Home

elsAxdt User Manual Python/CGNS interface for elsA

1. CGNS name or CGNS like name elsA name Notes Coordinatex x CoordinateY y CoordinateZ z Density ro MomentumX rou MomentumY rov MomentumZ row MomentumxX rovx MomentumY rovy MomentumZ rovz EnergyStagnationDensity roe TurbulentSANuTilde k 1 TurbulentSANuTildeDensity rok 1 3 TurbulentEnergyKinetic k 2 TurbulentDissipation eps 2 TurbulentDissipationRate eps 2 3 TurbulentEnergyKineticDensity rok 2 3 TurbulentDissipationDensity roeps 2 3 RSDDensityRMS residual ro 3 RSDMomentumXRMS residual rou 3 RSDMomentumYRMS residual_rov 3 RSDMomentumZRMS residual_row 3 RSDEnergyStagnationDensityRMS residual_roe 3 Velocityx u VelocityY v VelocityZ w ViscosityMolecular visclam ViscosityEddy viscturb Viscosity EddyMolecularRatio viscrapp SkinFrictionx frictionvectorx SkinFrictionY frictionvectory SkinFrictionz frictionvectorz SkinFrictionMagnitude frictionmodulus TurbulentDistance walldistance TurbulentDistanceIndex wallglobalindex 3 Mach m Pressure p PressureStagnation stag_pressure EnthalpyStagnation stag_enthalpy NormalHeatFlux normalheatflux Temperature tsta TemperatureStagnation tgen 44 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA a ee THE FRENCH AEROSPACE LAB Reference Tables CGNS name or CGNS like name elsA name Notes MomentumXFlux flux rou 3 MomentumYFlux flux_rov 3 Mo
2. We suggest you to use these constants then your code would have less dependency to a change to the module implementation We detail here these constants 3 6 Enumerates and constants Dictionaries The values listed here are defined as Python We encourage you to use these constants instead of actual string or integer values Moreover we strongly suggest you always prefix the constants with the module name 3 6 1 Adapter The Adapter enumerates are parameters to the Xdt class translator That is when you create an Xdt object using the base class you have to specify what kind of adapter you want M PYTHON Xdt Python translator read writes Python objects in memory O cens Xdt CGNS translator read writes CGNS ADF or CGNS HDF5 files O USER XML MEMORY Reserved for future usage Actually the CGNS and PYTHON adapters are the most used and you would prefer to create Xdt objects using the specialized classes and not the base one 3 6 2 ErrorCode The error enumerate gives you the level of the returned error M OK Nothing bad happens O WARNING Something bad happens but this is not that important we just inform you on this and you continue to work O ERROR Something wrong happens but you can decide to continue to work O FATAL Something wrong happens no way for elsAxdt to continue to work you have to stop 3 6 3 ParseMode The parse mode tells the Xdt object what kind of operation to perform after reading the CGNS tree M C
3. ZoneBC ZoneBC_t O Symmetry South BC_t PointRange IndexRange_t Wall Front BC_t PointRange IndexRange_t Solver BC UserDefinedData_t wall temp DataArray_t E InflowSubsonic West BC_t Fig 18 BC with CGNS type The user defined BC indicates an elsa BC with or without data described in the next section A In the case of a PointList the min and the max of the point indices will be used simulating a structured window The PointList is only allowed in order to make life easier for connections with unstructured solvers A Some CGNS BCs without arguments are translated in elsa BCs with solver specific arguments For example wallslip can have the two extrap and prescor arguments In that case the user can either 1 supress these arguments from his CGNS tree and the solver will use its own default values or 2 use a Solver BC node as described hereafter In the wallslip example one would find two extra nodes Solver BC extrap and Solver BC prescor which are integer DataArrays with a dimension of 1 If the BC type does not exist in CGNS or if it does not fit exactly with the elsa BC you should use a complete Solver BC node In that case the BC type is UserDefined and the type is one of the arguments given in the BC pattern 4 6 1 b Boundary with global data In that case a list of values can be associated with the CGNS BC This data so called DataSet is defined without any knowledge of the applicati
4. cans ONERA Se Package THE FRENCH AEROSPACE LAB mpirun np 1 elsA x c import elsAxdt print elsAxdt ranksize 6 2 4 Post processing common pitfalls The elsAxdt module does not save the CGNS result by itself You have to explicitely ask for a disk save The CGNS result tree is in memory and is lost at end of computation The input CGNS tree is not copied into the result tree For example you won t find the grid coordinates in the result tree unless you explicitely ask elsa to produce such an output 6 2 5 Current version fingerprint The elsAxdt version number appears in the standard output at startup You can also find this version number in the package sources elsAxdt version py major and minor version numbers The test report has been generated for the elsAxdt module the elsAxdt test directory contains a test suite that produces atest report This report is stored as a file and is released with thedistribution tarball You re generate the test report by running yourself the test suite on your platform Some of the test are taken from our customers usual application without proprietary data of course You can refer to the remarks in section find out the specific application pattern you are looking for 6 3 Known issues and common troubleshootings The list below gives the issues or limitation known at release time 6 3 1 a CGNS tree not taken into account As the elsAxdt parser gain features it can r
5. PointRange IndexRange_t FamilyName FamilyName_t Fig 21 BC defined by a family We can see the family gathers a set of BCs each BC declares its application range in its zone and is associated to a family The family itself is found in the base level with the parameters to set for all the BCs that belongs to the family Er WALL 01 Family _t FamilyBC FamilyBC_t Solver BC UserDefinedData_t wall temp DataArray_t JM SYM 01 Family_t FamilyBC FamilyBC_t J INJ 01 Family_t FamilyBC FamilyBC_t o gt Solver BC UserDefinedData_t PressureStagnation DataArray_t EnthalpyStagnation DataArray_t tv DataArray_t sty DataArray_t stzv DataArray_t OUT 01 Family _t OUTPUT WALL 03 Family _t Fig 22 Family BC and proprietary parameters 4 6 2 Partial fields The partial fields are stored in BC structures So far the applications we have with partial fields are related to surfaces and can be understood as BC values In the example below the results have been defined by the skeletton found in FlowSolution _t Solver SolutionSetInBC as described in a section before A CGNS BC with data set is created with the compliant sub tree to the actual value The BC data set is in its owner zone and has a PointRange Or PointList and a GridLocation defining the location of the values in its zone Base Zone ZoneBC Wall 1 IcingWallData Gaz NeumannData NormalHeatFlux Base Zone ZoneBC Wa
6. O Path a list of strings giving the name of the output value w r t the BC data sets structures O Protocol in the case of code coupling defines the states of the automata where the input output should occur It is a list of strings each string is a state keyword fbefore begin iteration subiteration end after A elsa only reads structured zones however all ranges can be defined using PointRange of PointList structures For example it makes possible the use of both structured and unstructured codes at the boundaries if you define you point ranges using PointList instead of PointRange t structure FlowSolution EndOfRun a Solver SolutionSubSetinBC ext Path YPointList Protocol FlowSolution_t UserDefinedData_t UserDefinedData_t DataArray_t DataArray_t DataArray_t 22 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA Js a p THE FRENCH AEROSPACE LAB Profile Fig 14 Surface output settings For example in ext1 the Path could be IcingWallData Gaz NeumannData NormalHeatFlux IcingWallData Gaz NeumannData SkinFrictionyYy IcingWallData Gaz NeumannData SkinFrictionZ and the Protocol could be after end of computation Such a definition means that we want the NormalHeatFlux and the SkinFrictiony and SkinFrictionz on the specified PointList of the zone at the end of the computation The Path strings should be separeted by newlines n all trailing blanks are removed by
7. Zone 002 Zone 003 Zone 004 Zone 005 Zone 006 Zone 007 Zone 008 Data SquaredNozzle 06 R s ext 0 0 0 0 Fri 1 1 1 You read here the Zlist list of zones per processor is used by the elsAxdt the selection attribute is set but not the distribution attribute 5 5 Remarks about data exchange 5 5 4 Shared arrays The CGNS trees used by elsAxdt are Python trees These are lists of lists with Python objects such as string arrays lists and these have the same life cycle as any other Python object The exception are some CGNS tree values used for code couping or run time exchange we detail this exception later in this section A Python object exists as far as it has a reference on itself This means that something else is using this object A variable name is a reference to an object Thus if you have a variable pointing to a CGNS tree this one would not be destroyed by Python If you want Python to destroy the tree and get back the memory used by this tree you have to explicitely delete the variable i e use the del Python function to destroy the variable The Python CGNS node values are arrays These arrays are pointing to actual memory allocated by elsAxdt or elsA depending on this is an input or an output CGNS tree In the case you delete the tree the arrays are deleted because the memory ownership is set to the CGNS tree not elsAxdt ot elsA There is one exception that is for run time CGNS t
8. The FlowEquationSet_t node also contains solver independant parameters All other nodes found as the CGNSBase t children are ignored 4 2 3 Zone structure SIDS 6 3 The physical discretization of space is separated in blocs or zones The zones are children of the Base t node and they have the Zone t type elsa only reads structured zones The Zone t nodes can have any user defined name Below the zone node we can find all grid related information boundary conditions solutions and connectivities Connectivities are allowed only between zones of the same base This means that the name of a zone refered to in a connectivity structure should be found in the same base This also means that if you do not have enough connectivity information to give to the solver i e for a Chimera computation you should not introduce a grid connectivity structure In the case no zone is found in the input tree the error 7101 is raised Dy B SquaredNozzle CGNSBase_t Er Zone 001 Zone _t ZoneType ZoneType_t GridCoordinates GridCoordinates_t ZoneGridConnectivity ZoneGridConnectivity_t 9 ZoneBC ZoneBC t E FlowSolution init FlowSolution_t FlowSolution EndOfRun FlowSolution_t EH 4Zone 007 one t H Zone 008 Zone_t Fig 2 Zone example The children of the Zone _t node are the grid itself GridCoordinates the connectivities between this zone and other zones ZoneGridConnectivity t the boundary conditions ZoneBc_t the
9. and errors The solver actually reads the CGNS tree without performing any control The user has to apply checking and CGNS compliance tools before the solver submission Thus most solver errors related to CGNS or XDT CGNS are forced as FATAL errors The table hereafter describes the possible troubleshootings you may have 7100 The XDT CGNS tree you gave as argument is empty This error can be raised if 1 there is no CGNSBase t node as father of all other nodes or 2 you did not ask to read the mesh 7101 No Zone t node has been found in the CGNSBase t node of your tree Check your tree the solver reads only the first CGNSBase _t and you may have a first empty base before the actual base 7120 One of the boundaries you gave is not known The solver cannot find a translation of the given BCType t into one of its own boundary condition types Check the BCTypes the error message gives you the path to the bad BC 7121 While reading a BC the required fields where not found Or while reading a BC one of the required field found was incorrect ELSA MU 02052 v2 0 elsAxdt 43 49 Reference Tables THE FRENCH Vys 7 2 Variable name mapping The mapping tables are printed when you use the trace option while reading any CGNS file The list here after details the most used variables
