy Do not use any machine dpedendent flags but use module selections o empty Compiler flags are composed from F90 PREFLAGS and F90 POSTFLAGS only e F90 BASEPATH Path for CO5BOLD base directory o The configure script tries to determine the base directory name automatically de fault This should work if the local directory is located somewhere below hd rhd o otherwise This string is used as base directory name e g home user for Some examples can be found in Sect 3 7 Compiler Macros Some of the modules of the CO5BOLD code with suffix F90 employ compiler macros to switch between code versions during compile time Typicall
67. d analysis tool called COBOLD AT abbrev CAT which will help to work with COSBOLD data without having to write and edit own IDL code The routines are stored if available in the directory IDL COBOLDAT and have to be started with cat A more detailed documentation is planned 110 8 DOCUMENT HISTORY 8 Document history 2002 02 16 First version on the web lots of extensions and changes in between 2004 02 23 3 8 13 2004 02 23 2004 03 02 2004 03 02 2004 03 03 2004 03 04 2004 03 04 2004 03 04 2004 03 04 2004 03 04 2004 03 04 2004 03 04 2004 03 04 SGI Origin compiler settings section modified for CINES machines 3 8 12 short example for uio datasetlist rd pro usage 4 7 3 Intel compiler settings section modified LD ASSUME KERNEL 2 4 19 SGI Origin compiler settings section modified IBM compiler settings section added Document history started pretty late Dedicated OpenMP section added Dedicated inlining section added Parameter for core drag force 5 3 4 New EOS table for solar composition 5 3 5 Trademarks List of new dust files the table was split into two 3 5 Installation of UIO UNIX scripts with configure 4 6 1 111 9 Glossary 10 CO5BOLD or COBOLD is the short form of COnservative COde for the COmputation of COmpressible COnvection in a BOx of L Dimensions with 1 2 3 EOS Equation Of State RHD Radiation HydroDynamics UIO th
68. d hyd gravcorr p01 5 rhd hyd entropyfix p01 1 rhd hyd upwind p01 0 rhd hyd roeid 101 0 rhd roeld step 101 0 rhd roeid slope 101 0 rhd roeld flux 101 undefined rhd bound t01 defined rhd roeld flux t01 defined rhd roeld step t01 defined See Sect 3 7 for more information about the meaning of the values The reading of the parameter file starts with ACTION Read parameter file After a parameter is read its value is printed see Sect 5 3 The line ACTION Load EOS data indicates the start of the reading of the equation of state data It is followed by some information about the EOS table in use Similarly the line ACTION Load opacity tables indicates the start of the reading of the opacity data The information that follows is taken directly from the header of the opacity table Currently the last file to be read is the start model which is announced by ACTION Read start model and followed by some information about the start model e g the number of grid points and a new section showing the quantities actually read e g 98 5 CONTROL AND DATA FILES Properties of start model time time s xci X1 coordinates of cell centers cm xc2 X2 coordinates of cell centers cm XCS x3 coordinates of cell centers cm xb1 X1 coordinates of cell boundaries cm xb2 X2 coordinates of cell boundaries cm xb3 X3 coordinates o
69. dtime out full 9 real dtime out mean 91 real dtime start 88 real endtime 87 real gamma 94 real grav 71 BE INDEX real luminositypervolume real mass star 71 real plustine 87 real gmol 94 real r0 grav 71 real ri grav 71 real ri rad real rho min 75 real s inflow real starttime real teff 70 tensor viscosity time step rhd stop Roe solver SGI short characteristics SHORTrad Sparc 111 Sun tensor viscosity tensor viscosity routine thermodynamic relations 6 adiabatic gradient enthalpy 6 entropy internal energy 6 polytropic gas sound speed specific heats 9 time step timing statistics total energy flux UIO conversion type example format Fortran IDL makefile record length unit uio mac module UNIX scripts uiocat uio datasetlist rd uio dataset rd uioinfo uio init uiolook uio struct rd UKAFF unformatted INDEX 115 UNIX xbi xb2 64 xb3 64 xc1 64 xc2 aed
70. e 0 5 3 0 Opacities 5 3 7 Hydrodynamics Control s sce oe to e e wue pE e E E e oea E 5 3 8 Tensor Viscosity Control sore sopa kos o o E E ee ee Te aee ole ue de dat he a RR b ge AL aes oe eS 5 3 10 Radiation Transport Control oo 5 3 11 Process Time Management es 5 3 12 Time Step Control 22r 5 3 13 Input Output Control egos EUR ye eco kh ana eR Gee OE 5 3 14 Additional Information Obsolete and Test Parameters 5 4 Additional Control and Status Files rhd stop rhd cont and rhd done 5 5 Text Output rhd out ss 6 Running a Simulation 6 1 Quickstart How to Run COSBOLDJ 2s 6 2 Running CO5BOLD on a Machine with Batch System 7 Data Analysis with IDL 7 1 Preparations 12 COSBOLD Data in IDL s oe ES DRA O UK AR BSE ae awe a 7 2 2 Loading CO5BOLD Data full sta end Eq RU B Fay a RAE E HORE Btn Sh aun BOR Bk ge p E E a a d OR E TT 71 3 IDL Data Structure 7 4 More IDL routines 8 Document history Glossary 10 Trademarks 58 58 60 60 61 63 63 64 64 65 68 69 69 70 72 75 76 TT 79 80 81 86 88 91 94 96 97 101 101 102 106 106 106 106 106 107 107 107 108 109 110 111 111 4 LIST OF TABLES List of Figures 1 Old directory scheme be Bw We NO VR ee e ee Be ae a 17 2 Performance tests on Hitachi SR800
71. e Universal Input Output format It is used in CO5BOLD for parameter model mean and EOS files Trademarks AMD is a trademark of Advanced Micro Devices Inc Compaq is a US trademark of Compaq and or Hewlett Packard Company Cray is a trademark of Cray Research HP is a US trademark of Hewlett Packard Company IBM is a US trademark of International Business Machines IDL is a registered trademark of Research Systems Inc Intel Itanium and Pentium are US trademarks of Intel Corporation Linux is a trademark of Linus Torvalds NEC is a registered trademark of Nippon Electric Company PGI is a trademark of The Portland Group Compiler Technology SGI is a trademark of Silicon Graphics Solaris Sun SunOS and SunFire are US trademarks of Sun Microsystems Inc Sparc is a US trademark of SPARC International Inc UNIX is a registered trademark of The Open Group All other product names mentioned in this manual are trademarks or registered trademarks of their respective owners Index alpha AMD batch gueue 102 big endian 47 49 boundary conditions chemical composition CINES CO5BOLD Compaq 111 compilation compiler alpha Cray Hewlett Packard Hitachi SR8000 IBM Intel NEC SX 5 PGI SGI UKAFF Sun compiler macro category gasinter 101 IDF MSrad raytas rhd shortrad lambda 101 33 rhd bound t01 rhd box bmag01 rhd box gravO1 rhd box qucO1 rhd hyd entropyfix p
72. e o f90 general default e F90 PREFLAGS Compiler flags to be put at the beginning of the list Usually the list of compiler flags produced by the configure script should be pretty complete But you might want to add special switches like Bstatic to enforce static linking of libraries o No extra flags e F90 POSTFLAGS Compiler flags to be put at the end of the list Usually the list of compiler flags produced by the configure script should be pretty complete However you might want to overwrite some settings This can be done by setting this variable to a none empty value because typically a compiler should interpret the flags from left to right o No extra flags e F90 PARALLEL Parallelization scheme o scalar no parallelization default o openmp OpenMP appropriate for COBBOLD o auto auto parallelization not implemented for all machines e F90 DEBUG Debugging level o 0 No extra debugging information produced full optimization is chosen default o 1 standard debugging mode typically switch g instead of fast o 2 other debbuging or array checking modes possible if implemented for the requested machine e F9O_LHDRAD LHDrad radiation transport o 0 do not activate compile and link this module default o 1 activate this radiation transport module 3 7 Compiler Macros 25 e F90 MSRAD MSrad radiation transport o 0 do not activate compile and link this module default o 1 a
73. e tentatively support implicit radiation transport It can be activated with the parameter real c_radimplicitmu f E15 8 b 4 amp n Implicitness parameter for radiation transport u 1 amp c0 0 0 explicit 0 5 time centered 1 0 fully implicit 0 0 Allowed values are o 0 0 Fully explicit radiation transport possible with all modules o 0 0 C 1 0 Partly implicit radiation transport o 0 5 Radiation transport time centered o 1 0 Fully implicit radiation transport Values outside this range do not have much meaning The implicit transport does not work efficiently yet It does not yield significantly larger time steps than possible with a sequence of purely explicit sub time steps Additionally it turns out that the hydrodynamics runs into trouble if a too large time step still well within the Courant condition is requested e real c raditereps With activated implicit radiation transport LHDrad module only the requested convergence accuracy of the iteration can be set e g with real c_raditereps f E15 8 b 4 amp n Relative accuracy for radiation iteration u 1 amp cO Typical value 1 0E 03 2 0E 03 e real c raditerstep With activated implicit radiation transport LHDrad module only the step size of the iter ation can be restricted with e g real c_raditerstep f E15 8 b 4 amp n Step size of radiation iteration u 1 amp cO Typical values 0 7 0 81 1 0 Allowed values are
74. e 1 0 might be reasonable real c_hptopfactor In the case of a transmitting upper or outer boundary the density stratification outside the model has to be extrapolated properly Assumptions about this density affects the amount of mass flowing into the model For the extrapolation it is assumed that the density scale H scales with the pressure scale height Hp as H H c_hptopfactor real c_hptopfactor f E15 8 b 4 amp n Correction factor for surface pressure scale height u 1 0 8 Possible values are o C 0 0 No effect actually a value of 1 0 is chosen o 0 0 lt C lt 1 0 The density scale height is enlarged to account for possible effects of turbulent pressure on the scale height The density decays less rapidly with height than in an isothermal hydrostatic stratification o C 1 0 Density scale height is pressure scale height o C gt 1 0 Density scale height is smaller than pressure scale height Not really useful real c radhtautop The MSrad radiation transport module needs the specification of the scale height of the optical depth at the upper boundary e g with real c_radhtautop f E15 8 b 4 n Scale height of optical depth at top u cm 60 0E 05 5 3 Parameter File rhd par 75 e real rho min During long periods of matter infall the density at an open outer boundary can become very low To limit the decrease of the density a lower limit in the extrapolated ghost cells can be set e g with real r
75. e basic handling of the variables I O in UIO format construction and modification of variables The module uio varfile module f90 contains a type definition for a file built of UIO flexible variables together with routines for the handling of these files 4 5 2 Use of UIO Modules in Fortran90 To make the UIO routines available in a Fortran program the appropriate modules have to be specified in a USE statement At maximum five modules play a role The uio bulk module contains the main part of the UIO routines and also uses the relevant sub modules Instead of uio bulk module the module uio table module is used if the UIO table routines are needed The modules uio siz module and uio nam module contain specifications about the size of some arrays and the length of strings and the names of types and keywords respectively The module uio filedef module contains some definitions in addition to the basic UIO standard as e g the label names which delimit a data set label dataset and label enddataset A typical case for the use of UIO modules is use uio bulk module use uio siz module use uio nam module 4 5 3 Compiling and Makefiles For a certain platform it is necessary to change the name of the module file with the machine dependent routines uio mac module f90 in a Makefile for the UIO routines For this purpose the environment variable UIOMAC has to be set to the name of the appropriate routine see Sect 3 3 For CO5BOL
76. e code compiles and runs fast but crashes after a few time steps A proposed compiling sequence is LHDrad DUST MHD do not compile export F90_MSRAD 1 export F90_SHORTRAD 1 configure make 3 8 7 Hitachi SR8000 The appropriate machine dependent UIO module is uio mac hitachi module f90 The appro priate machine dependent RHD module is rhd mac hitachi module f90 conti199 Up to 199 continuation lines can be interpreted otherwise not more than 39 continuation lines are accepted limit Limits the amount of time and memory for compilation opt ss use highest possible optimization level nopredicate this option switches off a sub option activated by opt ss It is necessary to disable the predicate option because the code crashes otherwise segmentation violation The switch must appear after setting opt ss pvfunc 2 References the pseudo vectorizing mathematical function and applies the tem porary array to reference the pseudo vectorizing mathematical function omp parallel 1 parallelize based on OpenMP directives only procnum 8 generated code for 8 processors on one node orphaned 1 Checks if the regions sequentially executed contain orphaned directives dur ing run time when PROCNUME 8 is specified If a sequentially executed region contains an orphaned directive the system outputs a message and terminates the program nestcheck 1 Checks for nesting errors in parallel regions If a parallel region is n
77. e g r16 i2 are used to modify the accuracy and standard memory size of variables If an existing file is opened for reading the file format and conversion type are determined automatically from the file if the conversion type of the data in the file is among the conversion types supported by the compiler If the file has a conversion type native but is created on a machine with different internal data representation the file header might be readable but an error will probably occur during the reading of a real variable 4 3 2 Data File Structure The UIO routines only handle sequential files Each file consists of a list of entries The first entry describes the file format conversion type and the machine who is responsible for it The following entries contain data scalars and 1D 4D arrays of type integer real single amp double precision complex single precision character tables with columns of type integer real single precision or character or structuring information labels Each entry consists of the header and the possibly empty data block Each header is a list of at most 20 terms separated by blanks or linefeeds The first term is the entry type e g real see table 7 followed by an identifier This identifier should follow the standard rules for variables lowercase letters numbers underscore starting with letter It is a name as e g rho v 1 The rest of the terms come in the form keyword val
78. e header of the parameter file e g fileform uio form formatted convert ieee 4 date 01 01 2002 amp program by hand can be abbreviated to 70 5 CONTROL AND DATA FILES fileform uio form formatted convert ieee 4 which indicates that the file is in UIO form and Fortran formatted ASCII The speci fication of the conversion type convert ieee 4 is more relevant for unformatted files These terms should not be changed But it can be of interest to append e g the date of the last modification e g date 01 01 2002 character file id The file identification string character file id f A80 b 80 n File identification rhd parameter indicates the intended use of the file as parameter file for the RHD code CO5BOLD Do not edit character description The header of the file can should contain a short description of the simulation as in e g character description d 1 4 f A80 p 1 b 80 n File description Parameter file for RHD code Full size 3D Betelgeuse model 5 M Sun 650 R Sun Start with st35gm04n03 09 end 12773 171 3 Run with SHORTrad This entry is optional it can be omitted completely but it is recommended to put at least some relevant keywords into this array If you change the number of lines between 1 and 20 you have to adjust the size specification d 1 4 in the example above character history The file history has a similar purpose as the previous en