10. array module When elsAxdt is imported a message is sent to the standard output This message recalls you the Python array module which has been used for elsAxdt installation You must insure this array module is the same as the one used by elsA or any other Python module you plan to import in your Python scripts 1 4 About this document The elsAxdt manual is at the same time a reference document and a users guide As we are supporting the CGNS standard many informations related to the standard are not detailled here The user should read the SIDS to have more complete information about the CGNS data structures The chapters of this document are gt Quick Start to be able to run elsAxdt on examples in five minutes gt Interface describes the objects and functions available through elsAxdt in order to drive elsA and to use CGNS Profile is the reference for CGNS trees elsAxdt can understand or produce Package is a technical chapter describing the module architecture and some installation remarks The features examples and interface detailled in this document are based on at least these versions elsAxdt v5 30 elsA v3 3 If you have an release with a greater version number compatibility is insured 4 49 elsAxdt ELSA MU 02052 v2 0 ND WNE Teens ONERA RE ee Quick Start THE FRENCH AEROSPACE LAB 2 QUICK START You want to use elsAxdt now Follow the small tutorial hereafter we assume you already
11. artviscosity DataArray_t avcoef_k2 DataArray_t savcoef_k4 DataArray_t savcoef_sigma DataArray_t efl DataArray_t convergence_level DataArray_t ode DataArray_t implicit DataArray_t time_algo DataArray_t niter DataArray_t inititer DataArray_t freqcompres freqcomptimestep multigrid Fig 24 Proprietary solver parameters DataArray_t DataArray_t DataArray_t Again the reader would refer to the tables to find the list and the meaning of every parameter 4 8 Chimera features The Chimera requires the definition of the overlapping boundary conditions and the definition of the neighbourhood for a set of zones A body is automatically detected in the case of a family BC defining a contiguous set of surfaces This family is used as the tag for the body identification A mask is created for this body The zones without a mask have an automatic generation of ghost mask A The CGNS standard defines overlapping connectivity in our case we are generating the overlapping data thus the standard structure is unused A The read write of overlapping information and Chimera coeffs will be availble in the next elsAxdt release The definition of the overlapping and the masks are using the families The mask are Zone level families and the overlapping are BC level families These should be different families it is not possible to have the same family name for Zone and BC levels at the same time We see the
12. continues If there is a requested stop the Python script specified as argument of the P command line argument is run The script is run in the same name space as the solver and the main script argument Thus you can share variables in these two scripts the three are the same process Some other states are specific to dedicated coupling computations such as the multi stages turbomachinery computations Two more states are used O pemwrite after the solver writes data in the xdt runtime tree O pemread after the solver reads data in the xdt runtime tree In the case of the multi stage computation the user does not have to request a stop at these states The solver forces a stop at the pcmwrite state when it detects the use of a multi stage boundary condition The states actually are set using a bit string each state is a bit in this string and the complete string gives the stop behavior of the solver before Ox0001 after 0x0002 start 0x0004 end 0x0008 iteration 0x0010 pcmwrite 0x0040 pcmread 0x0080 5 3 Coupling script The code coupling script is the piece of Python code executed when the solver reaches a coupling state A coupling state is declared by the user as a state where the solver has to stop see command line arguments in the previous section The coupling script can either be a separate script or a function Such a function so called Callback is declared and set into a special Python di
13. data with respect to the computation equation model The table below details the possible patterns Density Always required Momentumx Always required Momentumy Always required Momentumz Always required EnergyStagnationDensity Always required TurbulentEnergyKineticDensity TurbulentDissipationDensity TurbulentSANuTilde TurbulentDistance TurbulentDistanceIndex 4 7 2 The flow equation set SIDS 12 1 The FlowEquationSet tree is ignored 4 7 3 Solver computation parameters We also added a proprietary sub tree required for the representation of all computation parameters used by elsa and not taken into account by the CGNS standard These are solver parameters and CGNS could not specify these parameters as common for all CFD solvers However CGNS allows the addition of specific sub trees such as our Solver Compute sub tree The Solver Compute node contains all numerical control of the elsa solver the user usually cannot avoid to have this sub tree 28 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA THE FRENCH AEROSPACE LAB Profile wif E SquaredNozzle CGNSBase_t EH Zone 001 Zone _t Er Zone 007 Zone _t 1 Zone 008 Zone _t Simulation Type SimulationType_t E FlowEquationSet FlowEquationSet_t Er Solver Compute UserDefinedData_t fluid DataArray_t phymod DataArray_t gamma DataArray_t freezing DataArray_t flux DataArray_t
14. file We want now to read a CGNS file for our computation This file is the input tree We call that a tree because it is organized with nodes having children each node is compliant with the CGNS data model and the whole tree itself is compliant We create an xdt tree in memory using the CGNS file given as argument The variable mytree is the xdt representation of the 001Disk cgns tree The mytree object can be used to query some practical informations about the loaded tree import elsAxdt mytree elsAxdt XdtCGNS 001Disk cgns print mytree filename The CGNS file can contain links to other CGNS files it is up to you to insure that all required files are available in the correct directory During the read the links will be followed the whole data is gathered into a single in memory representation of all the computation data you have put in the CGNS file A We strongly suggest you use only file names without any relative or absolute path reference in your link paths A You can specify a directory search path for the linked to files read the section 6 3 The actual read is not performed in line 2 we see in next section how to perform the actual read and to run the computation ELSA MU 02052 v2 0 elsAxdt 5 49 SJ A U1 amp amp NN H ONERA Z NEKA NS Quick Start nanas S S THE FRENCH AEROSPACE LAB 2 3 Running elsA The e sA solver requires a complete simulation description You can define all the simulation dat
15. flow solutions FlowSolution t In the case of a parallel computation the zones taken into account by the current processor should have a complete structure The zones which are not located on the current processor may only contain the connectivity information Dr B Plane 846 v4 CGNSBase t E Zonel Zone _t ZoneType ZoneType t 3 ZoneGridConnectivity ZoneGridConnectivity_t 3 GridCoordinates GridCoordinates_t 3 FlowSolution init FlowSolution_t 3 FlowSolution EndOfRun FlowSolution_t E Zone2 Zone t ZoneType ZoneType_t ZoneGridConnectivity ZoneGridConnectivity_t Zones Zone_t A Zone4 Zone_t Fig 3 Complete and incomplete zones parallel computation In the figure Fig 3 you can see the Zonel is complete the Zone2 is not it only has the connectivity information Such a tree can be given to the processor having in charge only the Zonel assuming another processor would have Zone2 to compute 16 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA a ee THE FRENCH AEROSPACE LAB Profile 4 3 CGNS tree grids and solutions The meshes and solutions are children of their associated zone The solution one can found in a zone contains the full field of data related to the mesh We detail here how to drive elsA for the output tree specification 4 31 Grids SIDS 7 1 The CGNS mesh is defined by the sub tree GridCoordinates containing the values for Coordinatex Coord
16. haf or cgns file However the file format you use should be the one the library you used during elsAxdt production An XdtCGNS adapter has a specialized attribute filename which contains the name of the CGNS file used O xdtcGNs creates a new object argument is the filename 3 4 Migration methods The following methods can be applied to an Xdt instance The methods with the migration tag are only useful if you want to mix a legacy elsa Python script and an Xdt tree parse O elsA2CGNS elsaname returns the CGNS name string as a path to the actual CGNS node used to create the elsA object with the name given as argument string O CGNS2elsA cgnsname returns the elsA name string in the case this CGNS node given as argument string was used to create an elsA object Otherwise raises an error O symboltable returns all the elsA names created from a CGNS node O e forceValue name attribute value set the value integer float string to the internal attribute string of the object with internal name string As the operation is performed on the actual elsA objects the method should be used with care No control of any kind is made on the attribute name nor the value O e getValue name attribute get the value integer float string from the internal attribute string of the object with internal name string The attribute name should be defined on the target object unless the operation fails during run time no control is made o
17. in memory 954 Asynchroneous data exchange case 952 On the fly storage of convergence data 953 External use of pyMPI module 954 BC input control with external script 955 BC actuator disk 956 BC periodic 957 BC input with user defined data on BC level family 958 BC mobile coef 959 BC Froude 960 Different ways to define grid location output using user defined data 961 Different ways to write grid location output using GridLocation 962 Different ways to read grid location output using GridLocation 963 Read CGNS transition information 964 Write CGNS transition information 065 Write CGNS Chimera information 970 Adjoint definitions as input 974 Adjoint definition for output 872 Unstructured mesh and connectivity read 073 Unstructured solution write ELSA MU 02052 v2 0 elsAxdt 47 49 ONERA Reference Tables ee OO a THE FRENCH AEROSPACE LAB Vys Forewords 1 1 CGNS Standard 1 2 Storage vs data model 1 3 Python arrays implementation 1 4 About this document Quick Start 2 1 The module import 2 2 Reading a CGNS file 2 3 Running elsA 2 4 CGNS trees layout Module User Interface 3 1 Python package interface 3 2 The Xdt class 3 3 Adapters methods 3 4 Migration methods 3 5 Module constants 3 6 Enumerates and constants Dictionaries Profile 4 1 Compliance and profile 4 2 Mai
18. in the solution skeletton the convergence for this variable is automatically added in the output tree The convergence would be produced at the CGNSBase t level GlobalConvergenceHistory t Dy B CASE11 CGNSBase_t El GlobalConvergenceHistory ConvergenceHistory_t IterationNumber DataArray_t RSDEnergyStagnationDensityRMS DataArray_t RSDDensityRMS DataArray_t RSDMomentumZRMS DataArray_t RSDMomentum YRMS DataArray_t RSDMomentumxRMS DataArray_t Zone Zone_t ao Zones Zone_t Fig 10 Global convergence for all Zones A For parallel configurations only the output tree on the processor 0 zero would have the convergence node Again it is up to the user application to gather parallel solution into a single tree Usually the post processing of a CGNS tree in parallel performs the merge of the per processor sub trees and adds links to the input tree grid coordinates With such a post processing a third party vizualisation tool can blindly read your CGNS solution A The window based residual should be associated with a Family 20 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA ee THE FRENCH AEROSPACE LAB Profile 4 4 3 Family based output The user can request convergence on other variables using a specific solver declaration with the families A Family_t node has to have a Solver Output node with the set of elsa parameters for such an ouput The example in figure 4 12 shows a request for the family named w