79. e i E o tat E P U A j p f B va 1 1 100 a n s j I y 1 a a00 Fi A E Eo a nmm i 900 F P ROO E _ J 103 104 os 108 chunk size my Oi 4 r a asli o ost A e A 1 lia g 4 r 1 d xa U Jou f 4 3 1 4 l ak a e gt o o egi L i ill 103 105 106 O r a e 8 na kak 252x292x188 3 time steps ea E 2 i il o oe 10 104 g chunk size r 1 Figure 2 Performance tests on Hitachi SR8000 at HLR Stuttgart For models with 128x128x192 and 252x252x188 grid cells different values for the hydrodynamics and viscosity chunk size pa rameters were used See text for more details 40 3 PROGRAM FILES INSTALLATION COMPILATION e q64 Activate the 64 bit mode In the default 32 bit mode a very small model with total memory requirements far below the 2 GByte limit might run However experienced IBM users might experiment instead with the compiler switches bmaxstack bmaxdata and qsmallstack with the UNIX commands limit or ulimit or with the settings in the header of the batch script e 03 Choose optimization level Higher levels that involve interporcedural analysis cause the compiler to stop with an error message e qarch auto qtune auto qcache auto To allow optimization specific for the local machine default in the configure script Cross compilation can be activated e g with qarch pwr3 qtune p
80. e new Courant condition etc would allow more A value value 0 0 deactivates this restriction e real c courant A typical Courant factor for each 1D hydrodynamics step can be specified with e g real c courant f E15 8 b 4 n HD Courant factor u 1 amp cO range 0 0 lt C_Courant lt 1 0 typically 0 5 0 5 From the minimum cell crossing time of a partial wave and this factor a recommendation for the next time step is computed A value of 1 0 is the upper limit which guarantees stability for some simple linear test problems Values around 0 5 are recommended for fully non linear simulations e real c courantmax A typical Courant factor for each 1D hydrodynamics step can be specified with e g real c courantmax f E15 8 b 4 n maximum HD Courant factor u 1 amp cO range C Courant lt C Courantmax lt 1 0 typically 0 9 0 8 From the minimum cell crossing time of a partial wave and this factor an upper limit for the current time step is computed If this limit is exceeded the computation is interrupted and resumed with a smaller time step based on c courant Usually this parameter should be restricted by c courant lt c courantmax lt 1 0 A value around 0 8 appears to be a good choice e real c maxeichange The relative change in internal during a single 1D hydrodynamics step can be used to restrict the time step by specifying real c maxeichange f E15 8 b 4 n maximum hydro energy change u 1 amp
81. e of the reconstruction methods VanLeer Superbee or PP see Sect 5 3 7 is activated the slope can be reduced by averaging with the results from a MinMod reconstruction by setting c slopered see Sect 5 3 7 to a positive non zero value This can improve the stability without significantly reducing the effective numerical resolution Switch values o 0 default no slope reduction o 1 slope reduction in case of expansion wave o 2 slope reduction in case of strong density contrast e IDF in rhd hyd module F90 Integer Delta Flux Category performance enhancement Number of padding cells for flux like variables This number was introduced to check whether the increase of the size of vectors for flux like quantities defined at cell boundaries can improve the performance especially on a CRAY machine The gain is marginal if present at all The parameter is usually set to zero or left undefined Values 28 3 PROGRAM FILES INSTALLATION COMPILATION o 0 default no padding cells o 1 2 3 extra padding cells e rhd hyd gravcorr pO1 in rhd hyd module F90 rhd hydrodynamics gravitation correction parameter 01 Category selection of approximation This parameter controls the way the Roe solver handles the source terms due to gravity A different choice results in different simulation results and not just in slightly faster or slower code The problem is that the original Roe solver interpretes the
82. ed Amazingly the UIO routines also used to work under PV WAVE Version 6 01 sun4 solaris sparc So far there exist three UNIX shell scripts calling Fortran routines useful to quickly examine data sets or to change the format or conversion type of files The installation of these scripts with a configure make make install sequence is described in Sect Without these scripts an UIO file in ASCII format can be examined with any text editor or with more To get an overview about the contents of a binary file the command strings 30 uio_example_file dat can be useful 4 2 Example of UIO Data File To give a first impression about the data structure here follows a simple test file which contains the header a label a couple of scalars an array and a short table fileform uio form formatted convert native version 0 1 1997 11 29 amp date 29 11 1997 21 23 39 835 system IRIX machine atlas osrelease 6 3 amp osversion 12161207 hardware IP32 language Fortran90 program uiotst label testdata n sample test data field date 29 11 1997 21 23 39 835 integer ia f I3 b 4 n This is the answer 42 complex ca f E13 6 E13 6 b 8 n This is a complex answer 0 400000E 01 0 200000E 01 48 4 UIO DATA FORMAT real da f E23 15 b 8 n precise answer 0 420000000000000E 02 real answer f F4 0 b 4 n answer u 1 42 real real2d d 100 103 200 204 f E13 6 p 4 b 4 0 100
83. ed on standard output or writes it into a file In the latter case the format change from formatted to unformatted or vice versa is possible OPTIONS Conversion type native ieee_4 machine dependent its specification is only relevant if an output file is specified with the o option and output format unformatted f Output format formatted or unformatted Its Specification is only relevant if an output file is Specified with the o option zh Help print usage of uiocat 1 List of entries to be copied E g uiocat 1 real rho uiocat 1 r al rho integer i uiocat 1 label real rho integer i uiocat 1 label real rho i Here copylist is a list separated by Each item consists of exactly two items separated by a blank No additional blanks are allowed Use copylist with as above 9 Output file name If omitted standard output is used and c and f are meaningless Suns 5 5 1 Last change 12 January 1998 1 4 6 4 Information about Conversion Types uioinfo The shell script uioinfo calls the Fortran program uioinf f90 The man page uioinfo 1V Misc Reference Manual Pages uioinfo 1V NAME uioinfo print machine dependent information SYNOPSIS uioinfo DESCRIPTION The routine uioinfo prints information about its environment and a list of possible conversion types OPTIONS Sun0S 5 5 1 Last change 12 January 1998 4 7 IDL UIO Routines The UI
84. efault versions of loops in the CO5BOLD code vectorize better and are preferred over the standard ones F Enable macro expansion Otask1 Parallelization Enable tasking in this case OpenMP Oinline3 Optimization enable high level of inlining e vector3 Oscalar3 General optimization Drhd hyd roe1d 101 1 Optimization Choose non standard set of routines for Roe solver See Sect Drhd hyd entropyfix p01 1 Optimization version with masks weights See Sect 3 7 Drhd hyd upwind p01 1 Optimization version with masks weights e Drhd shortrad operator 101 2 Optimization short characteristics operator with masks weights e Drhd shortrad dir 102 1 Optimization OMP PARALLEL statement outside of outer loop in rhd shortrad dirX 3 8 4 Compaq alpha Documentation about the Compag Fortran compiler The appropriate machine dependent UIO module is uio mac decalpha module f90 It al lows the reading and writing of files in little endian and big endian format assume byterecl Necessary for the UIO routines specify that the length of a record is measured in bytes and not in words e cpp invoke the preprocessor on all source files e inline speed V Force automatic inlining optimized for speed e 04 General optimization Drhd hyd roeid 101 0 Optimization Choose standard set of routines for Roe solver See Sect e DMSrad raytas 0 Optimization choose default version of loop in SUBROUTINE
85. el amp cO none default nosource dust simple Ol co componentO1 01 Y ci dust k3mon 01 dust k3mon 02 dust k3mon 01 Possible values are the list is hopefully expanded in the near future o none None No handling of any dust molecule density at all o nosource Skip source term step for dust molecules entirely but do the transport o dust simple 01 Simple and unrealistic dust formation model only for testing of the numerics o co componentO1 01 Simple CO formation from Matthias Steffen with one compo nent only but realistic time scales o dust k3mon 01 Simple C rich dust formation routines from Susanne H fner with one component only but realistic time scales o dust k3mon 02 Simple C rich dust formation routines from Susanne H fner with two components for dust density and free carbon density o dust moment04 c2 C rich dust chemistry 4 moments routines from Susanne Hofner e real c dustOX There are five parameters real c dustO1 to real c dustO5 to control each dust for mation scheme in detail A parameter can be given as in real c_dust01 f E15 8 b 4 n Dust parameter 1 0 0 The meaning and unit can vary from scheme to scheme The default value is 0 0 in each case 5 3 10 Radiation Transport Control In this part of the parameter file the radiation transport module has to be selected Depending on this selection a couple of additional parameters have to be specified Table 13 giv
86. erly For instance if you want to enable debugging options type export F90_DEBUG 1 Restart the configure script after every change in the control variables With e g export F90 MACHINE dummy export F90 PREFLAGS Oprettyfast Qsomethingelse configure it is possible to specify all machine dependent settings yourself see Sect 3 6 and Sect 3 6 This is useful when dealing with a compiler hitherto unknown to the configure script 9 Start the compilation with make to produce the executable rhd exe 16 3 PROGRAM FILES INSTALLATION COMPILATION A simple sample installation may look like the following the sub directory for is put into the home directory Choose base directory cd HOME 4 Put the tar file there 4 Expand the tar file tar zxvf for tar gz Go into default master directory cd for hd rhd YOUR MACHINE Activate OpenMP und MSrad radiation transport export F90_PARALLEL openmp export F90_MSRAD 1 Start the configure script configure 4 Compile make echo Voila If you want to compile in a directory in a completely different place not in a sub directory of for as described above you have to set the environment variable F90 BASEPATH see Sect 3 6 to make the paths to the source files known to the configure script That might look like mkdir SOME WEIRD PLACE cd SOME WEIRD PLACE export F90_BASEPATH HOME
87. es a list of the parameters and the modules they apply to The standard routines are now in the MSrad module for local models and the SHORTrad module for global Star in a Box models The LHDrad module is not maintained very much anymore e character radscheme So far there exist three different radiation transport modules The active on can be selected e g with character radscheme f A80 b 80 n Radiation transport scheme amp c0 LHDrad MSrad SHORTrad amp ci None skip radiation transport step entirely SHORTrad Possible values are o None Skip radiation transport entirely o LHDrad old supergiant module It uses long characteristics and is restricted to an equidistant grid and open boundaries at all surfaces Note that the switch Drhd r01 1 has to be set during compilation see Sect 3 7 82 5 CONTROL AND DATA FILES Parameter Section LHDrad MSrad SHORTrad radscheme il n radminiter 5 3 10 s si d n raditer 5 3 10 i ii n radmaxiter 5 3 10 m T x radraybase 5 3 10 T radraystar 5 3 10 n n radtheta 5 9 10 n radphi 5 9 10 n radsubray 5 9 10 n radthinpoint n radthickpoint n radtaurefine 5 3 10 Li n radband 5 3 10 c radimplicitmu 5 3 10 ii si c_raditereps 5 3 10 x i c_raditerstep 5 3 10 i i c_radtvisdtau 5 3 10 j c_radtvis li c_radhtautop c_radcourant 5 3 12 ii ii c radcourantmax 5 3 12 ii i u c radmaxeichange 5 3 12 i T Table 13 List
88. es that control the CO5BOLD configure script are also rec ognized see the header of the UIO configure script The command make install generates a directory HOME local and sub directories An init script is put into HOME bin And the resource files cshrc and bashrc are modified to call it Therefore this installation step is potentially dangerous because its effect is not restricted to the local directory 4 6 2 Quick Examination of Files uiolook The shell script uiolook calls the Fortran program uiolok f90 The man page UIOLOOK 1V Misc Reference Manual Pages UIOLOOK 1V NAME uiolook print entry headers of file in uio form SYNOPSIS uiolook h p filename AVAILABILITY DESCRIPTION The routine uiolook reads each filename file in uio form in sequence and displays the headers of the entries in pretty form OPTIONS h Print usage of uiolook p Entry header keywords in long pretty form SunOS 5 5 1 Last change 27 November 1996 1 4 6 3 Transformation of Files uiocat The shell script uiocat calls the Fortran program uiocat f90 The man page UIOCAT 1V Misc Reference Manual Pages UIOCAT 1V NAME uiocat concatenate file s in uio form SYNOPSIS uiocat c conversion f format h 1 1 copylist o outputfilename filename 58 4 UIO DATA FORMAT DESCRIPTION The routine uiocat reads each filename file in uio form in Sequence and displays its contents formatt
89. essing data in UIO format within IDL the initialization routine uio init see Sect 4 7 1 and the high level reading routines uio struct rd pro uio dataset rd pro and uio datasetlist rd pro see Sect might suffice 4 7 1 Initialization of UIO Routines under IDL The directory containing the IDL UIO routines should be added to the IDL variable PATH This could be done by a program segment in the startup procedure like Try to determine language if n elements X TICKV eq 150 then langua WAVE else langua IDL Add user IDL directory to search path if langua eq IDL then begin amp addpath expand path C UIOPATH idl amp endif else begin amp addpath home supas024 uio idl home supas024 wave amp endelse if strtrim addpath 2 ne then path addpatht path delvar addpath Alternatively one might want to set the IDL path variable accordingly like export IDL_PATH UIOPATH id1 for example in the bashrc file Or one just copies or links the UIO IDL routines to a location in the standard IDL search path It is reasonable to include the UIO initialization in the startup procedure as e g Initialize wio routines uio_init progrm by hand IDL can handle the conversion types native ieee_4 ieeele_4 ieee idl xdr compare Tab 6 in Sect 1 3 1 Here ieee 4 is the default and should be used as a standard Attention The IDL type l
90. ested the system returns an error and terminates the program Without this option the code aborts with an error message indicating illegal nesting Compiler bug pmpar Collects the performance monitor information for each parallelization unit pmfunc Collects the performance monitor information for each procedure Drhd hyd roeid 101 1 Optimization Choose non standard set of routines for Roe solver See Sect DMSrad raytas 0 Optimization choose default version of loop in SUBROUTINE raytas in file M9rad3D F90 See Sect Important note The UIO routines need in addition the compiler option subchk Array bound checking Without this checking option some UIO routines are not working properly compiler bug 38 3 PROGRAM FILES INSTALLATION COMPILATION A proposed compiling seguence is only default modules activated export F90_PREFLAGS subchk configure make UIO export F90_PREFLAGS configure make Performance tests on hwwsr8k Some tests have been performed on the machine hwwsr8k at HLR Stuttgart in order to de termine the optimum chunk sizes which are set by the parameters n_hydcellsperchunk and n_viscellsperchunk see Sect 5 3 7 and Sect 5 3 8 T wo different models have been used one consisting of 128x128x192 grid cells the other of 252x252x188 respectively Grey radiative trans fer has been performed with the MSrad module Different values for the chunk size s have been assumed where the hydrody
91. f cell boundaries cm rho Density g cm 3 ei Internal energy erg g vi Velocity 1 cm s v2 Velocity 2 cm s v3 Velocity 3 cm s quc001 Number density of CO 1 cm 3 advect 1 bb1 Magnetic field 1 G bb2 Magnetic field 2 G bb3 Magnetic field 3 G It might follow ACTION Initialize MS radiation transport routines lt lt lt lt lt lt lt lt And finally ACTION Open output files lt lt lt lt lt lt lt lt which indicates that the rhd full file see Sect 5 1 and the rhd mean file see Sect 5 2 have been opened and now contain a header The end of the initialization phase and the beginning of the proper simulation is marked by e g 5 2 S5 S S Start Computation Eee Time step number itime 47050 time 2 5821813E 08 t job 8 510000E 00 The output for a typical simulation time step can look like for a supergiant model with SHORTrad radiation transport Time step number itime 49048 time 2 6822680E 08 t job 1 272180E 06 dtime 5 3047E 03 HD 1 4838E 04 RAD 5 3047E 03 VIS 1 0723E 05 Luminosity per core volume 4 49999049E 02 HYD 1 N_cellsperchunk n chunks 10000 1410 HYD 2 N_cellsperchunk n chunks 10000 1410 HYD 3 N_cellsperchunk n chunks 10000 1410 VIS3D N_cellsperchunk n chunks 10000 1360 Start of rhd shortrad step n subdtime 1 minmax T 1 111651E 03 1 630733E 05 Main 1 3 ray direction 2 0 000000 1 000
92. f threads should equal the number of available processors and has to be set at run time with the environment variable OMP NUM THREADS e g with export OMP NUM THREADS 16 3 8 2 General Inlining Candidate routines for inlining are i e they should be inlined if anyhow possible e file rhd hyd module F90 rhd_hyd_avg rhd hyd upwind rhd hyd predO rhd hyd predm rhd_hyd_predp rhd hyd alpha rhd hyd constanteq rhd_hyd_minmodeg rhd_hyd_minmod rhd hyd vanleereq rhd hyd vanleer rhd_hyd_superbeeeg rhd hyd superbee rhd hyd ppeq rhd hyd pp rhd hyd hdflux e file rhd 1hdrad module F90 rhd rad3d raylhd rhd rad3d solve rhd_rad3d_solveeq 3 8 Optimization Compiler Switches 35 e file rhd shortrad module F90 rhd shortrad operator rhd shortrad dtauop On some machines the makefile generated by the configure script contains this list explicitely On others one has to rely on automatic inlining see the following sections All routines that should be inlined are contained in the same modules as the calling routines Therefore no inter procedure inlining is needed 3 8 3 Cray SV1 On craSHi in Kiel a CRAY SV1 20 32768 SN9542 now out of service CO5BOLD could use all 4 processors per board Documentation about the system and the compiler can be found with the CRAYdodi system The new configure script still includes a branch for this system even if has never been tested on that machine In some cases the non d
93. fer 4 freguency bands and a large one with 315x315x81 grid points employing grey radiative transfer Rather short runs of 10 small model and 3 large model time steps were performed Even for the large model the memory demand was ca 800Mb which is very modest considering that every sub node of the machine consisting of 4 processors has 2Gb of memory The results are summarized in the following three figures The black lines give the scaling of the total time the green lines the scaling of the time needed by the hydrodynamics routines and the red lines the scaling of the radiative transfer routines The scaling is presented as the increase of processing time per processors as the problem is distributed among more and more processors The times are normalized to the time that is used in a scalar i e single processor setup Ideally one would like a constant behavior which stays close to one 2 0 Machine grand 1 8 125x125x81 grid points 4 OBM band 10 time steps 1 6 1 4 0 5 10 15 20 25 Number of processors Per processor execution time increase Figure 3 UKAFF machine grand small model The perhaps most interesting result is that the speedup on ukaff is about 11 for the large model on 16 processors Perhaps not ideal but within the range of practical interest In general the hydrodynamics routines scale more favorably than the radiation routines This is perhaps sim ply related to the fact that in expli
94. file MSrad3D F90 See Sect e Drhd hyd roeid 101 0 Optimization choose standard set of routines for Roe solver See Sect ipi Peace 3 8 Optimization Compiler Switches 41 e Drhd shortrad operator 102 1 Optimization use the manually inlined version of the short characteristics operator e Drhd shortrad dtauop 102 1 Optimization use the manually inlined version of the optical coupling operator Everything compiles The MSrad radiation transport module does not work with OpenMP activated In this case it is recommended to use the Intel compiler instead see Sect 3 8 10 3 8 10 Linux Intel Compiler With Version 7 0 and 7 1 of the Intel compiler CO5BOLD compiles Version 8 0 still causes touble The native format on Intel machines is little endian With export F_UFMTENDIAN big to be set at runtime after compilation before running CO5BOLD the default can be changed to big endian The appropriate UIO modules are uio mac intel module f90 in the little endian case and uio mac sun module f90 in the big endian case The compiler is called with ifc Important switches are Vaxlib Link proper library to make the machine understand e g call flush 6 fpp Activate the preprocessor silently e 03 General optimization flag e tpp6 xK Optimization especially for Pentium III and Athlon includes SSE vector com mands e tpp7 xW Optimization especially for Pentium IV includes SSE2 vector command
95. for In s F90 BASEPATH hd rhd conf configure configure The variable F90 BASEPATH also has to be set explicitely if the main directory for should have another name Renaming the sub directories with the source files is not a good idea it requires modifications of the configure script itself 3 3 Directory Structure The files necessary to compile CO5BOLD are distributed over a few directories A typical setup would be to put everything into the main directory for Then the source files would be located as in Tab The executables and makefiles object files module information files are usually located in subdirectories of the source code directories These subdirectories typically have the name of the machine architecture or operating system the executable is compiled for The former directory organization can be found in Tab 2 and Fig 3 4 Old Setup File for Paths For the previous version of CO5BOLD all paths were stored in environment variables and could be set with the scripts 3 4 Old Setup File for Paths setarcdeppaths csh setarcde ppaths sh TMAC MAC RHDEXEPATH HSTEXEPATH rhd RHDPATH MAC MAC 3HDWEXEPATH MEANEXEPATH hdw mean 3HDWPATH IMEANPATH mat Main directory MATPATH FORTRANDISK str f90 jSTRPATH MATSRCPATH uio eos UIOPATH EOSPATH FMAC MAC jSTREXEPATH jSTREXEPATH bin f90 gas f90 UIOSRCPATH 3GASPATH SECSSRCPATH SMAC mani UIOEXEPATH MAC MAC GASEKEPA
96. formation Obsolete and Test Parameters e real abux This optional information parameter can be specified with real abux f E15 8 b 4 n hydrogen abundance number fraction u 1 amp cO standard solar mixture 0 90851003E 00 It has no practical consequences because the actually used chemical composition is deter mined by the files for equation of state and opacity e real abuy This optional information parameter can be specified with real abuy f E15 8 b 4 n helium abundance number fraction u 1 amp cO standard solar mixture 0 90850003E 01 It has no practical consequences because the actually used chemical composition is deter mined by the files for equation of state and opacity e real qmol This optional information parameter can be specified with real qmol f E15 8 b 4 n mean molecular weight u u amp c0 standard solar mixture 0 13018000E 01 It has no practical consequences because the actually used chemical composition is deter mined by the files for equation of state and opacity e real gamma This optional information parameter can be specified with real gamma f E15 8 b 4 n Adiabatic coefficient u 1 amp c0 0 0 1 6666666666666 1 4 0 0 It has no practical consequences because the actually used chemical composition is deter mined by the files for equation of state and opacity e real c visneul real c visneul f E15 8 b 4 amp n Linear viscosity parameter von Neumann Rich
97. g cm amp ds 0 0 0 0 0 1 frhov13b xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Momentum x flux x3 direction amp u erg cm 3 amp ds 0 0 0 0 0 1 frhov23b_xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Momentum x2 flux x3 direction amp u erg cm 3 amp ds 0 0 0 0 0 1 68 5 CONTROL AND DATA FILES real frhov33b xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Momentum x3 flux x3 direction amp u erg cm 3 amp ds 0 0 0 0 0 1 real feipb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Enthalpy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 real fekb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Kinetic Energy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 real fegb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Gravitational Energy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 real fepb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Pressure Energy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 real fevb_xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Viscous Energy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 real ferb xmean d 1 1 1 1 1 121 f E13 6 p 4 b 4 amp n Radiative Energy Flux amp u erg cm 2 s amp ds 0 0 0 0 0 1 label endbox The above list was slightly edited by adding blanks to improve readability The identifier of an entry together with the name n and the unit u should give a first hint about the meaning
98. g measurement are called very frequently and might slow down the code It should not be used in conjunction with OpenMP Values o undefined default no timing information o defined call subroutines to measure elapsed time rhd rad3d step t01 inrhd Ihdrad module F90 rhd radiation 3 dimensions step timing 01 Category additional output Produce timing information with main 3D radiation transport routine It can be used together with OpenMP and should cause no noticeable performance loss Values o undefined default no timing information o defined call subroutines to measure elapsed time rhd shortrad operator 101 inrhd shortrad module F90 rhd short characteristics radiation operator loop 01 Category performance enhancement selection of approximation Choose type of short characteristics operator The operators usually come in pairs 1 2 3 4 5 6 There is a development from 1 2 over 3 4 to 5 6 towards higher stability Both members of each pair should do the same operation but use different ways to do a case distinction The even operator has in most case the better performance But the odd operator might be saver to use Values o 0 simple test operator fast but results are utterly useless o 1 case distinction with if then else construct o 2 case distinction with masks weights 0 0 or 1 0 o 3 case distinction with if then else construct slope reduction of source function o 4
99. gnored by COS5BOLD but read by the IDL input routine They can be used to provide comments additional information about the model or control parameters for further processing 5 3 Parameter File rhd par 69 5 3 1 Quickstart How to Make a Proper Parameter File You will never write a new parameter file from scratch Typically you take an old file e g the one controlling the simulation which produced the model which is used to start the new run and edit it 1 Take the parameter file corresponding to the model you want the new simulation to start with 2 Most of the parameters should be already OK E g most of the parameters controlling the boundaries do not have to be changed 3 Write a brief description of the purpose of the planned simulation into the character description array You might but a remark about the parent file of the pa rameter file under construction and the current date into the character history array see Sect 5 3 2 4 Check modify the name of start model and output files infile start outfile end outfile full outfile mean On a system with batch queue this has not to be done in the parameter file itself but in the external command file see Sect 5 3 13 5 Check modify the fundamental parameters including boundary condition specifiers see Sections and 5 3 4 respectively e effective temperature control s inflow teff luminositypervolume C radHtautop e gravity grav mode grav mass
100. he native format is equal to the conversion type crayxmp_8 but also the conversion types ieee_4 ieee 4 limit and jeee 8 can be chosen The last three conversion types correspond to the CRAY internal types jeee 32 ieee_dp and ieee 64 respectively 4 3 Structure of UIO Files 49 On a machine with an internal data representation not within the list in the existing uio mac module f90 files one could use the standard file uio mac module f90 and is restricted to the native conversion type But it is better to invent an appropriate name for the new data format and to build a proper machine dependent UIO file e g from uio mac ieee module f90 conversion type I R D description native 7 7 17 internal data format on all machines sometimes useful but not recommended ieee 4 4 4 8 standard IEEE big endian format recommended ieeele 4 4 4 8 IEEE little endian format ieee 8 8 8 16 double precision IEEE big endian format on some machines possible crayxmp 8 8 8 16 CRAY internal data format idl 4 4 8 IDL format but IDL also supports ieee 4 xdr 4 4 8 format possible with IDL ieee 4 limit 4 8 8 standard IEEE format ieee 4 ieee To Y Y IEEE format unknown length not recommended Table 6 UIO conversion types with length of integers single precision reals and double precision reals in bytes and an explanation Some attention has to be paid if weird compiler switches as
101. he routine uio datasetlist rd pro is con venient as in the short example 62 4 UIO DATA FORMAT ierr 0 delvar listdata Loop over all datasets while ierr eq 0 do begin amp Read the next dataset ful uio datasetlist rd testmodel 0 full listdata listdata ierr ierr amp if ierr eq 0 then begin amp print ful z time format A E15 8 Now do the data handling demo print Mean density avg ful z rho amp endif amp endwhile or in the more detailed example model st33gm06n03 modelident amp parmodelident 01 modeldisk getenv HOME dat rhd d model modelfile modeldisk model modelident ae sofa L parfile modeldisk model parmodelident par Read parameter file par uio_struct_rd parfile Open first dataset to get some information about array sizes delvar listdata ful uio datasetlist rd modelfile listdata listdata ierr ierr uio closrd listdata channel delvar listdata nxci n elements ful z xc1 nxc2 n elements ful z xc2 nxc3 n elements ful z xc3 n timestep 1000 Some huge value to get everything Reduce for tests ierr 0 i 0 Loop over all datasets while ierr eq 0 and i 1t n_timestep do begin amp Read the next dataset ful uio datasetlist rd modelfile listdata listdata ierr ierr amp if ierr eq 0 then begin amp print i
102. he work load distribu tion onto the threads in case of parallel runs with OpenMP Due to the special han dling of boundary cells the overhead per call increases significantly for small chunks Typically larger chunk sizes compared the the hydrodynamics chunk sizes set with integer n_hydcellsperchunk see Sect are adequate On the other hand they should not be too large to limit the usage of temporary memory and to allow parallelization the distribution of chunks to threads For simulations with activated OpenMP on a par allel machine the chunk size has to be made small enough to allow at least as many chunks as processors available This is particularly important for models with a small number of grid points e g 2D models An example is given for the Hitachi SR8000 in Sect 5 3 9 Dust Molecules CO5BOLD can now handle a number of additional density arrays They can be used to describe e g the mass density of dust distribution moments or number densities of molecules These species are properly advected with the gas density There is also already a small number of dust molecule formation models available These models have to be improved in the future and the influence on the radiation field opacities radiation pressure on dust has to be taken into account 5 3 Parameter File rhd par 81 e character dustscheme A scheme for dust or molecule formation and transport can be selected e g with character dustscheme f A80 b 80 n Dust mod
103. ho min f E15 8 b 4 n Minimum boundary density u g cm 3 1 0E 25 The density within the model will typically not fall much below this value A value of 0 0 default or below deactivates this feature e real c coredrag To damp the flow in the core of models with central potential a drag force restricted to the inner part of the model r rO0 grav can be applied It is controlled e g with real c coredrag f E15 8 b 4 n Core drag force parameter u 1 1 0 A value of 0 0 default or below deactivates this feature 5 3 5 Equation of State e character eosfile The equation of state file together with the opacity file implicitly determine the chemical composition The EOS file can be specified for instance with character eosfile f A80 b 80 n E0S file name amp cO eos gamma140 eos eos mm20 1 eos eos mm00 13 eos eos mmOO 15 eos There exists an increasing number of files o o o eos mmOO 13 eos Standard EOS file for solar composition with extra large density range towards low densities There exist two other files for the same composition but smaller density range eos mm00 eos eos mm00 1 eos eos XXXX 14 eos Several EOS tables for various compositions eos mmOO 15 eos EOS file for solar composition with temperature range extended to very low values by keeping u fixed below 500 K The number of points for the density sampling has been reduced The number of points for the energy sampling has been increased