19. increases the experts you can find the documentation tutorials examples tools Many data formats are restricted to plain float arrays or even discrete data Thus the user has to be aware of the meaning and the structure of the data before any use of the actual values The CGNS standard is a data specification or data model it comes with a detailled and well defined meaning for every part of the data structure Such a standard could be seen as a restriction to proprietary formats because many of these formats are exactly defined for the targeted applications The standard is more a consensus the largest common set of agreements between people making solvers making tools or using them Then while not perfect the CGNS standard is the ground for elsA data model interface 1 1 2 CGNS version The reference document for CGNS is the so called SIDS which stands for Standard Interface Data Structure The reference version is v2 3 4 which is described in the document AIAA R101A 2006 publicly available on http www cgns org official CGNS web site ELSA MU 02052 v2 0 elsAxdt 3 49 ONERA pene el ei NS Forewords a ee yv THE FRENCH AEROSPACE LAB 1 2 Storage vs data model A data model is a logical view of data Such a view can be mapped to a physical view For instance saying that a 3D point with cartesian coordinates have three coordinates X Y and Z is a logical or conceptual view of a point An array with three 64 bi
20. profile ZoneBC_t SIDS 9 2 structure only supports BC_t children 4 6 1 Boundary conditions SIDS 9 3 The elsa profile BC_ structure only supports BCType_t PointRange and PointList and BCDataSet_t children The location of values is ignored at this level see BCDataSet_t 24 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA a O U THE FRENCH AEROSPACE LAB Profile 4 6 1 a Boundary without data The most simple BC definition is a simple type without data The Bc _t contains a value a string which contains the CGNS BC type name The actual name of the BC is left to the user For example you can find a BC with the name Wall Up and with the value BCWallInviscid The PointRange is the application patch or window for this BC A PointList can replace the PointRange as far as the list defines a structured patch The CGNS BC types without values and taken into account by elsa are listed hereafter All other CGNS BC raises the error 7120 In the following table the legacy elsa type is given for migration purpose BCFarfield nref BCWallInviscid wallslip Normal velocity is ignored BCWallViscous walladia BCWall BCOut flowSupersonic outsup BCSymmetryPlane sym BCDegenerateLine inactive BCDegeneratePoint inactive UserDefined any Can be used to defined any elsA BC mandatory for overlap BC definition BCOverlap overlap The use of this value is not recommanded see Chimera section
21. should be of enumerate LoadMode Other parameters are usual Python types such as string integers etc O xdtParser creates a new object arguments are the adapter type and the parameters The parameters type change depending on the adapter type for exemple a CGNS adapter will require the filename where the Python adapter waits for a Python CGNS tree i e a list of nodes O load flags actually creates the Xdt tree taking into account the set of flags passed as arguments LoadMode with default value READ_ALL The return value is the Xdt object itself O compute starts the elsA computation The return value is a new Xdt object with the resulting computation tree O save filename saves the output Xdt object in a CGNS file All the tree is stored in a single file If the filename already exists the adapter automatically adds a trailing to the filename and tries again to save until it succeed Thus if you run several times your script you will have many files and the newest one has the highest number of O hollow returns a Python CGNS tree representation of the current Xdt object O dump a pl p2 creates a new Xdt object with adapter a Adapter and parameters pl and p2 CGNS file with name filename string O add filename reads a CGNSTree and adds the CGNSBase to the already parsed file s In addition with these methods the adapters have specific methods for the migration purpose See next section for these specific me
22. space characters SlatOverlap Family_t MT FamilyBC FamilyBC_t cl BCoverlap J Solver Overlap UserDefinedData_t MT NeighboutrList DataArray_t c1 fluid WingOverlap fluid Background Fig 27 Typed BC overlap definition FamilyBC FamilyBC_t cl UserDefined B Solver Overlap UserDefinedData_t MT NeighbourList DataArray_t cl fluid WingOverlap fluid Background Fig 28 UserDefined BC overlap definition A Note the the BC FamilyName does NOT contain the CGNSBase_t name The specific parameters you want to set for the mask have to be set at the family level with the Solver Mask node The elsA attributes are used as is the NeighbourList has the same pattern as the previously described node FUSELAGE Family_t MT YFamilyBC FamilyBC_t ci BCWallViscous a Solver Mask UserDefinedData_t MT type DataArray_t ci cart_elts proj_direction DataArray_t y dim1 DataArray_t I4 500 dim2 DataArray_t a6 500 NeighbourList DataArray_t c1 GOAHEAD 003 MainRotor MRB GOAHEAD 003 TailRotor TRB Fig 29 Mask parameters 30 49 elsAxdt JELSA MU 02052 v2 0 cans ONERA THE FRENCH AEROSPACE LAB Profile 4 9 Zone parameters The are a lot of elsA parameters the user can set at the Zone level Most of the time the parameters are set into a zone level family node The parameters are applied to all zones matching the family The Solver motion uses the elsA attributes for the motion it appl
23. the tree is explained in the next chapter we assume here you have a correct tree available You have several ways to create and modify an Xdt tree depending on your application you may have a complete tree in a CGNS file or you may use full Python trees for in memory code coupling An Xdt object is created with different arguments depending on the adapter An adapter is a translator which helps you to create or copy trees Five adapters are available but two of them are really useful with the current solver version O PYTHON uses the Python CGNS tree representation O cens uses the CGNS HDF or ADF tree on disk Depending on the adapter the arguments are different The most simple adapter is the CGNS one which reads a CGNS file mytree XdtCGNS 010Disk cgns The file is not read the mytree object is created but it waits for more options before reading the file mytree XdtCGNS 010Disk cgns mytree load READ ALL Now the file is read the complete file as specified by the READ ALL option You can perform the same command in a single line mytree XdtCGNS 010Disk cgns load READ ALL At this time you have in memory a C Xdt tree with the structure and the data of your CGNS file You can also ask to a Xdt tree to save itself into a CGNS file mytree duplicate 010Disk cgns The Xdt tree knows it has to save as a CGNS file because it was created wich a CGNS adapter in other words mytre
24. they have the Solver prefix 41 3 About screenshots The CGNS trees screenshots have been produced by CGNS NAV which can read HDF5 ADF binary files or Python in memory CGNS trees a pyS7 GOAHEAD_mrb View 01 O amp 3 ANT A BS 5 a a wa ln di at it CGNSLibraryVersion CGNSLibraryVersion_t 3 E GOAHEAD 003 MainRotor CGNSBase_t 3 3 E t block 1 Zone_t 1113 205 33 112 204 32 0 0 0 E 1 block 2 Zone _t A 97 21 13 96 20 12 0 0 0 E wt block 3 Zone_t A 29 61 13 28 60 12 0 0 0 ZoneType ZoneType_t ASstructured ZoneBC ee ZoneBC_t GridCoordinates amp GridCoordinates_t Er ZoneGridConnectivity ZoneGridConnectivity_t MT Ensegs3 3 GridConnectivityito1_t C1 Al block 2 Transform int indexDimension 14 1 2 3 PointRange IndexRange_t mM 1 11 1 29 1 13 PointRangeDonor IndexRange_t 4 69 21 1 97 21 13 seg 3 4 GridConnectivityito1_t 1 Al block 4 seg 3 5 GridConnectivityito1_t C1 At block 1 seg 3 6 GridConnectivitylto1_t C1 Al block 1 FlowSolution EndOfRun FlowSolution_t ur ee Flow pyS7 Control panel v0 9 BEE 1 block E gt T O 8 B 2 block S 7 View File Dir 5 Vode Fa 2 block 1 GOAHEAD mrb stck4 poinot tmp GOAHEAD 2 data 7 GOAHEAD str stck4 poinot tmp GOAHEAD 2 data 2 block GOAHEAD hub stck4 poinot tmp GOAHEAD 2 data 2 block GOAHEAD_mrb stck4 poinot tmp GOAHEAD 2 data GOAHEAD 0
25. this is a command line argument for a code coupling computation with a legacy script elsA X 10 C xdt runtime tree P A06ax py t py The same computation with a full CGNS computation would be run with the following command line Python is now the interpretor instead of elsA python A06a py X 10 C xdt runtime tree P A06ax py And the A06a py script itself should contain import elsAxdt as E import sys mode code here E args sys argv 32 49 elsAxdt ELSA MU 02052 v2 0 HR H 0 OO J O 0 1 cans ONERA aa a Interoperability Features THE FRENCH AEROSPACE LAB mode code here mytree E XdtCGNS FluidPlate cgns load E READ ALL mytree compute mode code here The important line in this script is the args method call with the command line arguments passed to the solver as it starts its computation These arguments are required by the sopver but it doesn t care how you give it For example you can give your arguments as hard coded string values import elsAxdt as E import sys E args X 10 C xdt runtime tree P A06ax py mytree E XdtCGNS sys argv 1 load E READ ALL mytree compute Here the filename is passed as arg while the other values are built in a simulated command line arguments list 5 2 Code coupling user interface 5 24 Command line arguments The solver has code coupling dedicated command line optio
26. 03 MainRotor 1 block 4 o3 bl gt pyS7 GOAHEAD_mrb View 03 a x 8 bl ET 8 bl 3 bl 4 bl J 4 bl 4 bl 4 bl MR F4 E Solver Mask UserDefinedData_t MT AN type DataArray_t xray dim1 DataArray_t 500 dim2 DataArray_t 500 blanking_criteria DataArray_t cell_intersect NeighbourList DataArray_t GOAHEAD 003 Fuselage ENGINE GOAHEAD 003 Fuselage HS GOAI OVERLAP Family _t E LE fh PHHHH DH H NAHEAN mrh zatek nnina mn GOAHEAN 2 fdata QAHEAN 003 MainBoto hlack ates Conrd LE 3 1 o 1 2 3 4 5 E 1 8 9 6133e 00 6133e 00 6132e 00 6132e 00 6131e 00 6130e 00 6128e 00 6127e 00 6126e 00 6126e 00 6133e 00 6133e 00 6132e 00 6132e 00 6131e 00 6130e 00 6128e 00 6127e 00 6126e 00 6126e 00 6132e 00 6132e 00 6132e 00 6132e 00 1 6132e 00 1 1 1 1 6132e 00 1 1 1 1 1 6132e 00 6132e 00 6132e 00 6131e 00 6130e 00 6129e 00 6128e 00 6127e 00 6126e 00 6132e 00 6132e 00 6132e 00 6132e 00 6131e 00 6130e 00 6129e 00 6128e 00 6126e 00 6126e 00 6132e 00 1 6132e 00 6132e 00 1 6132e 00 6132e 00 1 6132e 00 6132e 00 6131e 00 1 6131 6130e 00 6129e 00 1 6129e 00 6128e 00 1 6128e 00 6126e 00 1 6126e 00 6126e 00 1 6126e 00 6133e 00 6132e 00 6132e 00 6132e 00 6131e 00 6130e 00 6129e 00 6127e 00 6126e 00 6126e 00 6133e 00 6133e 00 6132e 00 6132e 00 6131e 00 6130e 00
27. 6129e 00 6127e 00 6126e 00 6126e 00 6133e 00 6133e 00 6132e 00 6132e 00 6131e 00 6130e 00 6129e 00 6127e 00 6126e 00 6126e 00 pp 6130e 00 6129e 00 6128e 00 6127e 00 6126e 00 m O co a on Nn e N M TE OAN Pere eae ee RRRREDRERER ol LEE RE LL a ER Pee Tae ee Per ae GOAHEAD 003 MainRotor 1 block 3 GridCoordinates Coordinatex In this view the columns name are displayed Other screen shoots in this manual may no have the column names However the contents is obvious the first column is the name of the node the second textual column is the type of the node In most cases this pair name type is enough to understand the tree structure The user defined names are in bold while the other are mandatory or strongly recommanded names 14 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Profile 4 2 Main CGNS tree structures The input data is a read only tree The input includes the geometry of the computation grids connectivities the boundary conditions the initialization fields the reference state the governing equations and the computation and numerical control attributes All the parameters and data required for a computation can be described in a single logical tree structure This tree can actually be stored into several CGNS files with links which are parsed by elsAxdt The standard data is defined using the SIDSrecommandation The proprietary in
28. Array_t tv DataArray_t Fig 20 Proprietary BC data with CGNS variable names A It is always possible to add solver specific values as described in the previous section A Solver Bc node has to be defined and the user can put a list of names and values A The conservative variables should have the CGNS compliant name and should of type DataArray of float with a dimension of 1 4 6 1 c Family boundary conditions The most user friendly way to define a boundary condition is to use the family logical set Some tools can generate BC using the families A Family is a label identification of a CAD surface segment or point The CAD tool allows the definition of BCs on families thus allowing the use of set of CAD entities As the CAD generates the mesh is produces blocs connectivities distribute BCs and generates BCs application patches The families in a CGNS tree can be found as BCs labels This information is useless to the solver as the 26 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a O O THE FRENCH AEROSPACE LAB Profile CAD distributes and correctly generates BC types patches or global data However identification of families can be useful for pre and post processing which is not the purpose of the elsa solver T zZoneg9 Zone_t Zonelype ZoneType_t 4 ZoneGridConnectivity ZoneGridConnectivity_t 4 GridCoordinates GridCoordinates_t ZoneBC ZoneBC_t B BC Zone89 732517 BC t