104. iables uiocop f90 Program to copy and transform UIO files uiolok f90 Program to look into UIO files uioinf f90 Program to give information about conversion types uiotst f90 Program to produce test UIO file Table 9 UIO Fortran90 files module contents uio cst module channel status information uio cvl module convert type list of current machine uio inf module information about environment uio nam module definition of names uio siz module string length table size uio base module basic set of UIO routines string processing header handling I O channel management Table 10 Contents of uio base module f90 4 5 Fortran90 4 5 1 Files The Fortran UIO package is a collection of Fortran90 modules and programs described in Table 9 The file uio base module f90 contains the basic set of modules see Table 10 The files uio mac module f90 Tab 9 contain machine dependent routines collected in the module uio mac module see Tab tip It comes in various flavors The machine independent version is uio mac module f90 which can be used for first tests but does not provide all possible features Therefore it should be discarded afterwards and replaced by a version more suitable for the platform in use The file uio_mac_ieee_module f90 is appropriate for all machines with IEEE big_endian binary format Additionally there exist files containing calls of machine library routines uio_mac_crayts_module f90 uio mac crayx
105. ierror ierror This operation adds the tags EOS and OPA to the data structure FUL which contain more quantities like e g the temperature see Sect 7 3 108 7 DATA ANALYSIS WITH IDL 7 3 IDL Data Structure The data structure FUL contains the following variables and substructures Use help str ful to get this information See also the short description of the contents of a model file in Sect and particularly the man page of the script uiolook in Sect which gives you even more detailed information directly from the file xx Structure lt 8287a0c gt 9 tags length 78404128 refs 1 TYPE STRING uio HEAD STRUCT gt Anonymous Array 1 DATASET ID STRING single box MODELTIME FLOAT 10050 1 MODELITIME LONG 64088 DTIME FLOAT 0 176381 TIME OUT FULL LAST FLOAT 10050 1 TIME OUT MEAN LAST FLOAT 10040 0 Z STRUCT gt Anonymous Array 1 If the command in Sect has been performed the following substructures are present EOS STRUCT gt Anonymous Array 1 OPA STRUCT gt Anonymous Array 1 The substructure FUL Z contains the original data arrays from the model file like the spatial axes density rho internal energy ei and the three spatial components of the velocity vi v2 v3 Structure 8274184 16 tags length 78403900 refs 2 TYPE STRING uio BOX ID STRING t DIMENSION LONG Array 2 3 TIME FLOAT 10050 1 ITIME LONG 64088 XC1 FLOAT Array 140 1 1 XC2 FLOAT Arrayli 140 1 XC3 FLOAT
106. imulation should not use the same parameter file it would stop immediately real plustime If the initial model should be advanced by a certain stellar time span this value can be set e g with real plustime f E15 8 b 4 n simulation advance time u s 5 0E 07 A value 0 0 cancels this halt condition it is not checked at all This condition assures if it is the only one that all individual simulation runs cover approximately the same stellar time integer endtimestep If the simulation should run up to a certain time step its values can be specified e g with integer endtimestep f I11 b 4 n total simulation time step number u 1 1234 88 5 CONTROL AND DATA FILES This might be useful to advance the simulation up to a point shortly before a previous simulation crashed A value 0 cancels this halt condition it is not checked at all integer plustimestep If the initial model should be advanced by a certain number of time steps their number can be set e g with integer plustimestep f I11 b 4 n simulation advance time step number u 1 2000 A value 0 deactivates this halt condition it is not checked at all 5 3 12 Time Step Control In this group parameters to control the time step restrictions can be set They are important because decide about performance and stability of CO5BOLD They should be tested and adjusted for a simulation of a new type of object But all the dimensionless parameters
107. ined The unformatted binary I O is much faster and gives usually more compact files with higher accuracy ideally exact in the numerical data representation But in principle the file format is machine dependent Fortunately the IEEE format turns out to become a quasi standard among a variety of machines Most workstations work internally with this format Some CRAYS which have a different internal data representation allow the hidden transformation between the internal and IEEE format during the I O process The UIO routines support this feature of CRAY FORTRAN compilers by means of a module uio mac module individually designed for two types of CRAY machines using certain CRAY specific system calls CRAY FFIO assign logic Nevertheless there is also a machine independent version of this module written completely in standard Fortran90 but providing less features than the machine dependent versions Besides the format the conversion type see table 6 has to be specified The native conversion type is the internal binary data representation which is also standard for unformatted Fortran output If this representation happens to be conformal with the IEEE standard the conversion type ieee 4 should be used It gives the same data format but in the header of the file the term convert ieee 4 instead of convert native describes the data format precisely in a way also understandable by other machines On CRAY machines t
108. istant grid and open boundaries at all surfaces old supergiant module Values 30 3 PROGRAM FILES INSTALLATION COMPILATION o undefined default LHDrad routines are deactivated o 1 LHDrad routines are recognized by the compiler rhd r02 in rhd F90 rhd radiation 02 Category feature activation Switch to include MSrad radiation transport module It uses long characteristics The lateral boundaries have to be periodic Top and bottom can be closed or open solar module Values o undefined default MSrad routines are deactivated o 1 MSrad routines are recognized by the compiler rhd_r03 in rhd F90 rhd radiation 03 Category feature activation Switch to include SHOR Trad radiation transport module It uses short characteristics and is restricted to an equidistant grid and open boundaries at all surfaces new supergiant module Values o undefined default SHORTrad routines are deactivated o 1 SHOR Irad routines are recognized by the compiler rhd rad3d toray 101 inrhd Ihdrad module F90 rhd radiation 3 dimensions to ray loop 01 Category performance enhancement There might be a performance gain by splitting the main loop in routine rhd rad3d toray into three separate loops Typically one big loop is to be preferred Values o undefined default One big loop o defined Three smaller loops rhd rad3d fromray 101 inrhd Ihdrad module F90 rhd radia
109. l dtime out mean f E15 8 b 4 n utput time step u s amp c0 dtime out mean 0 0 gt no output 0 5E 06 92 5 CONTROL AND DATA FILES Allowed values are o 0 0 No output to this file o 0 0 Output at every time step o gt 0 0 Output to mean file approximately every dtime out mean seconds Because the size of one mean dataset is much smaller than one full dataset it is possible to request a higher sampling rate without using too much disk space character infile start The filename of the initial model is specified e g with character infile start f A80 b 80 n File name of start model rhd sta Default is rhd sta for a parameter file used within a batch system Typical filenames are st35gm04n01_01 sta or gt57g44n20dz end character outfile end The file name for the final model can be specified with e g character outfile end f A80 b 80 n QOutput file name rhd end The default is rhd end Leaving it empty means that no final model is written This of course inhibits follow up simulations but can be useful to save time and disk space for some tests character outfile full The name of the file for the output of additional full models at regular intervals see Sect 5 1 can be given with e g character outfile full f A80 b 80 n Output file name rhd full Leaving it empty means that no file of this type is written character outfile mean The name of the file for the output of additional
110. l routines hd rhd rhd hyd module F90 RHD hydrodynamics routines hd rhd rhd vis module F90 RHD tensor viscosity routines hd rhd rhd rad module f90 RHD interface for Matthias radiation routine hd rhd rhd lhdrad module F90 RHD radiative transfer routines long characteristics supergiant case hd rhd rhd shortrad module F90 RHD radiative transfer routines short characteristics supergiant case hd rhd rhd F90 RHD main program Table 5 For historical reasons list of all old modules the table shows the file name with part of its path the shortcut for the directory and its description 24 3 PROGRAM FILES INSTALLATION COMPILATION 3 6 Configure Script The configure script produces a Makefile It is controlled by environment variables see below It tries to use reasonable default values if they are not set properly In the script the machine type is determined with uname m According to the control variables and the machine architecture the compiler name and its compiler flags are composed These are written into the header of a Makefile which is produced in the end An existing Makefile is appended to Makefile old Additionally the compilation command is written into the file gt compiler_flags info in a form ready to be included in a Fortran program The environment variables that control the script are e F90 COMPILER Fortran compiler o a machine dependent default is chosen individually for each architectur
111. lly smaller than 2 0 1 0 Its value should be slightly above c viscourant and below 2 0 5 3 13 Input Output Control With this group of parameters the start model and the type and amount of output can be specified Parameters with the suffix start describe the initial model these with suffix end the corresponding final model Additional data can be written into the file described by the parameters with suffix full full 2D 3D model dumps huge see Sect or into the file described by the parameters with suffix mean additional information see Sect 5 2 e real dtime out full The interval between datasets in the full file can be set e g with real dtime out full f E15 8 b 4 n Output time step u s amp c0 dtime out full lt 0 0 gt no output amp c1 dtime out full 0 0 gt output every time step 2 0E 06 Allowed values are o lt 0 0 No output to this file o 0 0 Output at every time step Attention This can produce HUGE files in no time o gt 0 0 Output to full file approximately every dtime out full seconds Some examples The classical value for this output for simulations of solar granulation is 20sec To save memory this can be increased to 30sec But in this case chromospheric shocks are very badly resolved To cover them properly a sampling rate of 10sec or below is needed e real dtime out mean The interval between datasets in the mean file can be set e g with rea
112. m start model o 0 default Take the initial time step count from start model real cputime Because of the long simulation time usually CO5BOLD will run in some sort of batch mode which might impose limits on the execution time per run On a CRAY the CPU time that is left can be accessed with a special subroutine in call tremain in rhd_mac_cray_module f90 On other machines it is possible to specify the allowed total time for the job e g with real cputime f E15 8 b 4 n CPU time u s 1000000 0 During the run of CO5SBOLD the leftover CPU time is computed by subtracting the used CPU time which is given by e time etime tarray in rhd mac sun module f90 from the specified total CPU time for the job real cputime_remainlimit Because CO5BOLD needs some time to finish the last time step it should start exiting well before all CPU time is used up This amount of buffer CPU time can be specified e g with real cputime_remainlimit f E15 8 b 4 n maximum remaining CPU time u s 2000 0 Its value depends on the size of the model and the speed of the machine more precisely the maximum CPU time per time step real endtime If the simulation should run up to a certain stellar time its values can be specified e g with real endtime f E15 8 b 4 n total simulation time limit u s 10000 0 A value 0 0 deactivates this halt condition it is not checked at all If this parameter is set to a non negative value a follow up s
113. mation The number of points in diffusion approximation can be set with e g integer n radthickpoint f I4 b 4 amp n Number of grid points with optically thick diff approximation amp c0 0 no diffusion approximation 0 The value should be choosen so that for all points in that region Ar gt 1 is valid Setting this value to 0 means that the diffusion approximation is not used in any part of the model integer n radtaurefine With the LHDrad and the MSrad module the number of points on the rays can be finer than the number of points in the basic numerical grid T he refinement can be set e g with integer n radtaurefine f I4 b 4 amp n Refinement factor 0 5 3 Parameter File rhd par 85 e integer n radband It can be specified whether the grey opacity table or the binned frequency dependent part of the opacity table is used during the computation The grey part contains only one bin The other possibly non grey contains one or more bins depending on the table chosen The parameter is specified with e g integer n radband f I4 b 4 n Number of frequency bins amp c0 1 grey opacities amp ci 2 non grey opacities if available from table 1 Allowed values are o 1 Use the grey part of the table o 2 Use the other possibly non grey frequency dependent part of the table Only the MSrad module so far can handle non grey tables e real c_radimplicitmu So far only the LHDrad and the SHORTrad modul
114. modules and the string handling module should be compiled in debug mode A proposed compiling sequence is MSrad does not compile all other modules are activated export F90 LHDRAD 1 export F90_MSRAD 0 export F90_SHORTRAD 1 export F90 DUST 1 export F90 MHD 1 export F90 PARALLEL openmp export F90_DEBUG 1 configure make UIO STR export F90_DEBUG 0 configure make 3 8 6 Hewlett Packard Itanium 2 The 2 processor system gunnar from Hewlett Packard is a dual Itanium 2 machine with two 900MHz ia64 CPU modules 4GB of RAM and 70GB user diskspace The single processor performance of COSBOLD is very good On two processor the code runs even faster but just stops after a few time steps The number of time steps varies even for simulations with the very same start model and parameter file Therefore OpenMP should not be activated currently The compiler settings are somewhat similar to the settings of the HP V2500 system in Sec tion 3 8 5 e U77 Link proper library to make the machine understand e g call flush 6 http devresource hp com devresource Tools lang html 3 8 Optimization Compiler Switches 37 cpp yes Switch on the C preprocessor Note that all Fortran90 files have to end with 90 The F90 suffix does not seem to work Ofast High optimization level And tOfaster is even higher Dopenmp Onoautopar Try to enable parallelization with OpenMP directives disable auto parallelization Th
115. mp module f90 uio_mac_sun_module f90 These make it possible to write information about the platform in use into the file header The CRAY versions allow unformatted I O in the CRAY specific format and additionally via the FFIO ASSIGN logic in IEEE format The file uio mac intel module f90 is appropriate for all machines with IEEE little endian binary format and no mechanism for automatic conversion 54 4 UIO DATA FORMAT routine purpose uio_getenv Get information about environment uio mkcvls Make list with possible conversion types uio uopen Open file with special handling for conversion type uio uclose Close file with special handling for conversion type Table 11 Contents of uio mac module The main set of routines is contained in uio bulk module f90 in the module uio bulk module The three files uio base module f90 uio mac module f90 and uio bulk module f90 comprise the standard set of UIO routines Additionally there exists a file uio table module f90 with the single module uio table module which permits the I O and manipulation of a certain table format see the example in section 4 2 The latest extension comes within the modules uio var module f90 and uio varfile module f90 The module uio var module f90 contains a type definition for a variable uio flexible variable of general type i e it may be a scalar integer value or a 1D character array or a 3D real array together with some routines for th
116. mplies 2 2 A collection of thermodynamic relations M Steffen AIP x Oe dp 5 d E p y y 1 e 1 e p ds dp 4 d 75 i o and the requirement that the mixed derivatives must be equal then yields CE ae 55 85 Se Cae 2 Similarly replacing de by em a Il in EqB3 we get or or 5 Ser e 2 1 E Op Alnp Olnp p Op p p Since 5 Ss u 5 2 u Olnp NOlnp PN Olnp mp id we finally obtain using Eqs 13 49 and The isothermal sound speed is then obtained as and 2 2 5 Ideal gas with constant specific heats polytropic gas In this case we obtain much simpler relations 1 Bee eb PT p I lese T pyel s Cy Inp ylnp const y 1 l2 3 const Cu p C Cy ER p Y pT 1 Cy pop cp R d 11 67 68 69 2 EQUATIONS 77 1 XT Xp 78 p p 05 79 P p 2 p r 13 3 Program Files Installation Compilation In this section all the files and modules CO5BOLD contains are listed The installation procedure is outlined and compiler switches necessary to compile CO5BOLD and to optimize its performance are described 31 Quickstart How to Compile CO5BOLD If you are going to install CO5BOLD on a machine with a known to setup script and makefile operating system and compiler see
117. mum cache usage and need the smallest amount of memory but result in additional overhead due to frequent subroutine calls Bigger and less chunks are to be preferred for vector machines and processors with large caches Very rough guide values may be o 2500 Pentium III processor o 20000 RISC processor o 100000 Vector machine Note For simulations with activated OpenMP on a parallel machine the chunk size has to be made small enough to allow at least as many chunks as processors available This is particularly important for models with a small number of grid points e g 2D models An example is given for the Hitachi SR8000 in Sect real c visdrag This viscosity parameter controls the drag force which is if requested applied inside the hydrodynamics routines themselves It does not act on velocity gradients as usual viscosity but applies a force proportional to the velocity itself but with the opposite sign The amount can be specified e g with real c visdrag f E15 8 b 4 amp n Drag viscosity parameter u 1 0 001 The value gives the fraction the velocity is reduced per time step Therefore reasonable values lie between 0 0 and 1 0 In almost every case the drag forces will be switched off c_visdrag 0 0 If e g strong pulsation have to be damped in the initial phase of a simulation a value around 0 001 0 01 seems appropriate real c visbound An additional drag force can be added locally in inflow cells in the outer layer
118. namics and the viscosity parameter were set equal In all cases three time steps have been computed The results are shown in Fig 2 The number of resulting chunks for step HYD1 the values for HYD2 HYD3 and VIS are very similar total memory perfor mance and the wall clock duration of the hydrodynamics and the viscosity routines are shown as functions of the chunk size parameter s Clearly the number of chunks decreases towards larger chunk sizes whereas the required memory increases in particular for very large chunk size values Moreover performance and CPU time can be optimised by choosing the right parameter values Interestingly the optimum chunk size is different for hydrodynamics and viscosity Based on these tests a larger value seems to be preferable for the viscosity n viscellsperchunk In the case of the smaller model 50000 seems to be fine for the hydrodynamics whereas the optimum viscosity chunk size is 200000 This difference explains the double peaked structure of performance and CPU time Note that the optimum values do not only depend on the architecture used but also on the dimensions of the model We recommend to test some chunk size values since it might lead to a higher performance 3 8 8 IBM Useful links IBM compiler documentation IBM XL Fortran especially IBM AIX compiler information center P Another source of the compiler documentation Documentation for AIX Systems especially IBM XL Fortran
119. nd modify table contents rearrange lines Merge two tables in different ways Read 1d array from 2d table array all types Write 1d array into table all types Close file with special handling for conversion type Open file with special handling for conversion type Transform a string to give a correct name of a variable Produce from write format string corresponding read format string Writing scalar and array data of all types Write file header Write header of variable input term table Write header of variable input line table Write label Write table of integer real and or character data to file 3 2k k k k ak k k ak k sk kok dd dd ed okokok ok okokok sk kok skok kok ed dd ak kakak dok skok ak ak akok kak ak ak ak ok ak ak ok ak ak ak ok ak ak ak Include Files uio pro 60 4 UIO DATA FORMAT A filedefinc uiocstinc channel status information common uio chainf uiocvlinc convert type list of current machine common uio_cvlist uionaminc names of types keywords identifiers default formats common uio defnam A uiosizinc length of strings size of tables trmtab lintab uiotabinc empty table structure common uio taborg 5 okokeokokokekekeokekekkkkekkkekelelelekeekeejelekekekekekelelelelelelelelejejejeleekekekekekelekelelelelelelelejeleeekekekelelelelelelelelejelejek Most of the routines are low level ones and do not have to be worried about because they rarely will be used directly For acc
120. nux then MAC linux elif uname n vxi then MAC vx1 else MAC sgi fi Individual libraries Timing TIMEPATH FORTRANDISK time f90 export TIMEPATH TIMESRCPATH TIMEPATH export TIMESRCPATH Constants amp units CONPATH FORTRANDISK con f90 export CONPATH CONSRCPATH CONPATH export CONSRCPATH uio UIOPATH FORTRANDISK uio export UIOPATH UIOSRCPATH UIOPATH f90 export UIOSRCPATH String handling STRPATH FORTRANDISK mat str export STRPATH STRSRCPATH STRPATH export STRSRCPATH Math MATPATH FORTRANDISK mat export MATPATH MATSRCPATH MATPATH f90 export MATSRCPATH gas GASPATH FORTRANDISK eos gas export GASPATH GASSRCPATH GASPATH export GASSRCPATH gt BOS EOSPATH FORTRANDISK eos export EOSPATH EOSSRCPATH EOSPATH f90 export EOSSRCPATH 19 20 3 PROGRAM FILES INSTALLATION COMPILATION Opacity OPTAPATH FORTRANDISK opa opta export OPTAPATH OPTASRCPATH OPTAPATH export OPTASRCPATH hydrostatic HSTPATH FORTRANDISK hd qf15 export HSTPATH HSTSRCPATH HSTPATH export HSTSRCPATH t fad RADPATH FORTRANDISK rad hdrad export RADPATH RADSRCPATH RADPATH export RADSRCPATH RHD RHDPATH FORTRANDISK hd rhd export RHDPATH RHDSRCPATH RHDPATH export RHDSRCPATH gt RHDB RHDBPATH FORTRANDISK hd rhdb
121. nvection a transmitting top boundary will be selected the closed one is an alternative The periodic condition is only recognized by the hydrodynamics routines and not by any radiation transport routine e character bottom bound The boundary condition at the bottom of the model is given for instance by character bottom bound f A80 b 80 n bottom boundary conditions amp cO closedbottom transmitting Possible values are 5 3 Parameter File rhd par 73 o reflective closed wall no gravity no radiation o constant open boundary with constant extrapolation of all values no gravity no radiation o closed closedtop closed wall can handle gravity open for outward radiation o closedbottom closed wall handles gravity radiation in diffusion approximation o periodic periodic boundaries for hydrodynamics and radiation o transmitting transmitting boundary for hydro and outward radiation The parame ters real c tchange real c tsurf and real c hptopfactor have to be specified o inoutflow classical open lower boundary for deep convection gravity and radiation possible The parameters real s inflow real c schange and real c pchange have to be specified In simulations of a solar like star with the MSrad radiation transport module the bot tom boundary is typically of type inoutflow A supergiant simulation will have a transmitting lower boundary e real luminositypervolume The luminosity of a Star in a
122. of macros only activate timing measurements and result in additional output Some of them are not thread save und should only be activated for runs on one thread as done by the configure script It is always save to switch any of them off by removing or undefining them The macros in the category test mark parts of code under development The default values should only be changed with great care typically by the author of that code segment The configure script does not touch these settings General e timing c factor in timing module F90 timing count factor Category account for property of machine To produce the timing statistics printed at the end of a simulation run the standard Fortran routine SYSTEM CLOCK is used The macro timing c factor specifies by how much the count rate of this routine is reduced when storing its count value This does not prevent all overflows but can make the output much more useful Values o 1 default count rate of SYSTEM CLOCK is used directly o otherwise e g 1000 count rate of SYSTEM CLOCK is reduced by this factor By a proper choice of this factor the timing measurements of individual routines can be made meaningful the reduction of the count rate prevents overflows due to the addition of several measurements An overflow during an individual measurement can not be prevented Therefore the count rate for the entire program still tends to produce overflows e gasinter 101 in gasinter
123. of radiation transport control parameters and the modules they are relevant for o MSrad solar module It uses long characteristics The lateral boundaries have to be periodic Top and bottom can be closed or open Note that the switch Drhd r02 1 has to be set during compilation see Sect 3 7 o SHORTrad new supergiant module It uses short characteristics and is restricted to an equidistant grid and open boundaries at all surfaces Note that the switch Drhd r03 1 has to be set during compilation see Sect 3 7 e integer n radminiter Usually the stability considerations dictate a radiative time step smaller than the hydrody namics or tensor viscosity time step To remedy this situation it is possible to allow several radiation transport steps per global time step Hitherto all three radiation transport mod ules support this iteration The minimum number of iterations radiative sub steps can be specified e g with integer n radminiter f I4 b 4 amp n Minimum number of radiation transport iterations c0 8 1 If less iterations are needed the time step limit for the next step is increased This value will in almost any case for explicit radiation transport be set to 1 In the implicit case it is set to a higher value typically 5 integer n raditer After each complete radiative time step the recommendation for the next time step will be chosen so that n raditer iterations will probably needed The parameter
124. ong corresponds to the standard Fortran type integer The IDL types byte and integer are not known in standard Fortran and are therefore transformed to the IDL type long before writing in the IDL routine uio wr Be aware of The parsing and interpretation of the entry headers can only be done by scalar operations which are comparatively slow in IDL 4 7 2 Reading Data with uio data pro The IDL routine uio data and the IDL function uio d were the first set of high level routines to read UIO data in IDL They were useful for the easy reading of not too complex data files By now they are replaced by the routines uio struct rd and uio dataset rd see see next Section and Sect 7 The old routines allow the opening 4 7 IDL UIO Routines 61 uio data mode open filename model dat uio data mode open filename model txt uio data mode open filename model dat family mod1 uio data mode open filename model txt family mod2 examination uio_data mode content uio_data mode files reading uio_data mode read value rho rho uio_data value temp temp uio_data value p p filename model dat uio_data value p p family modi plot oi uo d p uio det and closing uio data mode close filename model dat uio data mode close filename model txt uio
125. ongly recommended o formatted slow machine independent output big files e character outconv full The conversion type see Sect 4 3 1 of the full model files can be specified e g with character outconv full f A80 b 80 n Output file conversion amp c0 ieee_4 ieee_8 crayxmp_8 native ieee_4 The allowed values depend on the machine Leaving this field empty means that the default is chosen that is build into the local UIO module If the type ieee_4 is supported which is always the case so far it should be chosen e character outform_mean The format see Sect 4 3 1 of the additional data files can be chosen e g with character outform mean f A80 b 80 n Output file format amp cO formatted unformatted unformatted Allowed values are o unformatted default fast compact possibly machine dependent output strongly recommended o formatted slow machine independent output big files e character outconv mean The conversion type see Sect 4 3 1 of the additional data files can be specified e g with character outconv mean f A80 b 80 n Output file conversion amp c0 ieee_4 ieee_8 crayxmp_8 native ieee 4 The allowed values depend on the machine Leaving this field empty means that the default is chosen that is build into the local UIO module If the type ieee 4 is supported which is always the case so far it should be chosen 94 5 CONTROL AND DATA FILES 5 3 14 Additional In
126. ory additional output Produce timing information for main short characteristics routine It can be used together with OpenMP and should cause no noticeable performance loss Values o undefined default no timing information o defined call subroutines to measure elapsed time MSrad raytas in MSrad3D F90 Matthias Steffen radiation ray tau s Category performance enhancement Values o 0 default Loop with IF THEN ELSE o 1 Loop with ABS SIGN o 2 Loop with MIN MAX Optimization Compiler Switches In this section some mandatory or useful compiler flags are described These have different functions Enable necessary macro processing expansion for the F90 files Force proper handling of binary I O Choose module for radiative transfer Activate module for dust formation and or magnetic field transport Enable parallelization with OpenMP directives Choose a version of a subroutine or loop which is optimized for a specific architecture Tell the compiler if and what to inline Improve the general performance 34 3 PROGRAM FILES INSTALLATION COMPILATION 3 8 1 General OpenMP settings To activate OpenMP you have to set the corresponding environment variable see 3 6 before calling the configure script like export F90 PARALLEL openmp configure make This will insert the corresponding compiler switch e g openmp omp mp confer the following sections into the compiler calls in the makefile see Sect
127. osen e g with character outform fine f A80 b 80 n Output file format amp cO formatted unformatted unformatted Allowed values are o unformatted default fast compact possibly machine dependent output strongly recommended o formatted slow machine independent output big files This parameter can be specified but there is no corresponding output file in CO5BOLD yet e character outconv fine The conversion type see Sect 4 3 1 of the files with frequent output can be specified e g with character outconv fine f A80 b 80 n Output file conversion amp c0 ieee_4 ieee_8 crayxmp_8 native ieee_4 The allowed values depend on the machine Leaving this field empty means that the default is chosen that is build into the local UIO module If the type ieee_4 is supported which is always the case so far it should be chosen This parameter can be specified but there is no corresponding output file in CO5BOLD yet 5 4 Additional Control and Status Files rhd stop rhd cont and rhd done Before each time step CO5BOLD checks in the working directory whether the file rhd stop exists If it has been generated e g with touch rhd stop the code exits gracefully i e it produces a proper final model which can be used to restart the code This method of stopping a simulation is to be preferred over a simple kill or qdel command because it allows to analyze the state of the model just at the end of the simulation and
128. outines rad hdrad rhd rad module f90 RAD interface for Matthias radiation routine rad hdrad MSrad3D F90 RAD Matthias radiation transport routines long characteristics periodic sides hd dust rhd dust module F90 DUST dust molecule formation hd dust dust k3mon module f DUST 1 or 2 component dust model dust momentc2 module f DUST 4 moment dust model C2 INC DUST dust include file C9 molecule C2H INC DUST dust include file C2H molecule C2H2 INC DUST dust include file C2H molecule CHPAR_CT INC DUST dust include file DINDEX INC DUST dust include file DKSPLINT INC DUST dust include file H2 INC DUST dust include file H molecule hd mhd rhd mhd module F90 MHD magnetic fields first version eos f90 gasinter_routines f90 EOS equation of state opa opta cubit_module f OPTA cubic interpolation opa opta opta par module f90 OPTA parameters for opacity routines opa opta opta routines f OPTA opacity Table 3 List of all high level modules the table shows the file name with part of its path the shortcut for the directory and its description 22 3 PROGRAM FILES INSTALLATION COMPILATION File and path Abb Description hd rhdb rhd action module f90 RHDB routines for control parameter passing hd rhdb rhd box module f90 RHDB box handling routines hd rhdb rhd dat module f90 RHDB handling of additional data averages hd rhdb rhd gl module f90 RHDB global parameters hd rhdb rhd io module f90 RHDB input output routines hd