29. Base test case 030 Complex case with an external update using Python for a NREF boundary limit 031 Protocol based output for BC values related to code coupling use 032 Output with bl_quantities boundary layer 033 Writes only the conservative values for restart 034 Reads only the conservative values for restart 035 A so called cellfict output 036 Output of ichim on BC defined surface 037 Writes only the conservative and the wall distance values for restart 038 Reads only the conservative and the wall distance values for restart 039 Writes only the wall distance values for restart 040 Reads only the wall distance values for restart 944 Parallel computation with per process file write 942 Parallel computation with per process file read each file is incomplete so that each process has not all the CGNS 46 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA Se THE FRENCH AEROSPACE LAB Reference Tables tree available 943 Parallel computation with a multi base family integration 944 Migration case reading only the mesh 945 Migration case with CGNS attributes overwrite 046 Migration case with a Python definition of computation attributes 047 Migration case with a Python definition of CFL function 948 Migration case with aeroelastic computation 949 Case with grid arrays shared in memory 950 Case with unsteady data arrays shared
30. OMPUTE Once the tree is read run the actual solver computation O TRANSLATE Load the tree but do not run the computation O UPDATE Update the tree instead of loading it These three enumerates can be set to the action attribute or you can use the methods compute translate and update instead 3 6 4 LoadMode flags These flags indicates which parts of the CGNS tree have to be taken into account This is mainly used for elsA legacy scripts migration or tuning Each flag can be or ed to build a bit string with options The default value for this parameter is READ_ALL O READ MESH Read Grids no connectivity O READ CONNECT Read connectivities O READ BC Read boundary conditions ELSA MU 02052 v2 0 elsAxdt 11 49 ONERA pene el D NS Module User Interface ET OR S Sy THE FRENCH AEROSPACE LAB O READ INIT Read initialisation fields O READ FLOW Read flow equation sets not flow solutions O READ COMPUTATION Read computation sub tree O READ OUTPUT Read output skeletton O READ TRACE Set the trace mode O0 READ NONE Empty option set O READ ALL Full options set The elsAxdt module loads a CGNS tree as a trustable and consistent set of data Thus there are nota lot of verifications and if an error occurs during the tree parse it often is a fatal error You can drive the tree parsing by saying wihch parts of the global tree you want to take into account or not This is the LoadMode set of flags Each flag can be added to the actual flag
31. a and control parameters in a single CGNS file or define only a part of these information in a CGNS file and complete the description in a Python script with e sA legacy commands This latter mode is so called migration mode The example below shows how to run a full CGNS computation import elsAxdt mytree elsAxdt xdtCGNS 001Disk cgns mytree mode elsAxdt READ ALL resulttree mytree compute mytree save 001Disk output cgns The line 4 sets a flag on mytree to indicate that all the CGNS tree has to be read The line 6 actually runs the computation the elsA solver should find in mytree i e in the CGNS tree all the required data The result of the computation is returned as a new xdttree containing all requested output such as result fields convergence The layout of a classical CGNS tree for a complete computation is detailed in the next section This result is not saved in a file unless you ask elsAxdt to do so The line 7 saves the result in a CGNS file This file doesn t contain the initial computation data such as grids reference state You have to use CGNS tools to add links from your resulting output CGNS file to the initial CGNS file in order to have grid information this is required for visualization tools for example A The target object for the save method is mytree not resulttree Because resulttree is a plain Python object and it has no specialized methods such as save This resulttree can be used in yo
32. aii 01 which is also used to gather a set of BC windows as explained later in this document in the boundary conditions section The elsa parameters as defined as is using DataArray_t node types to store float integer or string values The var is the space separated list of the elsa variable to use e g convflux_rou Please refer to the elsa reference manual for the attribute values E WALL 01 Famiyt FamilyBC FamilyBC_t BCWallViscous E Solver BC UserDefinedData_t Er Solver Output UserDefinedData_t var DataArray_t convflux_rou convflux_row diffflux_rou diffflux_row loc DataArray_t 1 fluxcoeff DataArray_t 1 0 period DataArray_t 5 E Solver Output Friction UserDefinedData_t var DataArray_t frictionvectorx frictionvectory frictionvectorz fluxcoeff DataArray_t 2 07765 heatfluxcoeff DataArray_t 9968330 0 oc DataArray_t 1 Fig 11 Output settings based on a family A lt BaseName gt GlobalConvergenceHistory as path for usual convergence outputs A The output specification node should have at least the pattern lt BaseName gt lt FamilyName gt Solver Output and it is possible to have more than one output specification The following figure shows an example of conservative variable convergence together with explicitely requested convergence In the case of a global integration of the value the result is set in GlobalConvergenceHistory E SquaredNozzle CGNSBase t Er G
33. al zones that is the zones computed on the current processor Thus each processor will produce a sub tree with the local zones it is up to the user application to gather the trees in a single one with all zones 5 4 8 Typical script start up The usual pattern in a parallel code using MPI is to identify the current processor and select the zones to compute on this processor import elsAxdt as e rank size e ranksize if rank 0 zlist Zone 001 Zone 002 Zone 003 Zone 004 else zlist Zone 005 Zone 006 Zone 007 Zone 008 In this example we know we have 8 zones and we know their names One would do something more general by reading the CGSN tree find the zones and their sizes balance the zones with respect to the processor number and then distribute the zone lists on these processors The actual distribution can be done in migration mode or not In the case of the migration a zone list is given to elsAxdt but the process number for each zone is given after the CGNS tree has been read 36 49 elsAxdt ELSA MU 02052 v2 0 Vys ONERA THE FRENCH AEROSPACE LAB Interoperability Features compute co J O U1 amp amp N H C C C C C C C C C C C C c e xdt e CGNS mode c mode e READ OUTPUT action e TRANSLATE selection zlist blocknode blocknode blocknode blocknode blocknode blocknode blocknode blocknode Zone 001
34. ant to retrieve in Python You want Xdt to dump the C data structure in Python but you do not want to duplicate memory During this dump operation elsAxdt sets the NPY_OWNDATA to zero for the large data arrays These large data arrays are of DataArray_t type in CGNS Once the NPY_OWNDATA is unactivated Python knowns it cannot release the array memory It can release the objects the lists strings all other objects created during the dump but not the memory zone of the array Because it is shared by the solver Any attempt to a del call to the Python tree you get by the dump e g get xdt output tree will release Python objects not DataArray_t memory zones and not Xdt objects 5 4 2 From Python to Xdt Second scenario you have a Python array you allocate by yourself using a Python function that feeds a numpy array The memory zone is allocated by numpy if you call del on this array it eventually deallocates the memory see Reference Counting below If you pass this array to the fluid solver you perform a load in elsAxdt e g XdtPython xdt runtime tree pytree and a complete Xdt tree is created This tree shares the memory zones of the large arrays In that case not only the DataArray_t are shared but all the numpy arrays This point means you cannot update an existing Xdt tree you only create new trees When Xdt loads the Python tree all the numpy arrays have their NPY_OWNDATA flags set to zera false Thus again the Python ob
35. ario During code coupling computation the user will read data from the solver and write data for the solver use Then it is mandatory to know how the memory is managed in these code coupling cases Unfortunatly there is no straightforward rule and one should change his policy depending on the data used during code coupling Most of the time small chunks of data are copies and the code coupling script has to release the data it reads from the solver ELSA MU 02052 v2 0 elsAxdt 35 49 U1 amp amp N H yon amp amp N H ONERA cans Interoperability Features PR ONU S THE FRENCH AEROSPACE LAB This is true for boundary based code coupling for example The elsAxdt module cannot manage automatically memory release because only your application knowns when arrays are no more useful We have the free method to free the Xdt data including the actual memory zones of the arrays you have the del Python method to decrement Python reference counters and eventually free the objects It is now your responsability to manage the memory in your Python code coupling script 5 4 6 Memory release examples In the first example below we show a coupling script with an Xdt tree that is only used for display Then it is released In that case the tree is an output tree which structure has been defined by the kind of outputs requested by the user The del call is not really useful because the tree variable has a global scope and it will be o
36. ctionnary The code snippet below shows such a declaration import elsAxdt as X a MY COMPUTATION s def myIterationCallBack if X iteration gt 1 xw X xdt E PYTHON Foo None xp xw dump E PYTHON E OUTPUT_TREE print xp X StateCallBack iteration myIterationCallBack c X XdtCGNS Input cgns c mode X READ ALL r c compute c save Output cgns The function mylterationCallBack will be run at every iteration state In that case of a function callback the user does not have to set the X command line option because elsAxdt knows it has to stop at iteration state as soon as it finds a function in its StateCallBack dictionnary 5 4 Memory ownership The goal of Xdt is to give access to the internal data of the solver The solver memory zones i e arrays are read write by means of pointers and there is no memory duplication whenever possible In other words the array you have in an Xdt tree node is the actual array used by the solver It is important to understand how the ownership of this memory zone is managed in particular when you use Python input output The Python arrays are numpy arrays We give some scenario hereafter 34 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Interoperability Features please read carefully these scenario to understand how to actually free the Xdt memory 5 41 From Xdt to Python First scenario you have a solver array you w
37. dLocation El Wall North PointRange GridLocation 5 BCDataSet NeumannData SkinFrictionZ SkinFrictionY SkinFrictionx PointRange IndexRange_t GridLocation GridLocation_t 0 FlowSolution EndOfRun FlowSolution_t Fig 13 Surface output Such a BC based partial solution storage is used for code coupling purpose The next section describes how to define such output IndexRange_t GridLocation_t BCDataSet_t BCData_t DataArray_t DataArray_t DataArray_t IndexRange_t GridLocation_t BC_t IndexRange_t GridLocation_t BCDataSet_t BCData_t DataArray_t DataArray_t DataArray_t 4 4 3 a The partial field exchange can be used in order to retrieve a partial field such as a surface but it is also designed for code coupling data exchange One should take care of the input tree declaration which defines a structure in the solution sub tree while the result itself is found in the ZoneBC_t sub tree Partial solutions stored as BC data A We are waiting for CGNS extensions so called Regions that would allow to store the partial fields in a flow solution still being compliant with CGNS SIDS The definition of the requested ouput is a private node Solver SolutionSubSetInBc under the FlowSolution EndOfRun This private node contains a list of UserDefinedData_t nodes each describing an expected output The output is described by O PointList or PointRange the indice of the partial field in the current zone