129. parameter one would try to adjust The total amount of viscosity should be controlled with real c vissmagorinsky and real c visartificial But the parameter can be used to tentatively switch off the off diagonal terms completely or to change its importance compared to the other terms e real c vistensordiv The factor in the stress tensor in front of of the divergence terms also on the diagonal can be set with e g real c vistensordiv f E15 8 b 4 amp n Divergence factor for viscous stress tensor u 1 amp cO typically 1 3 0 0 This is not really parameter one would try to adjust The total amount of viscosity should be controlled with real c vissmagorinsky and real c visartificial But the parameter can be used to switch off the divergence terms completely or to change its importance compared to the other terms These divergence terms can be used to reduce the effect of the tensor viscosity in the case of isotropic compression But this reduction c vistensordiv 0 333333 in 3D c vistensordiv 0 5 in 2D is usually switched off e integer n viscellsperchunk The number of cells per box or chunk treated by the tensor viscosity scheme at one call and by one thread can be set e g with integer n viscellsperchunk f 19 b 4 amp n Number of cells per viscosity chunk amp c0 70 gt old chopping amp c0 12000 reasonable value 20000 It can be adjusted to improve cache efficiency and to modify t
130. pf function u 1 cO q T tau Teff 4 0 75 tau tauross tau5000 t pgas pel vmic q 2 00E 07 6 539 3900 0 769 3 140 0 00 0 27637 2 50E 07 6 279 3920 1 171 2 752 0 00 0 28208 5 00E 07 5 868 3970 1 598 2 342 0 00 0 29675 1 00E 06 5 588 4030 1 842 2 105 0 00 0 31510 2 00E 06 5 334 4080 2 042 1 910 0 00 0 33103 5 00E 06 5 001 4160 2 279 1 674 0 00 0 35776 1 00E 05 4 747 4210 2 450 1 508 0 00 0 37527 2 00E 05 4 486 4270 2 618 1 341 0 00 0 39712 5 00E 05 4 131 4340 2 835 1 128 0 00 0 42377 1 00E 04 3 856 4400 2 999 0 967 0 50 0 44765 2 00E 04 3 577 4460 3 162 0 804 0 50 0 47248 5 00E 04 3 200 4530 3 377 0 596 0 50 0 50256 1 00E 03 2 912 4590 3 541 0 437 0 50 0 52925 2 00E 03 2 621 4640 3 704 0 279 0 50 0 55173 5 00E 03 2 233 4720 3 919 0 070 0 50 0 58792 1 00E 02 1 939 4800 4 083 0 094 0 50 0 62415 2 00E 02 1 645 4900 4 245 0 266 0 65 0 66867 5 00E 02 1 256 5080 4 460 0 504 0 85 0 74558 1 00E 01 0 961 5260 4 622 0 705 1 00 0 81447 2 50E 01 0 571 5560 4 830 1 002 1 25 0 89163 4 00E 01 0 371 5850 4 926 1 251 1 40 0 99911 7 00E 01 0 133 6260 5 022 1 613 1 50 1 13453 1 00E 00 0 019 6570 5 070 1 879 1 60 1 22581 1 50E 00 0 191 6880 5 114 2 140 1 70 1 17659 2 00E400 0 312 7160 5 140 2 363 1 80 1 13964 4 00E 00 0 597 7920 5 191 2 889 1 80 0 70033 6 00E 00 0 761 8250 5 213 3 092 1 80 0 46595 8 00E 00 0 877 8420 5 229 3 196 1 80 1 99552 1 00E 01 0 967 8500 5 242 3 245 1 80 3 76404 52 4
131. pilers Version 7 4 was installed Most of the compiler switches given in the previous section were used except for the following modifications which either gave empirically a better performance or were recommended by the UKAFF Hints for users e Ofast Replaces 03 gave better performance e LNO cs1 32k 1s1 32 cs2 8M 1s2 128 Explicit cache architecture added The option Ofast is now the default selected by the configure script for all SGIs with IP35 architecture T he cache architecture settings are activated for the UKAFF machines only A glitch in the system libraries made it necessary to add a work around to the source code file rhd F90 A bus error occurred whenever the system routine flush was trying to flush an empty file buffer The temporary work around was simply to add a write statement before every call of flush This made the log file look less nice but did the job Now the few flush statements that are not neccessarily preceded by a write statement are removed The main goal was to investigate the scaling of the performance of CO5BOLD with the number of processors This was done only for the MSrad module considering local surface convection http www nsc liu se systems sgi3k hitp techpubs sgi com library tpl cgi bin init cgi 1 44 3 PROGRAM FILES INSTALLATION COMPILATION models Two model sizes were tested a small one with 125x125x81 grid points employing non grey radiative trans
132. pressure gradient in a hydrostatic stratification a fluctuation due to shock waves In case of strong stratification this can lead to weird effects With activated correction the Roe solver treats only the deviations from a hydrostatic stratification as due to waves or shocks Several correction formulas have been tried The latest is the recommended default Values o 0 No pressure correction terms in Roe solver o 1 Simple correction with rhomean no new average pressure o 2 Simple correction with rhomean new average pressure o 3 Correction with local rho limited new average pressure o 4 Correction with local rho new different formula average pressure o 5 default Correction with local rho new limit new average pressure rhd hyd entropyfix pO1 in rhd hyd module F90 rhd hydrodynamics entropy fix parameter 01 Category performance enhancement The entropy fix can be done in one of two ways to get optimum performance with essentially the same results Values o 0 default if then else construction o 1 use a mask and the signum function rhd hyd upwind pO1 in rhd hyd module F90 rhd hydrodynamics upwind parameter 01 Category performance enhancement The determination of the upwind direction can be done in one of two ways to get optimum performance with essentially the same results Values o 0 default if then else construction o 1 use a mask and the sign
133. pressure mode amp c0 on off on Allowed values are o on Radiation pressure on o off Radiation pressure off The scheme is pretty slow and wrong in the optically thick Do not use real c_radtintminfac In the LHDrad module The fraction the interpolated temperature at a point on the ray may exceed the minimum temperature at its four neighbors on the HD grid can be set e g with real c_radtintminfac f E15 8 b 4 amp n Temperature interpolation parameter u 1 amp c0O lt 1 0 only bilinear 1 1 reasonable weighting between min und bil 0 0 The introduction of this parameter was an attempt to reduce the negative cooling effect of a single hot cell on its cool neighbors It should be switched off e g by setting it to 0 0 integer dtimestep_out_fine This parameter can be specified but there is no corresponding output file in CO5BOLD yet integer dtimestep out fine f I4 b 4 n Output time step number u 1 amp cO dtimestep out fine O gt no output 1 character outfile_fine The name of the file for the output of additional information at regular small intervals can be specified with e g 96 5 CONTROL AND DATA FILES character outfile fine f A80 b 80 n Output file name rhd fine Leaving it empty means that no file of this type is written Specifying it means the same yet e character outform fine The format see Sect 4 3 1 of the files with frequent output can be ch
134. processing of COSBOLD data Moreover there are already prepared IDL scripts for this purpose but most of them are rather complex and still have to be edited e g changing file names Huuhh For the beginning it is more clear to use single commands We give a short overview of some essential commands 7 21 Loading the Parameter File IDL gt parfile mymodel par IDL gt par uio_struct_rd parfile All control parameters are provided in the structure PAR 7 2 2 Loading CO5BOLD Data full sta end IDL gt modelfile home user mymodel full amp n 0 IDL gt ful uio_dataset_rd modelfile n n First the name and full path if not in the actual directory of the model file and the wanted time step should be defined Here time step means the consecutive number of the model snapshot in the file Declaring a time step number greater than the number of snapshots 7 2 COSBOLD Data in IDL 107 contained in the file will cause an error Otherwise all data of the particular time step n will be provided in the structure FUL Loading more than one timestep from the same file could be achieved as follows IDL gt uio_openrd nc modelfile outstr ierr IDL gt for i 0 ntime 1 do begin amp IDL gt ful uio dataset rd modelfile channel nc ierr ierr outstr err msg IDL gt endfor amp IDL gt uio_closrd nc Again this operation would cause an error if the wanted time step is not contained in the file Via checking the error
135. ption d 1 1 f A14 p 2 b 14 n File description character history d 1 20 f A80 p 1 b 80 n File history character version f A80 b 80 n Program version label dataset n RHD model label enddataset label dataset n RHD model label enddataset 52 File with Additional Data rhd mean 65 Each dataset has the following structure for a supergiant simulation label dataset n RHD model date 25 05 2001 09 41 29 405 label box date 25 05 2001 09 41 29 408 character box id f A80 b 80 n Block identification rad label endbox label box date 25 05 2001 09 41 29 983 character box id f A2 b 2 n Block identification zi label endbox label box date 25 05 2001 09 41 30 078 character box id f A2 b 2 n Block identification z2 label endbox label box date 25 05 2001 09 41 30 170 character box id f A2 b 2 n Block identification z3 label endbox label box date 25 05 2001 09 41 30 260 character box id f A1 b 1 n Block identification r label endbox label box date 25 05 2001 09 41 30 359 character box id f A80 b 80 n Block identification Zz label endbox label enddataset date 25 05 2001 09 41 30 489 There a six sub blocks delimited with box and endbox labels They contain surface intensity and flux arrays rad averages in the 23 plane z1 the 13 plane z2 the 12 plane z3 and over spherical shells r and a 2D slice through the model z 5 2 2 Con
136. r an identifier e g box time description and ad ditional information about array size e g d 63 63 63 63 63 63 data format e g E13 6 p 4 b 4 and properties of the quantity e g n Density u g cm 3 Each start rhd sta or final rhd end model file has a structure as shown above The rhd ful11 file usually contains a sequence of these data sets which of course can also be used as start model of a simulation The axes xci xc2 and xc3 describe the positions of the cell centers The axes xbi xb2 and xb3 contain the positions of the cell boundaries they have one element more than the corresponding cell centered quantity Cell boundaries should be centered in the middle between cell centers for the best represen tation of radiative fluxes with MSrad3D 5 2 File with Additional Data rhd mean A rhd mean file contains derived data averaged fluxes other averaged quantities surface intensities in addition to the complete data sets in rhd full files It has more entries than a full model file However they are much smaller Therefore one can afford a higher output sampling rate Its format is usually Fortran unformatted binary 5 2 1 Organization of rhd mean File A mean file usually consists of several datasets The overall structure is fileform uio form unformatted convert ieee 4 character file id f A8 b 8 n File identification character descri
137. r they are still used e g by the makefile that produces the executables that are called by the UIO scripts The environment variables for the UIO routines can be e g UIOSRCPATH home user for uio f90 UIOEXEPATH home user for uio f90 sun A script to set all necessary variables and paths can be here for the Bourne shell bin sh Disk where all Fortran programs are located FORTRANDISK HOME for export FORTRANDISK if uname s craSH then 4 Kiel cra5Hi UIOMAC uio mac crayxmp module export UIOMAC RHDMAC rhd mac cray module export RHDMAC elif uname s craSH then Kiel craSH UIOMAC uio mac crayts module export UIOMAC RHDMAC rhd mac cray module export RHDMAC elif uname m SR8000 then 3 4 Old Setup File for Paths Potsdam Hitachi UIOMAC uio mac hitachi module export UIOMAC RHDMAC rhd mac hitachi module export RHDMAC else Default Suns MAC files UIOMAC uio mac sun module export UIOMAC RHDMAC rhd mac sun module export RHDMAC fi Architecture dependent sub directory names for object file and executables if uname s craSH then MAC crash elif uname s craSHi then MAC crashi elif uname s SunOS then MAC sun elif uname s HP UX then if uname m ia64 then MAC hpia64 else MAC hp fi elif uname s Li
138. rad For a Star in a Box and particularly when only simple ray directions are allowed in the radiation tranport step the temperature in the outer corners of the box tends to become very small To artificially increase the effect of radiative heating the parameter real r1 rad can specify a radius beyond which only postive contributions of the radiative energy transport to the energy budget are taken into account This ruins the conservativity of the code in these layers and should be applied only in very remote corners which are then considered only as sort of extended boundary region but not as part of the real model The parameter can be specified e g with real ri rad f E15 8 b 4 n Outer radiation transport radius u cm amp c0O 0 0 Not used 8 00000et13 A value of 0 0 default or below deactivates this feature 72 5 CONTROL AND DATA FILES 5 3 4 Boundary Conditions The boundary conditions at the six sides of the computational box cannot be specified indepen dently For the naming convention of the boundaries a gravitational acceleration in x3 direction is assumed Accordingly there is a bottom a top and four side boundaries e character side bound The boundary condition at all four sides is given by e g character side bound f A80 b 80 n side boundary conditions amp cO closed transmitting periodic transmitting Possible values are o reflective closed wall no gravity no radiation
139. raytas in file M9rad3D F90 See Sect http www cray com craydoc http www compaq com fortran docs index html 36 3 PROGRAM FILES INSTALLATION COMPILATION 3 8 5 Hewlett Packard V2500 The 12 processor machine zeipel from Hewlett Packard is a V2500 PA 2 0 system Now there is a first success to force the compiler to accept the OpenMP directives in CO5BOLD Yet when running on several processors only some routines e g rhd_shortrad_dirsimple1 in CO5BOLD can benefit while others rhd_shortrad_dirsimple2 rhd_shortrad_dirsimple3 are significantly slower than on one processor In addition the single processor performance is not very good partly because the achievable optimization level is not very high Some macros which seem to be necessary e U77 Link proper library to make the machine understand e g call flush 6 e cpp yes Switch on the C preprocessor Note that all Fortran90 files have to end with 90 The F90 suffix does not seem to work e tOparallel tOopenmp Onoautopar Try to enable parallelization with OpenMP direc tives disable auto parallelization e tOnoinline Disables inlining This can simplify things With a proper choice of routine versions inlining is not really necessary anymore e 03 0limit General optimization with limited resource usage during compilation Some modules should only be compiled with 02 others compile even with 03 Onolimit The UIO
140. rd ncin termt ntt rho ident ilb ilow 1 1 Alternatively it is possible to search in the file for a special entry or to search in a specially generated entry list with call uio srhd ncin termt ntt type real ident rho outstr outstr ierr ierr Additionally the module uio var module makes it possible to read any entry into an UIO flexible variable and the module uio varfile module allows the reading of a complete file into a special file structure of UIO flexible variables To close a file after reading use uio closrd ncin There are several examples of programs with UIO routines like uio var test f90 uio varfile test f90 uiotst f90 uio demo f90 and of course CO5BOLD 4 6 UNIX Scripts So far there exist three UNIX shell scripts useful to quickly examine data sets uiolook to change the format or conversion type of files uiocat or to print some information about the conversion types possible on the local machine uioinfo 4 6 1 Installation of UIO UNIX Scripts Recently the installation procedure for the UIO scripts has been updated to make use of its own configure script Therefore the procedure should now look like tar zxvf for tar gz cd for uio f90 YOUR MACHINE configure make make install 4 6 UNIX Scripts 57 or tar zxvf for tar gz cd for uio f90 mkdir YOUR MACHINE cd YOUR MACHINE In s conf configure configure make make install Some of the environment variabl
141. ressiveness of the reconstruction scheme with e g character reconstruction f A80 b 80 n Reconstruction method amp cO Constant ci Minmod VanLeer Superbee c2 PP Minmod Possible values are o Constant The run of the partial waves inside the cells is assumed to be constant A highly dissipative first order scheme results This values will usually only be used for test or comparison purposes o Minmod Chooses the smallest slope which still results in a second order scheme It is the most diffusive and most stable one in this class o VanLeer default The recommended second order scheme o Superbee The most aggressive stable 2nd order scheme It results in the steepest shocks which works well in some test cases but might be to difficult for the radiation transport module to handle o PP Chooses the piecewise parabolic reconstruction of the PPM scheme Piecewise Parabolic Method Colella amp Woodward 1984 Results in 3rd order accuracy for the advection Usually the VanLeer reconstruction is a good choice If a more stable and diffusive scheme is needed take Minmod The PP reconstruction gives the highest accuracy However it tends to produce somewhat noisy models with small wiggles e g in the velocity real c slopered When Drhd roeid slope 101 2 is set see Sect 3 7 a new extra stabilization mechanism can be activated If one of the reconstruction methods VanLeer Superbee or PP see Se
142. rhdb rhd mac cray module f90 RHDB machine dependent routines CRAY hd rhdb rhd mac default module f90 RHDB machine dependent routines default hd rhdb rhd mac hitachi module f90 RHDB machine dependent routines Hitachi O hd rhdb rhd mac sun module f90 RHDB mac dependent routines Sun others hd rhdb rhd mean module f90 RHDB averaging routines hd rhdb rhd prop module f90 RHDB box properties hd rhdb rhd sub module f90 RHDB additional routines con f90 const module f90 CON physical and mathematical constants mat str str module f90 STR string handling time f90 timing module f90 timing routines zizie D0Z S uio f90 uio_base_module f90 I O routines uio f90 uio_bulk_module f90 I O routines uio f90 uio_filedef_module f90 UIO I O routines uio f90 uio mac crayts module f90 UIO I O routines machine dependent part uio f90 uio mac crayxmp module f90 UIO I O routines machine dependent part uio f90 uio mac decalpha module f90 UIO I O routines machine dependent part uio f90 uio mac hitachi module f90 UIO I O routines machine dependent part uio 90 uio mac ieee module f90 UIO I O routines machine dependent part uio f90 uio mac intel module f90 UIO I O routines machine dependent part uio f90 uio mac module f90 UIO I O routines m d minimal version uio f90 uio mac nec module f90 UIO I O routines machine dependent part uio f90 uio mac sun module f90 UIO I O m d works in most cases Table 4 List of all low level module
143. routines F90 gas interpolation 101 Category performance enhancement This switch determines how temporary arrays are handled to improve performance Values o 0 default Temporary coefficient arrays are actually copied o 1 Temporary coefficient arrays just get a pointer link into the big arrays e rhd box grav0l in rhd box module F90 rhd box gravitation 01 Category feature activation Switch to activate the array for the gravitational potential in the box structure If the switch is set to 1 a 3D array for the potential is created copied removed There is no module to compute the gravitational potential yet Therefore the entire thing has no practical value yet Values o 0 default no handling of array o 1 array handling activated e rhd box qucO1 inrhd box module F90 and rhd F90 rhd box quantity centered 01 Category feature activation 3 7 Compiler Macros 27 Now CO5BOLD is able to handle a number of additional quantities e g density arrays in addition to the basic hydrodynamics quantities p ei if this compiler switch is activated These additional quantities can be e g densities of dust distribution moments or densities of molecules Values o 0 default no handling of additional quantities density arrays o 1 handling of additional density arrays is activated To actually include dust formation in a simulation it is necessary to 1 set the switch Drhd box guc0
144. rs but that already the immediate neighbors have longer timescales and can damp the most sensitive points Based on real c radcourant the recommended typical radiative time step is computed real c radcourantmax With this parameter the maximum allowed radiative time step is prescribed as e g in real c radcourantmax f E15 8 b 4 n maximum RAD Courant factor u 1 amp cO range C radCourant lt C radCourantmax typically 2 0 2 0 This value will typically be somewhat larger than real c radcourant real c radmaxeichange The relative energy change per radiative sub step can be restricted e g with real c radmaxeichange f E15 8 b 4 n maximum radiative energy change amp u ti c range 0 01 1 0 0 25 The default is 0 5 Values between 0 1 and 0 5 seem reasonable A value lt 0 0 deactivates this time step check However the check of the radiative energy change should usually be performed A way to maximize the radiative time step and therefore the performance of the entire code can be to first set real c radmaxeichange to a proper value say 0 25 Then real c radcourant and real c radcourantmax are adjusted by trial and error in a way that the radiative time step is almost always restricted by the Courant condition and only sometimes in extreme cases by the maximum energy change restriction The computed output intensity should be checked for the size of its fluctuations due to a possibly too large value of
145. s e ip Optimization activate interprocedural optimization within each source file This enables inlining e DMSrad raytas 2 Optimization choose non default version of loop in SUBROUTINE raytas in file M rad3D F90 See Sect e Drhd shortrad dirl 101 1 Optimization Transpose arrays and use routine rhd shortrad dir3 for rays in x1 direction See Sect openmp Parallelization OpenMP directives are activated Note that the UIO routines should be compiled without OpenMP support even if they do not contain any OpenMP directives themselves e i dynamic Helpful against undefined reference to __ctype_b errors Using the Intel compiler there is a problem with the UIO modules when OpenMP is activated This a bit weird because the UIO modules do not contain any OpenMP directives However this means that OpenMP can be safely deactivated for these modules A proposed compiling sequence is all modules activated export F90_LHDRAD 1 export F90_MSRAD 1 export F90_SHORTRAD 1 export F90 DUST 1 42 3 PROGRAM FILES INSTALLATION COMPILATION export F90_MHD 1 export F90 PARALLEL scalar configure make UIO export F90_PARALLEL openmp configure make In some cases it migh be helpful to set export LD ASSUME KERNEL 2 4 19 when encountering problems with OpenMP see the Release Notes of the Intel compile 3 8 11 NEC SX 5 First attempts to compile COSBOLD on neSH at the Rechenzentrum Kiel and on hwwsx5 at
146. s the table shows the file name with part of its path the shortcut for the directory and its description 3 5 Fortran Files 23 File and path Abb Description mat str str module f90 STR string handling uio f90 uio base module f90 UIO I O routines uio f90 uio bulk module f90 UIO I O routines uio f90 uio filedef module f90 UIO I O routines uio f 90 uio mac sun module f90 UIO I O routines machine dependent part eos f90 gasinter routines f90 EOS equation of state rad hdrad cubit module f RAD cubic interpolation rad hdrad opta par module f90 RAD parameters for opacity routines rad hdrad opta routines f RAD opacity rad hdrad MSrad3D F90 RAD Matthias radiation transport routines long characteristics periodic sides hd rhd timing module f90 RHD timing routines hd rhd rhd const module f90 RHD physical and mathematical constants hd rhd rhd gl module f90 RHD global parameters hd rhd rhd action module f90 RHD routines for control parameter passing hd rhd rhd box module f90 RHD box handling routines hd rhd rhd dat module f90 RHD handling of additional data averages hd rhd rhd mean module f90 RHD averaging routines hd rhd rhd io module f90 RHD input output routines hd rhd rhd mac cray module f90 RHD machine dependent routines CRAY hd rhd rhd mac default module f90 RHD machine dependent routines default hd rhd rhd mac sun module f90 RHD machine dependent routines Sun hd rhd rhd sub module f90 RHD additiona
147. s can be edited to account for individual choices and the target machine With FORTRANDISK HOME for export FORTRANDISK 3 5 Fortran Files 21 the master directory is specified With UIOMAC uio_mac_sun_module export UIOMAC RHDMAC rhd_mac_sun_module export RHDMAC you set some machine dependent modules The sun modules work for most machines e g for Linux Intel AMD machines With MAC linux you specify the name of the subdirectories with the makefiles The other lines only have to be edited if you want to organize the directories in a completely different way In this case you have to adapt the configure script too 3 5 Fortran Files Tables 3 and 4 show a list of all source files necessary to compile the complete version of CO5BOLD Table 5 shows the former list File and path Abb Description hd rhd rhd F90 RHD main program hd rhd rhd hyd module F90 RHD hydrodynamics routines hd rhd rhd Ihdrad module F90 RHD radiative transfer routines long characteristics supergiant case hd rhd rhd shortrad module F90 RHD radiative transfer routines short characteristics supergiant case hd rhd rhd shortrad dtauopO1 f90 RHD short characteristics tau coupling hd rhd rhd shortrad dtauop02 f90 RHD short characteristics tau coupling hd rhd rhd shortrad operator00 f90 RHD short characteristics operator hd rhd rhd shortrad operator08 f90 RHD short characteristics operator hd rhd rhd vis module F90 RHD tensor viscosity r
148. se of a constant gravity the amount of the acceleration has to specified with real grav f E15 8 b 4 n Gravity u cm s 2 27500 0 Setting this value to zero switches off gravity oh wonder real mass_star In the case of a central the mass in cgs units of the star has to be specified with real mass star f E15 8 b 4 n Stellar Mass u g 9 94500et33 real r0 grav To avoid the central singularity in a 1 r potential it is smoothed in the center to give a central potential of 1 r0 grav specified with real ro grav f E15 8 b 4 n Inner Smoothing Radius u cm 9 45833e 12 This parameter should always be non zero for a central potential real ri_grav The density in an atmosphere in hydrostatic equilibrium can decline to very low values To artificial enlarge the pressure and density scale height in the outer layers of the star the corners of the box the gravity can be reduced by defining the potential at infinity to be 1 ri_grav specified with real ri grav f E15 8 b 4 n Quter Smoothing Radius u cm amp c0 70 0 Not used 11 35000et13 Setting this parameter to zero gives the usual 1 r behavior of the potential in the outer layers but also chooses another smoothing formula in the central part where real r0 grav is relevant But a value somewhat larger than the remotest corner of the box effectively cancels this artificial smoothing in the outer layers without changing the formula for the potential real rl
149. star e abundances eosfile opafile check the paths 6 If the gravity of the new model and therefore the characteristic time scale signifi cantly deviates from the old one the time specifications controlling the output frequency dtime out full dtime out mean the total length of the simulation if specified as stel lar time endtime plustime and absolute boundaries specifications for the time step dtime min dtime min stop dtime max dtime start have to be scaled Look for pa rameters with units u s see Sections 5 3 12 and 5 3 13 T l he rest of the parameters controls additional details Most of the constants are specified in dimensionless form and keep their value in a class of related simulations The previously used values will probably be reasonable for the new simulation too Of course a complete control of CO5BOLD is only possible after studying of the meaning of the parameters in detail e g by reading the following pages AND unfortunately an accompanying look into the source code itself 5 3 2 Header The header of the parameter file contains information about the file format and contents The description array can be used to specify the goal of the simulation special model characteristics or important parameter changes compared to a previous or standard model The history array may contain the predecessor of the parameter file to simplify a tracing of parameter changes e fileform uio Th
150. tents of Individual rhd mean File Entry An individual box inside a dataset entry in a mean file can have e g the following contents describing horizontal averages in a plane parallel model With uiolook chro2D03c008 01 mean less you get this and more label box date 06 11 2002 17 58 05 533 character box id f A2 b 2 n Block identification integer dimension d 1 2 1 3 f I7 p 6 b 4 66 real time f E13 6 b 4 integer itime f I10 b 4 real xci d 1 1 1 1 1 1 f E13 6 p 4 b 4 ds 0 xc2 d 1 0 0 1 0 1 real 1 1 1 1 1 f E13 6 p 4 b 4 ds 0 xc3 d 1 TA PO e o e O real ds 0 xb1 d 1 0 1 f E13 6 p 4 b 4 BR is e O real ds 0 xb2 d 1 z1 real 1 f E13 6 p 4 b 4 Ne ds 0 xb3 d 1 212031 real 1 1 gt ds 0 1 0 1 0 1 real rho_xmean d 1 1 1 1 1 120 f E13 6 real real real real vi xmean2 d 1 1 1 1 1 120 f E13 6 real v2 xmean2 d 1 1 1 1 1 120 f E13 6 real v3 xmean2 d 1 1 1 1 1 120 f E13 6 real rhovi xmean d 1 1 1 1 1 120 f E13 real rhov2 xmean d 1 1 1 1 1 120 f E13 real rhov3 xmean d 1 1 1 1 1 120 f E13 real bci xmean d 1 1 1 1 1 120 f E13 6 real bc2 xmean d 1 1 1 1 1 120 f E13 6 real bc3 xmean d 1 1 1 1 1 120 f E13 6 120 f E13 6 p 4 b 121 f E13 6 p 4 b vi xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 v2 xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 v3 xmean d 1 1 1 1 1 120 f E13 6 p 4 b 4 5 CONTR
151. ters at CINES See e g the excellent SGI Fortran90 manual Information about the CINES machines can be found under CINES or CINES Introduction Important switches are macro expand Enable macro expansion e mp Enable parallelization with OpenMP directives INLINE aggressive ON INLINE list INLINE preempt 0N General keywords for in lining INLINE must Optimization routines that should be inlined see Sect 3 8 2 e Ofast OPT Olimit 0 General optimization On older compiler versions 03 was the achievable optimum e IPA plimit 5500 Even more optimization This option requires lots of memory 1 GByte To get it it might be necessary to ask for more than one processor for the compi lation especially on the CINES machines e CG longbranch limit 60000 This switch limits needed compiler resources It is sug gested by the compiler on the CINES machines itself e Drhd roeid step 101 1 Slight performance improvement 3 8 13 SGI Origin 2000 3800 at UKAFF CO5BOLD has been also compiled and tested on up to 22 processors on the machines of the UK Astrophysical Fluids Facility UKAFF Jin Leicester England UKAFF operates two machines an SGI Origin 3800 with 128 processors named ukaff and an older SGI Origin 2000 with 22 processors named grand which is mainly used for development and test purposes Both machines are binary compatible At the time of testing April 2003 the SGI MIPSpro Com
152. that should usually be chosen real dtime_min_stop Sometimes a simulation can run into a pathological state where the time step decreases rapidly without recovering To prevent a simulation in such a case from running forever or until some other process time restriction applies without actually advancing significantly in time it is possible to specify an absolute minimum time step e g with 5 3 Parameter File rhd par 89 real dtime min stop f E15 8 b 4 n Minimum time step u s amp c0 dtime min stop 0 0 gt no restriction amp ci dtime lt dtime min stop gt program stop 1 0 If the actual time step falls below this value the simulation finishes gracefully This val ues has to specified as absolute time and has to be chosen carefully for each individual model or each group of models This time step restriction can be switched off by setting real dtime min stop 0 0 But in general one should keep it activated and try to find a proper positive value e real dtime incmax Sometimes a time step restriction can lead to a sudden drastic drop in the time step To prevent unwanted oscillations in the size of the time step its increase can be restricted e g with real dtime incmax f E15 8 b 4 n Maximum time step increment factor u 1 amp c0 dtime max 1 0 gt no restriction ci typically 1 1 1 2 This value specifies the maximum factor by which the time step can be increased from step to step even if th
153. ting during a simulation and the total length of a job If either one of the halt conditions below is met CO5BOLD finishes the current step writes a final model plus some final information to other files and stops execution For example on a CRAY one typically wants to use most of the CPU time given for an individual batch job In this case one can set e g real cputime remainlimit 2000 0 and the values for the other halt conditions to 1 0 or 1 e real starttime The start time of a simulation is usually taken from the start model file But sometimes is simulation is to be started with the final model of a previous run but should start at time 0 0 This can be achieved by setting the start time with real starttime f E15 8 b 4 n Start time u s 0 0 Allowed values are o gt 0 0 Set the initial time of the simulation to this value and override value from start model o lt 0 0 default Take the initial time from start model integer starttimestep The start time step count of a simulation is usually taken from the start model file But sometimes is simulation is to be started with the final model of a previous run but should start at time step 0 This can be achieved by setting the start time step count with 5 3 Parameter File rhd par 87 integer starttimestep f I11 b 4 n Start time step number u 1 0 Allowed values are o gt 0 Set the initial time step of the simulation to this value and override value fro