38. data you will give as input to the solver are compliant and all data you get from the solver are compliant too This doesn t mean that elsa can read and write all CGNS data An obvious example is unstructured data The elsa solver doesn t read the unstructured grids even if these grids are described in SIDS the reference document for CGNS data model 41 1 Trees and nodes A tree is a node with children nodes The first node is said to be the root and the nodes at the first level below the root are the children nodes The CGNS logical structure is a tree where nodes have a name a type and may have a value For example we want to represent the X coordinates for a mesh we use a node with the name Coordinatex its type is DataArray_t and it contains an array of float values representing these X coordinates The node can be identified by its full name i e the path Base Zone GridCoordinates Coordinatex which also indicates that the coordinates are related to a given zone in a given base Most CGNS node names are the names of the value itself instead of the usual CFD variable For example SpecificHeatRatio is used in the standard instead of the usual Gamma 4 1 2 The solver profile Then we have to define the restriction of elsa to SIDS In other words what is the list of CGNS structure that can be understood by elsa The mandatories structures should be there but there are many optional structures and fixed choices we have to describe All th
39. e but cannot be used with the current elsAxdt version These will be useful for the next release Some test cases look the same or are using the same features We do not remove these similar tests you just have to know that in some cases you may have more than one way to do the same thing A test case has its input file in Tutorial Data XxxxDisk cgns with xxx your test case number The Python script to run is Tutorial Script XXX lt script title gt py and the output is set in Tutorial Output XXX lt Script title gt cgns or similar ELSA MU 02052 v2 0 elsAxdt 45 49 ONERA Paves i Bist a A NS Reference Tables ee Sy THE FRENCH AEROSPACE LAB The Tutorial Reference contains the expected output for all cases you can pass all tutorial cases as elsAxdt test suite and compare with reference The Tutorial Script coverage py script performs such a coverage Tutorial files 001 Simple computation 002 Asimple output definition reused as linked to internal FlowSolution EndOfRun nodes 003 An family based output definition made using the Family _t nodes 004 A surface output definition distributed on Bc_t nodes 005 An output defined with FlowSolution EndOfRun and producing a writing solution file for restart 006 A file with a CGNS symbolic link to output 005WriteInitSolution cgns as FlowSoluti
40. e is an Xdt tree with CGNS input ouput capabilities A Please do not mismatch save and duplicate functions For backward compatibility concerns the save function is reserved for the computation result i e the output tree while duplicate actually saves the current tree In some complex computation cases for exemple in Chimera you may have to parse different trees each having a part of the simulation grids The add function purpose is to read a CGNS tree and to add it to the existing CGNS trees In the next example gfiles is a list of files the first file is used to create the Xdt object then the other files are parsed all the CGNSBase_t node are taken into account mytree xdt XdtCGNS gfiles 0 for gfile in gfiles 1 mytree add gfile Another useful adapter is the Python adapter The idea is the same as the previous CGNS adapter in that case the Xdt tree knows how to do input output with Python objects Then you have to represent a CGNS tree using Python see 5 5 4 scopetree CGNSLibraryVersion 2 4 CGNSLibraryVersion t SquaredNozzle 09 array 3 3 CGNSBase t mytree XdtPython scopetree The CGNS tree we use in this example only as a CGNSBase_t node and it is quite tedious to define a complete tree by hand Thus you can use the pyCGNS Python module to create and process Python CGNS trees This module is not required by elsAxdt but it would help if you want to do some operat
41. ead new CGNS structures If you want to known which structure you actual parser is able to understand please refer to the footnotes in the section The e sA Profile and check your parser version 6 3 1 b File read fails Check your file name is correct in the case of CGNS links check all the files are in the directory where you actually run the computation You can also check the ADF_LINK_PATH This variable contains the list of directories where CGNS is going to look for files the syntax is the same as the usual PATH PYTHONPATH Or LD_LIBRARY_PATH 6 3 1 c No way to import modules Sometimes we have a weird behavior when the PYTHONPATH ends with a trailing colon character for example PYTHONPATH home tools local usr local You remove the trailing colon character and try again 6 3 1 d No coupling script check In coupling mode the coupling script file name is not checked If you give a bad file name you have a weird error such as File CoupingStop py line 1 R A SyntaxError invalid syntax 6 3 1 e Core dump while Python array access Check the Python array package you use in your Python scripts it should be the same as the one used by elsAxdt see elsAxdt startup message ELSA MU 02052 v2 0 elsAxdt 41 49 ONERA RE TD NS Package PR TON RS S THE FRENCH AEROSPACE LAB 6 3 1 f Grid coordinates are missing in output file If you ask for grid coordinates in the result CGNS tree the coo
42. efer to SquaredNozzle Zone 001 instead of Zone 001 even if you have only one base SquaredNozzle CGNSBase_t 4 Zone 001 Zone_t H Zone 007 Zone _t 4 Zone 008 Zone _t Simulation Type Simulation Type_t H 4FlowEquationSet FlowEquationSet_t 4 4 Solver Compute UserDefinedData_t Er Walls Family_t olin Family_t Out Family_t ReferenceState ReferenceState_ t Fig 1 Base level children The elsa solver ignores the CellDimension and PhysicalDimension attributes The computation dimensions are set by the dimensions of the first zone found So far only 3D meshes can be read by elsa CGNS A 2D mesh has to be build with two indices in the K plane to obtain a 3D base you have to do that with your own tools then load the resulting 3D base with elsAxdt The CGNSBase_t node should have at least one Zone t The single ReferenceState_t and the list of Family_t are taken into account The familly structure is stored by the elsAxdt CGNS tree parser these families can be used for BCs or ELSA MU 02052 v2 0 elsAxdt 15 49 es y Profile THE FRENCH AEROSPACE LAB output commands features these points are detailled hereafter in this document You can see Family_t nodes are all at the Base level Thus one should take care of the families names The user defined node Solver Compute should be found at the CGNSBase_t level This node contains the solver specific parameters for the computation
43. elsA Ref ELSA MU 02052 ONERA NERA Version Edition 2 0 a IST a elsAxdt User Manual Date January 04 2010 DSNA elsAxdt User Manual _Python CGNS interface for elsA Release v5 30 2 SA Quality Author For the reviewers Approver Function Quality manager Project head Name M Poinot A M Vuillot S Heib Visa Sign par M Poinot Sign par A M Vuillot Sign par S Heib Software management ELSA SCM Applicability date immediate Diffusion Authors e SA users elsA Ref ELSA MU 02052 ER Version Edition 2 0 elsAxdt User Manual Date January 04 2010 DSNA HISTORY version DATE CAUSE and or NATURE of EVOLUTION edition 2 0 January 04 2010 Edition for release 5 30 of the Python CGNS interface for elsA 1 1 July 19 2007 Edition for release 5 3 of the Python CGNS interface for elsA 1 0 May 30 2002 Creation ONERA THE FRENCH AEROSPACE LAB Guide to CGNS Python interface for elsA Marc POINOT ONERA CS2A ELSA MU 02052 v2 0 January 2010 e H Tass THE FRENCH AEROSPACE LAB Copyright 2006 2010 ONERA France All Rights Reserved ONERA BP 72 29 Avenue de la Division Leclerc F 92322 Ch tillon Cedex FRANCE The elsAxdt module is the Python CGNS interface for the elsA CFD solver kernel It is a standard Python module that can be imported from usual Python scripts The elsAxdt module is required for CGNS based computation or code coupling with elsA Thi
44. ese specfication is also known as the solver profile The restrictions and extensions of the elsa CGNS data specification with respect to the CGNS standard are listed in the same order that the SIDS document e Specific non matching grid connectivities e Extension for partial or extended fields data e Specific boundary conditions e Computing and numerical control attributes The specific non matching connectivities have been added to take into some elsa extra capabilities The Extensions for partial fields data have been added in order to store 1D or 2D data in a 3D field for example The CGNS standard defines how to organize the data associated with the whole zone fields but there is no guidelines for the definition or partial fields results such as a surface The extension describes the way the elsa solver defines these partial fields The Specific boundary conditions are the elsA boundary condition that have no matching types in the CGNS standard The Computing and numerical control attributes are all the elsa solver specific control values such as number of iterations multigrid cycles artificial viscosity or any other parameter related to the solver but not taken into account in the CGNS standard The elsA solver specific nodes are defined to store information dedicated to elsA most of them have ELSA MU 02052 v2 0 elsAxdt 13 49 ONERA Sar A NS Profile ET OR y THE FRENCH AEROSPACE LAB the UserDefinedData_t node type and
45. formation is represented using the SIDS user defined structures A All float values are double floats except the CGNSLibraryVersion t value which usually is reserved for internal use A The order of the children for a given node is not significant and has no effect on the parsing or the interpretation of data by the solver 4 21 Common structures SIDS 4 and 5 The elsa solver performs its computation without any knowledge about the dimensionality of data The DataClass_t DimensionalUnits_t DimensionalExponents t sections 4 1 4 3 and 4 4 are unused In the case such structures are found in the input tree these are ignored Thus all data is understood as non dimensional data normalized with dimensional quantities Thus the DataArray t pattern used as input as well as output doesn t have the dimensional informations and the DataConversion_t are ignored 4 2 2 Entry level structures SIDS 6 2 The first node below the root node is the Base _t and it can have any user defined name You can read as many CGNS trees you would like too elsAxdt will add the CGNSBase_t to the previously loaded bases In the case no base is found in the input tree the error 7100 is raised The Base_t defines the physical and topological dimensions of the whole data in the CGNS tree a 3D base cannot contain 2D zones As elsAxdt can have multiple CGNSBase t the path to the actual CGNS object should include the CGNSBase_t name In the example below you should r
46. have installed the e sAxdt Python module If this is not the case please refer to the chapter Package where a section explains the installation 2 1 The module import We are writing a Python script it can be run using Python or elsA x as interpreters the example listed here after can be run with both but we recommend to use the elsA x interpreter the required shell environment to run it is easier to set than the plain Python environment The first statement you have to write in your Python code is the e sAxdt import The module has few functions you can call For example in the case of a parallel computation you might want to know which are the current process number of your running program and how many processes there are in your computation import elsAxdt rank size elsAxdt ranksize You write these two lines in a myscript py file you run it with the command elsA x myscript py You can do the same with the help of an MPI Python module e sAxdt provides this information to allow you to run a parallel computation without importing other modules but this is restricted to these two usual values rank and size A If you have some error return at this point you should read the section 6 3 Known issues and common troubleshootings in this manual A If you don t use MPI ranksize always returns the value 0 1 A You MUST use a specific Python interpretor or mpirun to run a parallel script 2 2 Reading a CGNS
47. ies these attributes to all zones that match the family 5 MRB FamilyBC a Solver Motion interpol interpol_tool motion axis_pnt_x axis_pnt_y axis_pnt_z axis_vet_x axis _vet_y axis vet_z omega transl_speed_x transl_speed_y transl_speed_z Family_t FamilyBC_t UserDefinedData_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t DataArray_t Fig 30 Zone motion family UserDefined The Solver param is used to set any attribute you want This is a straightforward translation to elsA ELSA MU 02052 v2 0 elsAxdt 31 49 U1 amp amp NN H ONERA Teens Interoperability Features THE FRENCH AEROSPACE LAB 5 INTEROPERABILITY FEATURES The elsAxdt interface is also during run time Some parallel and code coupling features can be used through elsAxdt there is no extra features in the module itself but only those implemented in the solver In other words you cannot add a feature such as making a BC change its state by reading a CGNS tree unless the BC is design to do so in the elsA kernel The elsAxdt provides the user with a Python interface to access code coupling data in a Python fashion or in the case of parallel a way to distribute data on a CGNS tree structure basis in a more general frame this is an interoperabilty feature the user can put in his Py
48. inateY and Coordinatez These names are mandatory The grids coordinates should be stored in a per coordinate basis The CGNS standard requires that any vector should be stored with a data array per vector coordinate the array indexing should have the Fortran indexing i e KJI loop 3 6 GridCoordinates GridCoordinates_t Coordinatex DataArray_t CoordinateY DataArray_t CoordinateZ DataArray_t Fig 4 Coordinate arrays for a 3D grid A Coordinates should be double precision data A The Rind information is not taken into account neither in grids nor in solutions A The grid coordinates are not duplicated in the output tree unless explicitely requested by the user application as described in the next section In the case of a moving grid the pattern Grid lt label gt is recommanded The elsAxdt parser would only read the GridCoordinates node and it is up to the user application to associate links between the actual grid to use at a given step 4 4 FlowSolution SIDS 7 7 The FlowSolution_t node is reserved for computation solution having the same size than the grid itself The solution nodes should use the e sAxdt reserved names at least for the initialisation and for the requested output solution The recommanded name pattern is FlowSolution lt label gt with lt label gt the iteration number Init or EndOfRun There is only one reserved and mandatory name in the case of a restart the initialization so
49. inition of some elsa output control parameters period integer sets output frequency w rt iteration refer to elsA manual writingmode integer refer to elsA manual 1 end of run 2 one each period writingframe integer refer to elsA manual loc integer Sets the GridLocation for the 0 CellCenter expected output 1 Vertex 2 cellfict Please note the important case of the so called cellfict output This output is not CGNS compliant and cannot be stored as a FlowSolution_t Then we add a UserDefinedData_t node named Solver Cellrict which stores the cellfict array values ELSA MU 02052 v2 0 elsAxdt 19 49 ONERA pee RE DRE NS Profile ee OT a THE FRENCH AEROSPACE LAB Er Zone317 Zone_t amp lt Zone318 Zone_t 5 Zone321 Zone_t ZoneType ZoneType E FlowSolution EndOfRun FlowSolution_t GridLocation GridLocation_t Er Solver CellFict UserDefinedData_t Momentumx DataArray_t MomentumY DataArray_t Momentumz DataArray_t Density DataArray_t EnergyStagnationDensity DataArray_t Fig 9 The proprietary so called Cel Fict solution The cellfict output will remain available for migration purpose however the user should note that in the future version of elsAxdt the storage of such a solution will be performed using a CGNS compliant structure The usual values used for output are detailled in the annex 4 4 2 Convergence output When you define a conservative variable
50. install base xdt elsA path path xdt elsA prod production xdt python array numpy numarray Numeric xdt path includes path1 path2 path3 xdt path libs path1 path2 path3 xdt libs 1lib1 11b2 1ib3 6 2 Installation and run time problems 6 2 1 Dynamic libraries The elsAxdt package gathers CGNS elsa with Python and a Python array package Thus you have to install all these libraries in your own platform and to define a correct run time environnement A common issue is to usedifferent packages version at compile time and run time Please take care of the installation and the run time environment 6 2 2 Python libraries The package has to be built with a Python array module This can benumpy numarray Numeric but we strongly recommand the use of numpy You have to insure that the Python scripts you are using actually import the same array module as the one used during elsAxdt production 6 2 3 Parallel mode There are some usual pitfalls most of them are quite simple to solve once you understand what s happening We give you there a bit of advice First you may try to run a elsa in parallel mode without using mpirun gt gt gt import elsAxdt elsAxdt v5 3 elsAxdt using numpy elsAxdt standard elsA startup mpirun must be used to launch all MPI applications Unless you use some specific flavours of MPI you should use mpirun 40 49 elsAxdt ELSA MU 02052 v2 0
51. ions on the Python CGNS trees The output of a Python adpater is a Python object it stays in the Python interpreter memory and you can process it with usual Python methods The arrays in the tree are numpy arrays If you want to know more about large array memory management please read the section 8 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA Sa ee THE FRENCH AEROSPACE LAB Module User Interface A The numpy Python arrays require you to define the array with explicit integer types when you use the ASCII representation of an array array 3 3 i You may also want to go from one representation to another one For example you have a CGNS adapter but you want a Python representation You use the hollow method mytree XdtCGNS 010Disk cgns pytree mytree hollow 3 3 Adapters methods The adapter base class is XdtParser the other adapters such as XdtCGNS and XdtPython are derived class You can always use the base class methods but the specialized classes have syntactic sugar that can simplify your code readability and reduce typo errors 3 3 1 XdtParser The base class of the adapters it provides methods for creating a tree setting options getting parameters or saving the tree You also have special functions that actually run the elsA solver or interact with it The enumerates or constants defined in the module are detailled in a next section For example the flags you can give as arguments to the load method
52. ject cannot release the memory of its array even if the array has been allocated by numpy itself 5 4 3 Actual Xdt memory release We know now that when you dump or load an Xdt tree to from Python the numpy arrays are no more owner of the memory zone The only way to free this memory is to release the Xdt tree using the free method This call deletes all the Xdt tree from the root node including all the node values and the array memory zones We can find some exceptions to this internal free of memory It is in the case the solver doesn t want Xdt to free the array For example if you have a direct access to an internal array used for coordinates a boundary condition or other data that should be protected from unwanted free These cases so called Kernel Ownership are quite rare 5 4 4 Reference Counting Now the memory used by the arrays is released or not and you cannot see your process memory decrease We need to detail two points here First Python manages reference counting on objects When you perform a del call this decrements the counter and when the counter reaches zero Python marks the object as no more used The reference counter is incremented each time you refer to the object for example when you add the same object in two lists or when you use a new variable to point to an existing object An object which is no more used is freed unless you have defined a specific behavior for the object 5 4 5 Code coupling usual scen
53. jects tree it has no proprietary classes and only requires the numpy module for actual arrays This CGNS tree representation can be understood by any Python script without required import The node values such as DataArray_t values are pointers to the same memory zone as your xdt tree shared with the actual numpy arrays This means that for example if you read large arrays of point coordinates these will not be duplicated when you ask for a Python hollow view of your xdt tree 6 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a O O THE FRENCH AEROSPACE LAB Module User Interface 3 MODULE USER INTERFACE The interface is first composed of a Python package with functions classes and other Python services and second with the description of the expected CGNS trees as input and produced CGNS trees as output The CGNS trees structure and the meaning of these structures are not explained in details in this chapter but in the next chapter This chapter describes the Python package interface and some helpful hints for dealing with the Python CGNS trees We assume you are familiar with both Python and CGNS tree structure 3 1 Python package interface The user interface is a Python module The whole elsa solver is loaded as a module the main class interface is the xdt class Keep clear in your mind that there is on one side the Python module so called elsAxdt and on the other side the xdt class This class is defined by elsAxdt and be
54. ll 1 IcingWallData Gaz NeumannData SkinFrictionyY Base Zone ZoneBC Wall 1 IcingWallData Gaz NeumannData SkinFrictionZ Base Zone ZoneBC Wall 2 IcingWallData Gaz NeumannData SkinFrictionyY Base Zone ZoneBC Wall 2 IcingWallData Gaz NeumannData SkinFrictionZ A So far there is no control to check if the BC patch actually is on a surface or not A There is no indication to tell wether the BC values are input or output values ELSA MU 02052 v2 0 elsAxdt 27 49 ONERA Sc A NS Profile nanas TS y THE FRENCH AEROSPACE LAB 4 7 CGNS tree miscellaneous structures There are CGNS sub trees not related to the zones but rather to a global view of the computation 4 7 1 The reference state SIDS 12 1 Er 4 Solver Compute UserDefinedData_t Cl ReferenceState ReferenceState_t Density DataArray_t 1 0 EnergyStagnationDensity DataArray_t H 2 98222994804 LengthReference DataArray_t 10 Mach DataArray_t H 0 829999983311 Momentum DataArray_t 0 98113 MomentumY DataArray_t 0 0 Momentumz DataArray_t 0 042837 Pressure DataArray_t 1 57110059261 Reynolds DataArray_t 6536200 0 Temperature DataArray_t 1 1377799511 TurbulentSANuTilde DataArray_t 1 502507984 7e 08 Fig 23 Reference State data Only one reference state named ReferenceState is allowed at the CGNSBase t level No other ReferenceState would be taken into account by the solver The ReferenceState should contain the relevant
55. lobalConvergenceHistory ConvergenceHistory_t IterationNumber DataArray_t convflux_rou DataArray_t convilux row DataArray_t difffluUx_rou DataArray_t diffflux_ row DataArray_t MomentumxXFlux DataArray_t MomentumZFlux DataArray_t RSDMomentumZRMS DataArray_t RSDMomentumYRMS DataArray_t RSDMomentumxRMS DataArray_t RSDDensityRMS DataArray_t Hl Zone 001 Zone_t Hl Zone 002 Zone_t 9 Zone 003 Zone_t EH 4Zone 004 Zone t Fig 12 Convergence example In the case of a partial field i e a surface on a wall the result goes into the related boundary condition if you have a set of BCs each BC has its own data The Fig 13 is the output you obtain when you have Fig 11 as input For example you have a set of Bc_t nodes defining a surface and you want to have the pressure on this surface You can create a Family _t named SURFACE all related BC_t are FamilySpecified with SURFACE aS FamilyName In the Family t node at the CGNSBase t level you add the Solver Ouput ELSA MU 02052 v2 0 elsAxdt 21 49 ONERA THE FRENCH AEROSPACE LAB cans reqesting the pressure output on this family elsAxdt parses all the corresponding Bcs and adds an output for each Profile EB Zone 004 Zone_t Zonelype ZoneType Er ZoneBC ZoneBC_t E Wall Front BC t PointRange GridLocation 5 BCDataSet Er NeumannData SkinFrictionz SkinFriction SkinFrictionx PointRange Gri
56. location where you have installed elsa are passed to elsAxdt production by means of the usual elsa shell variable ELSAPROD ELSADIST As the elsa libraries are used the consistency between your elsa Python and then elsAxdt installations should be insured In particular the elsAxdt module requires the double precision for float numbers r8 andthe shared libraries so The use of MPI is also strongly recommanded while not mandatory When elsa is built and installed it generates and copies a configuration file named elsAconfig py into the directory SELSADIST Dist bin SELSAPROD if you have a specific elsa build process please contact elsa support The configuration file is a Python script used by elsAxdt to set its own parameters In case of production problem on your platform please first look at this elsAconfig py file We give some details about its contents with the example below but please do not change values in this file unless you know what you are doing class eCONFIG def init self self majorversion self minorversion self production self platform 3 3 04g intelx86 mpi so self mpi True self shared True self pynum numpy self libs eDescp eFact eObf eTmo eRhs eLhs eBnd eJoin eTur eSio eSou eFxc eFxd eDtw eBlk eGeo eEos eMask eGlob eOper eTbx eFld eAgt ePcm eDef eAel eOpt eLur eSplit eChim eXdt config eCONFIG config cc
57. longs to the interface but it actually is something else that the package itself 3 1 1 The elsAXdt module The module can be imported from a plain Python interpreter or from the elsA solver itself Make sure your interpreter or your solver has the parallel capabilities if you want to use the MPI services The module level functions and constants can be used to get set information before the actual creation of an xdt object O args list of strings is used to set the command line argument and pass it to the solver In the case you are using elsAxdt from a plain Python interpreter you have to mimic the command line argument for the elsA solver This function is mandatory in the case of code coupling O iteration is a function call to the elsA solver which returns the current iteration number This is used in the scope of code coupling applications when your coupling script wants to know at which iteration it is performed O get tree name as string returns a solver CGNS tree with the given name if found The returned tree is a Python CGNS tree The reserved names are INPUT_TREE the tree you read at first OUTPUT_TREE the tree used to store all output results and RUNTIME_TREE the tree used as I O shuttle during code coupling computations O output returns the current output CGNS tree found in the solver This can be performed without creation of an xdt object The returned tree is a Python CGNS tree Same as get OUTPUT_TREE lis
58. lution should have the name FlowSolution Init The expected output solution in the input tree and the actual output solution in the output tree have a name with the pattern FlowSolution lt label gt When elsAxdt finds such a flow solution node in the input three it generates the same skeletton for the output tree This output tree will have a FlowSolution_t node with the name FlowSolution lt label gt Zone 007 Zone_t ZoneType ZoneType_t 0 GridCoordinates GridCoordinates_t Ho ZoneGridConnectivity ZoneGridConnectivity_t ZoneBC ZoneBC_t 4 FlowSolution init FlowSolution_t 3 FlowSolution 03000 FlowSolution_t H 4FlowSolution EndOfRun FlowSolution_t Fig 5 Flow solutions in a Zone You can have as many FlowSolution lt label gt solutions as you want A good practice is to use FlowSolution lt iteration gt for the output name and set a link from FlowSolution Init to the initialisation solution A FlowSolution_t without the FlowSolution prefix is ignored by elsAxdt ELSA MU 02052 v2 0 elsAxdt 17 49 ONERA Teens Profile THE FRENCH AEROSPACE LAB The FlowSolution Init would define the conservative variables at the cell center location In the case elsAxdt finds more variable than required the extra are ignored Of there is less variable than expected an error is raised by the solver kernel The distance to the wall required by the turbulence functions can also be set in the initialisation FlowSolutio
59. match connectivity 24 Fig 18 BC with CGNS type 25 Fig 19 Proprietary BC data 26 Fig 20 Proprietary BC data with CGNS variable names 26 Fig 21 BC defined by a family 26 Fig 22 Family BC and proprietary parameters 27 Fig 23 Reference State data 28 Fig 24 Proprietary solver parameters 29 Fig 25 Two kinds of families 29 Fig 26 Family specified overlapping BC 30 Fig 27 Typed BC overlap definition 30 Fig 28 UserDefined BC overlap definition 30 Fig 29 Mask parameters 30 Fig 30 Zone motion family 31 ELSA MU 02052 v2 0 elsAxdt 49 49
60. mentumZFlux flux_row 3 Torquex torque rou TorqueY torque rov Torquez torque row WallCellSize yplusmeshsize BladeSectionForce bladesecforce 3 BladeSectionForcex bladesecforcex 3 BladeSectionForceY bladesecforcey 3 BladeSectionForceZ bladesecforcez 3 BladeSectionTorque bladesectorque 3 BladeSectionTorquex bladesectorquex 3 BladeSectionTorqueY bladesectorquey 3 BladeSectionTorqueZ bladesectorquez 3 BladeSectionForceSquareMach bladesecforcem2 3 BladeSectionForceXSquareMach bladesecforcem2x 3 BladeSectionForceYSquareMach bladesecforcem2y 3 BladeSectionForceZSquareMach bladesecforcem2z 3 BladeSectionTorqueSquareMach bladesectorquem2 3 BladeSectionTorqueXSquareMach bladesectorquem2x 3 BladeSectionTorqueYSquareMach bladesectorquem2y 3 BladeSectionTorqueZSquareMach bladesectorquem2z 3 IterationNumber iteration 3 RadialPosition ronr 3 ChordLength chord 3 BladeAzimuth azimuth 3 ChimeraCellType ichim 3 VortexGeneratorForce vortexgeneratorforce 3 VortexGeneratorLocation vortexgeneratorlocation 3 Notes 1 Available only for Spalart turbulence model the elsA variable name doesn t reflect the actual value used for output 2 Available for K turbulence models 3 Not a CGNS name as described in Annex A of CGNS SIDS 7 3 Tutorial contents The Tutorial files are small test cases which only illustrate the elsAxdt use The tutorial has cgns files you can use as patterns and script files Please note the striked through file names are availabl
61. n CGNS tree structures 4 3 CGNS tree grids and solutions 4 4 FlowSolution SIDS 7 7 4 5 Grid connectivity SIDS 8 1 4 6 CGNS tree boundary conditions 4 7 CGNS tree miscellaneous structures 4 8 Chimera features 4 9 Zone parameters Interoperability Features 5 2 Code coupling user interface 5 3 Coupling script 5 4 Memory ownership 5 5 Remarks about data exchange Package 6 1 Production and installation 6 2 Installation and run time problems 6 3 Known issues and common troubleshootings Reference Tables 7 1 Warnings and errors 7 2 Variable name mapping 7 3 Tutorial contents CONTENTS BARBW OO OO U U 33 34 34 37 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Reference Tables FIGURES Fig 1 Base level children 15 Fig 2 Zone example 16 Fig 3 Complete and incomplete zones parallel computation 16 Fig 4 Coordinate arrays for a 3D grid 17 Fig 5 Flow solutions in a Zone 17 Fig 6 Flow solution used to initialize a Zone 18 Fig 7 Flow solution result a end of computation 18 Fig 8 Proprietary settings for a solution output 19 Fig 9 The proprietary so called CellFict solution 20 Fig 10 Global convergence for all Zones 20 Fig 11 Output settings based on a family 21 Fig 12 Convergence example 21 Fig 13 Surface output 22 Fig 14 Surface output settings 23 Fig 15 Surface output result 23 Fig 16 One to one connectivity 24 Fig 17 Proprietary so called no
62. n t The list of expected variables are detailled in the next table FlowSolution init FlowSolution_t GridLocation GridLocation_t Density DataArray_t Momentumx DataArray_t Momentum DataArray_t MomentumZ DataArray_t EnergyStagnationDensity DataArray_t Fig 6 Flow solution used to initialize a Zone A The turbulence variables are declared using their actual name not the variable name For example one should use TurbulentEnergyKineticDensity instead of the elsa v6 The usual variable used in the init are Density MomentumX MomentumY MomentumZ EnergyStagnationDensity e TurbulentSANuTildeDensity Spalart turbulence model only e TurbulentEnergyKineticDensity TurbulentDissipationDensity K turbulence models e TurbulentDistance TurbulentDistancelndex only for distance computation The can be set without conservative variables You have four options for the initialization A ReferenceState and no distance information you may have to set it as a Solver compute attribute e A ReferenceState and the TurbulentDistance TurbulentDistanceIndex in the FlowSolution Init e A FlowSolution Init with conservative variables and no distance information you may have to set it as a Solver compute attribute AFlowSolution Init with conservative variables and distance information 441 Empty solution as ouput skeletton Any solution with a name starting with FlowSolution EndOfRun is understo
63. n the attribute name These migration methods have pure CGNS names or pure elsa names as arguments you should refer to the documentation before any attempt to retrieve an actual object or its name from the elsa solver We suggest you use the migration patterns you can find in this document or in the module examples Some more migration methods exist however we do not recommend to use them The use of migration methods can be dangerous is you set bad parameters or if you want to access to an object before it is actually created If you really need a migration method and you cannot use one in the list above please contact elsA support 3 5 Module constants Some predefined constants are set in the module dictionnaries These constants are enumerates and can be found in two dictionaries For example you use the constant CGNS as a function argument a elsAxdt CGNS There are cases you only have the enumerate value itself i e 4 but you want to know which constant it is you use the look back dictionary a elsAxdt CGNS b elsAxdt Adapter CGNS c elsAxdt Adapter elsAxdt CGNS 10 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Module User Interface A reverse lookup dictionnary has the dictionnary name with a trailing underscore The module dictionnaries are described in the section hereafter All dictionnaries of the elsAxdt module are in the elsAxdt Enumerates top dictionnary
64. ns These options have to be specified if you want to use the solver for a connection and a data exchange with another software P lt filename gt py Gives the name of the Python file to import at each state stop C lt connection name gt Sets the name for the code coupling connection One should use this string to identify the data in the coupling script In the case each solver instance produces the same sub tree the connection name is used as key for identification The python function ccname returns its value X lt state stop flag gt A bit string flag indicating the state stops This is an hexadecimal OR between values see the code coupling examples D Sets the debug mode on 5 2 2 State stops The code coupling interface is specified by the input and output data structure and the time at which this exchange takes place The finite state machine FSM described in next figure shows the very simple automata of the solver interface The states are O before before any processing in the solver O start before the first iteration O iteration before any iteration O end after the last iteration O after after all processings in the solver before start iteration end after ONERA cans Interoperability Features ee cime CR THE FRENCH AEROSPACE LAB Each time the solver passes from one state to another it checks if there is a request to stop before entering the new state If there is no stop the computation
65. ntaining the CGNS label the CGNS type The SCOPE tree defines the root node of a tree as None None lt children gt CGNSTree t which is not standard All elsAxdt trees are SCOPE trees 38 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Package 6 PACKAGE 6 1 Production and installation The elsAxdt parser is a true Python module It is build and installed with the Python standard Distutils The package contains pure python scripts C code source files CGNS data files various scripts tests The actual installation is performed by Python itself thus it is performed with the requirements you have set for your Python installation 6 1 1 The package structure The Python main tree structure has three directories Doc contains this documentation source files pdf and html versions of the current release Misc directory has useless file for end user Then elsAxdt is the actual module The first level of the directory and the utils have the pure Python files The demo directory is not installed but should be useful for a first try with elsAxdt The src contains the C and C source files for the elsAxdt drivers The test directory is not installed it provides a Go script to run the tests Read the README file for more information on passing tests on your own platform 6 1 2 Production requirements The elsa solver is required You have to build and install elsa The production type and the
66. od as an output specification by elsAxdt Such a solution should be found in each zone where the user wants an output When elsAxdt finds a FlowSolution EndOfRun name it generates an output skeletton for the computation In other words you give a skeletton of ouput tree and elsa uses it as the ouput tree model The children of this FlowSolution_t node can of DataArray_t of UserDefinedData_t types Only the name is used as the variable name to add to the output tree 0 FlowSolution lnit FlowSolution_t amp FlowSolution EndOfRun FlowSolution t GridLocation GridLocation_t Density DataArray_t Momentumx DataArray_t Momentum DataArray_t MomentumZ DataArray_t Pressure DataArray_t Mach DataArray_t Fig 7 Flow solution result a end of computation 18 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA ee THE FRENCH AEROSPACE LAB Profile A If you define a DataArray_t you will have to set the zone dimensions in order to be CGNS compliant Thus you will loose a lot of space if you do not lay on an existing solution you want to keep Moreover as using the DataArray_t does not require the zone dimensions you can use a link to a given zone and then factorize your output declaration A In the output tree no links are created to the grids coordinates reference state or any other sub trees of the input CGNS tree The user as to post process the result and add the links to the grids A As mentionned in the next sec
67. ompiler config cflags g g fPIC D_E FORTRAN LOOPS DE SCALAR COMPUTER D_ELSA COMPILER_INTEL_IA64_ D_E USE STANDARD IOSTREAM DE RTTI I home tools include mpich U_E MPI_ DE MPI DMPICH IGNORE CXX SEEK DE DOUBLEREAL icc shared fPIC zmuldefs IMPR EI E Nb ENre a VibteaLILY METCh PR C OTeMAMIS CAC PEL EMEA CS config cshared config shflags config extralibs ELSA MU 02052 v2 0 elsAxdt 39 49 ONERA Re ES NS Package PR TON RS THE FRENCH AEROSPACE LAB config extralpath home tools local x86 64 1ib opt intel fce 9 0 1ib 6 1 3 Procedure You have to run the following command to produce the module export ELSAPROD export ELSADIST ELSAPROD IA64 mpi so ELSADIST usr local tools python setup py build The installation can be performed in a user defined directory However we suggest to avoid the installation in the Python installation directories but rather to use the elsa installation directories As the packagecontains some machine dependant shared libraries the best is to install the whole set of files using the command line python setup py install prefix SELSADIST Dist bin SELSAPROD You can produce the package with your own set of search directories and library names The following snippet shows the command line specific options elsAxdt specific options see README file for details on installation options help prefix exec prefix
68. on Init to read fora restart Couldn t be used without this linked to file 007 Two different output for an integration based on a same family 008 Simple parallel computation with 2 MPI processes 009 Simple chimera case 010 Simple chimera writing interpolation and masking data as non CGNS separate files 011 Simple chimera reading interpolation and masking data as non CGNS separate files 012 Simple chimera reading interpolation and masking data as CGNS separate files 013 A chimera computation with a doubly defined boundary condition 914 Multiple reference states used for both initialisation and BCs 015 User defined Grid Location for FlowSolution_t 916 Ouput to get GridCoordinates in absolute frame as FlowSolution_t and GridCoordinates t 917 So called coarsen mesh example grid coordinates are read every other point 948 Output to get an unsteady or time dependant as final result of the computation 919 Case of a skeletton file re using links ar restart 020 NCT case with CellListDonor as PointRange mono base 021 NCT case without CellListDonor mono base 022 NCT case without CellListDonor multi base 023 NCT case with PointList multi base 924 Rigid motion on the zone level family 925 Block parameters on the zone level family 926 Case with a Vortex generator write 927 Case with a Vortex generator read using the case 026 928 Use of the GoverninEquationSet tree 929 A 2D CGNS
69. on patch This means the values are global to the whole application patch and is independant of any re meshing Most of the BC pattern is the same as the BC ELSA MU 02052 v2 0 elsAxdt 25 49 ONERA pee A NS Profile ans S THE FRENCH AEROSPACE LAB without data pattern We add a DataSet subtree starting with a BCData_t node The name of this node is left to the user and has no meaning to the solver Then depending on the kind of BC you have you find a Dirichlet or Neumann node and the list of CGNS variables to define The CGNS BCs with global data accepted by elsa are listed in the table hereafter BCInflowSupersonic insup Requires the conservatives variables BCInflowSubsonic injl BCTunnelInflow BCOutflowSubsonic outpres BCTunnelOutflow BCWallViscousIsothermal wallisoth BCWallViscousHeatFlux wallheatflux E ZoneBC ZoneBC_t lt oSymmetry South BC_t lt oSymmetry Back BC t amp InflowSubsonic West BC t PointRange IndexRange_t Solver BC UserDefinedData_t type Descriptor_t stagnation_pressure DataArray_t stagnation_enthalpy DataArray_t tv DataArray_t tyv DataArray_t tv DataArray_t Fig 19 Proprietary BC data E ZoneBC ZoneBC_t EH Symmetry South BC t Wall Front BC t E InflowSubsonic West BCt PointRange IndexRange_t Bo Solver BC UserDefinedData_t PressureStagnation DataArey t f EnthalpyStagnation DataArray_t txv DataArray_t tyv Data
70. one patch PointRangeDonor Or PointListDonor The other zone is identified by the ZoneDonorName The Transform node gives the transformation matrix E Zone 002 Zone t ZoneType Zonelype_t GridCoordinates GridCoordinates_t ZoneGridConnectivity ZoneGridConnectivity_t El Zone 002 Zone 004 GridConnectivityito1_t Transform int IndexDimension PointRange IndexRange_t PointRangeDonor IndexRange_t H Zone 002 Zone 006 GridConnectivityito1_t H Zone 002 Zone 001 GridConnectivity1to1_t ELSA MU 02052 v2 0 elsAxdt 23 49 ONERA eee o y THE FRENCH AEROSPACE LAB Profile Fig 16 One to one connectivity A No PointXxxx and PointXxxxDonor Mix With Range and A Ranges are list of indices by points i e in 3D we have Imin Jmin Kmin Imax Jmax Kmax This allow the use of the same structure for 2D and for unstructured list of points indices 4 5 1 a Non matching connectivity The mismatch connectivity is described in a GridConnectivity_t node The elsa solver requires the declaration of GlobBorders for such a connectivity The connectivity is defined even if you do not have information about the connected zone s the CellListDonor is not used and the connected to zone information is ignored too The mandatory sub node Solver Property of type UserDefinedData t contains the information about the GlobBorders these are type nomatch Descriptor t or DataA
71. rdinates are stored in the FlowSolution sub tree not the GridCoordinates 6 3 1 9 The solver crashes when he leaves the last line of Python script If you use SWIG generation for elsA check your version and in particular the options nodefaultctor nodefaultdtor that should be set 6 3 1 h Installation does not match build You should call the build and the install installation commands with the same command line options Otherwise the install would take the default values of the setup which were overriden by your command line options during the build 6 3 1 i Avoid attribute errors using name scopes You can improve the readability and the portability of your Python code using the import xxx as x statement instead of the simple import However be sure you cannot have name collision never do this from elsAxdt import Using a single letter as new module name can cause troubles which could have unexpected effect on large or imported scripts as described in the bad script below this really leads to big problems import elsAxdt as r r xdtCGNS 010Disk cgns load r s r ranksize oups Overwrite previous Python object named r with an integer The use of so called scope prefix leads to a better readability and reduces name scope problems 42 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA Se Reference Tables THE FRENCH AEROSPACE LAB 7 REFERENCE TABLES 7 1 Warnings
72. ree 5 5 2 Reserved CGNS tree names You do not have to set the tree names if you use usual elsAxdt script patterns Actually the input and the output trees would be enough and you would not need to define new names The default and recommanded tree names are O xdt input tree contains all computation data such as meshes connectivity initialisation solutions boundary conditions The tree is read at start time O xdt output tree contains all requested results at end of computation O xdt runtime tree is the exchange tree used for code coupling it can be read and write but it is up to the application to make sure the modifications are relevant to the computation 5 5 3 CGNS HDF or CGNS ADF files The low level storage used for CGNS files is not managed by elsAxdt but by the CGNS library used at link time In other words if you link elsAxdt with an ADF based mid level library you will obtain ADF low level storage files Then if you want to use HDF5 instead of ADF for your files please refer to the CGNS library production setup When a CGNS file is open with read or write mode the file is actually open or created on disk The usual system constraints have to be remind for example the file ownership file access file quotas etc ELSA MU 02052 v2 0 elsAxdt 37 49 ONERA cans Interoperability Features eae en a THE FRENCH AEROSPACE LAB No control is done by the Xdt The CGNS file can contain symbolic links
73. refering other CGNS files The access control should be done at this level too furthermore the link refers to a name that could be an absolute or a relative path leading to common problems related to opening files in a current directory We suggest to use three levels of files linked together The mesh can be shared by several computations The computation defines equations reference state numerical controls This file actually has links to the mesh file The results which refers to the previous file You can also add a level for the initialisation fields Such a linked files design leads first to a better access time during the read but it also insures a concurrent read of meshes and an exclusive write of the solutions at the same time 5 5 4 SCOPE The SCOPE initiative is an Open Specification of a Python implementation of a CGNS tree A SCOPE tree is a Python tree composed of lists of lists A node is a list and each node has a list of children This children list is composed of nodes which are lists and so on A node has four objects a string a value a children list and a type e Name A Python string should not contain or a single and should not be empty Except these requirements a name is whatever you want Some names are reserved by the CGNS standard e g CGNSLibraryVersion GridCoordinates e Value A value is a numpy array of type 14 18 R4 R8 or C1 e Children a list of nodes e Type A Python string co
74. rray_t globborder string value Descriptor t Or DataArray_t globborderdonor string value Descriptor _t Or DataArray t lt any other attribute gt any value DataArray t MT ZoneGridConnectivity ZoneGridConnectivity_t ae match3_3 GridConnectivity1to1_t ci zonel 5 match3_7 GridConnectivity1to1_t zone2 3 match3_5 GridConnectivity_t zone4 PointRange IndexRange_t 4 1 20 1 16 20 2 GellListDonor IndexArray_t s 1 1 1 11 1 2 Solver Property UserDefinedData_t MT type DataArray_t c1 nomatch globborder DataArray_t ci GBO1A globborderdonor DataArray t C1 GBO1B nomatch_special DataArray_t cr none YGridConnectivityType GridConnectivityType_t Abutting Fig 17 Proprietary so called nomatch connectivity Some defaults values are set by elsAxdt you can always overwrite these values using attributes at the same node level A The CellListDonor has to be present but can have fake values and should be of size 3x2 but in all cases elsAxdt ignores it 4 6 CGNS tree boundary conditions The boundary conditions defined in CGNS are very general BCs The elsa solver has a lot of dedicated BC and most descriptions in this section are extensions to the standard Of course