154. tion 3 dimensions from ray loop 01 Category performance enhancement There might be a performance gain by splitting a big loop in routine rhd rad3d fromray into two separate loops Typically one big loop is to be preferred Values o undefined default One big loop o defined Two smaller loops rhd rad3d r02 in rhd Ihdrad module F90 rhd radiation 3 dimensions radiation 02 Category test Module rhd Ihdrad module contains a routine for the handling of periodic boundaries It is in an experimental state and is deactivated by default Values o undefined default Skip routine rhd_rad3d_dirper during compilation o defined Compile routine rhd rad3d dirper rhd rad3d solve t01 in rhd Ihdrad module F90 rhd radiation 3 dimensions solve timing 01 Category additional output Produce timing information for the routines which solves the 1D radiation transport egua tion along single ray This routine is called very freguently The timing measurement might slow it down somewhat It should not be used in conjunction with OpenMP Values 3 7 Compiler Macros 31 o undefined default no timing information o defined call subroutines to measure elapsed time rhd rad3d dir t01 in rhd Ihdrad module F90 rhd radiation 3 dimensions direction timing 01 Category additional output Produce timing information for the routines which solves the radiation transport equation for one direction field T he timin
155. tmyer type u 1 0 0 e real c visneu2 real c visneu2 f E15 8 b 4 amp n Quadratic viscosity parameter von Neumann Richtmyer type u 1 0 0 e real c radkappasmooth In the LHDrad module the opacity along each ray can be smoothed The amount of smooth ing can be set e g with 5 3 Parameter File rhd par 95 real c_radkappasmooth f E15 8 b 4 n Opacity smoothing parameter u 1 amp c0 70 0 no smoothing 0 25 light smoothing 0 666 strong smoothing 0 0 The smoothing can perhaps reduce the noise in the intensity images somewhat but has no general beneficial effect and should usually not be used real c radtsmooth In the LHDrad module the 3D temperature array can be smoothed The amount of smooth ing can be set e g with real c_radtsmooth f E15 8 b 4 n Temperature smoothing parameter u 1 amp c0 70 0 no smoothing 0 5 reasonable smoothing 1 0 max smoothing 0 0 The smoothing can sometimes reduce the noise in the intensity images but causes amplifies some anomalies of the radiative Greens function Some cool cell just above the sharp sub photospheric temperature drop are not heated but cool further down Negative temperature spikes may result This smoothing should not be used anymore character radpressure In the LHDrad module there exists a simple prescription for the radiative pressure reason able in the optically thin which can be activated with character radpressure f A80 b 80 n Radiation
156. tput Produce timing information for the routine which computes the Roe fluxes It should not be used in conjunction with OpenMP Values o undefined default no timing information o defined call subroutines to measure elapsed time e rhd roeid step t01 in rhd hyd module F90 rhd roe 1 dimension step timing 01 Category additional output Produce timing information for the routine which performs the Roe step It should not be used in conjunction with OpenMP Values o undefined default no timing information o defined call subroutines to measure elapsed time Hydrodynamics tensor viscosity e rhd vis density pO1 in rhd vis module F90 rhd viscosity density parameter 01 Category selection of approximation Choose formula for density average at cell boundary in tensor viscosity routines Values o 0 rhomean min rholeft rhoright o 1 default rhomean 0 5 rholeft rhoright e rhd vis t0l in rhd vis module F90 rhd viscosity timing 01 Category additional output Produce timing information for 2D 3D tensor viscosity routines It should not be used in conjunction with OpenMP Values o undefined default no timing information o defined call subroutines to measure elapsed time Radiation transport e rhd r01 in rhd F90 rhd radiation 01 Category feature activation Switch to include LHDrad radiation transport module It uses long characteristics and is restricted to an equid
157. try It can be used to keep infor mation about the parent parameter file as in character history d 1 2 f A80 p 1 b 80 n File history Taken from st35gm04n03 09 par Last Modification 01 01 2002 Its use is optional 5 3 3 Fundamental Model Parameters e real teff The effective temperature is one of the basic model parameters and is specified e g with real teff f F13 3 b 4 n Effective Temperature u K 3500 0 for a relatively cool star Note that the actual effective temperature can only be deter mined a posteriori and that the entropy of the instreaming entropy see below is more important than teff itself In fact teff is only used to control material properties at the outer boundary Its value should be close to the expected effective temperature of the model character grav mod Gravity is another characteristic of a stellar atmosphere The type or geometry of the external gravity field has to be specified e g with character grav mode f A80 b 80 n Type of gravity field amp cO constant central central 5 3 Parameter File rhd par 71 Two values are possible so far o constant In the standard solar case the constant gravity specified with real grav is directed downward in x3 direction o central For the supergiant case a central potential is assumed with an origin at x 0 The stellar mass as well as inner and outer smoothing radius have to be specified real grav In the ca
158. try or if it gives only additional information and is optional and therefore not necessary to specify A header line has a maximum length of 80 characters A continuation line is indicated by amp at the end of the line A header consists of 20 lines at maximum It can be preceded by empty lines except for the file header entry Example real time f F9 2 b 4 n Time u s c0 Simulation time in seconds amp ci Time count starts at 0 0 12 34 The entry header is followed by the entry data block This block is empty for labels and the fileform entry but non empty otherwise In an unformatted file each header line is an individual record containing a string with exactly 80 characters The following data block scalar or array is one single record In a formatted file each header line is a string of at most 80 characters delimited by a LINEFEED of whatever the operating system decided to be appropriate as EOL character sequence The following data block is written as sequence of lines The number of items per line is specified by the p keyword in the header 4 3 3 Tables For a table the entry header is followed by a list of headers for the individual table columns a single table header line consisting of abbreviations of the table entry identifiers and the table itself see the example in section 4 2 The dimension keyword gives the number of columns and rows in the form d 1 columns l rows The Holweger
159. ue See Tab 8 for some pre defined keywords entry type entry contents fileform file description first entry integer scalars 1D 4D arrays real scalars 1D 4D arrays single amp double precision complex scalars 1D 4D arrays single precision character scalars 1D 4D arrays table table with integer real character columns label label entry for file structuring Table 7 UIO entry types 50 4 UIO DATA FORMAT keyword description example descriptor info necessary b byte number 4 format yes d dimension 0 9 format yes arrays ds dimension shift 1 1 yes f Fortran format E13 6 format yes p values per line 4 format yes arrays t transformation log10 format n name density yes u unit g cm 3 yes date date 1 1 98 yes c0 9 comment Dichte yes form file format formatted file yes file header convert conversion ieee 4 file yes file header version version 0 1 1997 11 29 yes system system IRIX yes machine machine name atlas yes osrelease OS release 6 3 yes osversion OS version 12161207 yes hardware machine hardware IP32 yes language program language Fortran90 yes program program uiotst yes XYZ user defined source yes Table 8 Standard UIO entry header keywords The keyword is given with a short description and an example The fourth fifth and sixth column indicate if the keyword is a mandatory descriptor in the file header or for the format of an en
160. um function rhd hyd roeid 101 in rhd hyd module F90 rhd hydrodynamics roe 1 dimension loop 01 Category performance enhancement The computation of the Roe fluxes can be done by either of two sets of routines to find the set which gives optimum performance with essentially the same results Values o 0 default lots of small routines acting on scalars inlining needed cache reuse is optimized o 1 routines acting on arrays more temporary arrays necessary vectorization is easier rhd roeid flux 101 in rhd hyd module F90 rhd roe 1 dimension flux loop 01 Category test By setting this switch an alternative way of computing the upwind centered Roe states is activated only for constant reconstruction for performance test purposes only do not activate Values 3 7 Compiler Macros 29 o undefined default Use standard method to compute the Roe states o defined Use non standard method to compute the Roe states e rhd bound t01 in rhd hyd module F90 rhd bound timing 01 Category additional output Produce timing information for inner boundary routine central potential or lower and upper boundary routines constant gravitation It can be used together with OpenMP Values o undefined default no timing information o defined call subroutines to measure elapsed time e rhd roeld flux t01 in rhd hyd module F90 rhd roe 1 dimension flux timing 01 Category additional ou
161. unzip for tar gz tar xvf for tar instead Unpacking the tar file creates a sub directory for in the local directory and possible overwrites existing files You get sub sub directories as described in Sect and files as listed in Tables 3 and 4 See the Readme file for README 14 3 PROGRAM FILES INSTALLATION COMPILATION 3 Change with cd for hd rhd into the main directory Look at the existing sub directories e g with ls og grep d to see if you find one that fits your machine The directory for hd rhd conf should not be used It contains only the configure script But any other directory will do If you don t like any of the existing directories create your own e g with mkdir YOUR MACHINE Change into this directory with cd YOUR MACHINE Check if there is a configure script or a link to it with ls og configure which should give something like lrwxrwxrwx 1 17 2002 12 04 17 39 configure gt If it is not there create the link with In s conf configure configure This gives you a screen output like Configuration script for CO5BOLD Makefile conf configure Start the configure script to create the first version of the Makefile No parallelization requested assume default F90 PARALLEL scalar No debugging requested assume default F90_DEBUG 0 No LHDrad module requested assume default F90 LHDRAD O No MSrad module requested assume default F90 MSRAD O No
162. with an IDL routine Edit the parameter file rhd par typically you will start with an existing file and edit it You should check that the paths and names of EOS opacity and start file are set correctly Watch the username For details about the parameter file see Sect 5 3 1 Start the simulation with nice nohup rhd exe gt rhd out amp You can see how the simulation proceeds with tail f rhd out The other data files usually are in a binary format this can be changed to ASCII for matted in rhd par Their contents can be read and analyzed with IDL routines see Sect 1 7 3 For machines with batch queue there is a script which can handle an entire sequence of simulations see Sect 6 2 102 6 RUNNING A SIMULATION 6 2 Running CO5BOLD on a Machine with Batch System For longer simulations it is inconvenient to restart the individual jobs by hand This task is done by a script originally from Hans G nter Ludwig Its basic function is sketched in Fig 6 CO5BOLD function diagram Figure 6 Program scheme Here comes an example of the script rhd1 job for a system without dedicated batch system The submission of a job is done via nohup rhd1 job bin sh PBS N rhdi PBS 1 walltime 72 00 00 PBS 1 nodes 1 ppn 4 Job file for the execution of RHD on gunnar Itanium 2 Uppsala Source rhdi job original from HGL gt 6
163. wr3 or qarch pwr4 gtune pwr4 by setting an environment vari able e g F9O MACHINE pwr3 before calling configure e Q To activate inlining and to specify the list of routines that should be inlined see Sect 3 8 2 e qsmp noauto omp Parallelization OpenMP directives are activated A job script rhd job on the Loadleveler batch system on io cines fr can be submitted with llsubmit rhd job The jobs in the queue for user testuser can be checked with llq u testuser A job with ID n34 56789 0 can be cancelled with llcancel n34 56789 0 3 8 9 Linux PGI Compiler So far under Linux the compiler of the Portland Group has been used mostly to compile CO5BOLD It is called with pgf90 Important switches are e byteswapio With this flag set binary files in big endian format the standard for UIO files are automatically transformed to little endian and vice versa e fast General optimization flag to choose close to optimum optimization for local ma chine e Mvect sse Optimization Allow Pentium III vector commands e Mcache align Optimization Align some data object on cache line boundaries e fastsse From compiler version 4 0 on this option can be used instead of the three previous ones It contains and supersedes them e Minline Optimization the routines that should be inlined are listed in Sect e DMSrad raytas 2 Optimization choose non default version of loop in SUBROUTINE raytas in
164. xed derivatives in Eql4 together with Eq 28 Entropy change at constant pressure 36 This relation is obtained from the equality of the mixed derivatives in Eq 5 together with Eq 30 Specific heat at constant density 37 To derive the specific heat at constant pressure we start from the relation OlnT OlnT InT dl din Sho ap 7s as 38 from which we get Os OlnT OlnT 39 am onda s p d Using Eqs 28 and we obtain 40 Now 0 0 E S l dl 41 i CR ap mp PD ji a a al 01 E s np np ds dl dla T d 42 s 2155 P amp amr T C 5 42 hence Os Olnp Os Os Olnp 1 dl dla T 4 as sina Os R t at mp ser P 5 and finally ECM 45 or Using Eqs we finally obtain the relation for the specific heat at constant pressure Alternatively c can be obtained from Eq 16 10 2 EQUATIONS once and x7 are known see below or from We can now express the thermodynamic coefficients provided by COSBOLD in terms of cy Ti T3 and Vag We consider again Ea 23 replacing de by de de de dT 4 d 55 ar 95 Ji 55 B 1 0 1 0 e e de 7 57 7 Ur E Tp he v9 The requirement that the mixed derivatives must be equal then yields 1 1 a x or ar 5 55 7 en Op VT NOT OT NT N Op TR OT 1 de 1 1 Op p 7m 7 x Gr amp j n Finally Comparison with Eq 53 i
165. y you define at least one of the three switches rhd_r01 rhd r02 or rhd r03 to choose a radiation transport module The others have reasonable default values To find the combination with the optimal performance you should look into Sect The macros are sorted into different categories Some activate a certain feature like a radiation transport module or the dust module They have to be selected by the user typically via environment variables and the configure script see Sect 3 6 each time the code is compiled for a certain purpose 26 3 PROGRAM FILES INSTALLATION COMPILATION Other macros are meant to improve the performance by offering the choice between e g different loop structures or case distinctions These macros are set by the configure script to the best knowledge of the author s Ideally they should be checked and modified if necessary each time CO5BOLD is compiled on a new machine It should be save to modify these settings the results between runs with different settings should only differ slightly due to round off errors Some macros select between different numerical approximations A change here should be visible in a more or less drastic change of the results of a simulation Usually the default values should be accepted Other settings typically only exist to allow the comparison with older versions of COSBOLD or because there are new developments going on which have not yet managed to become the default A couple

CO5BOLD User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents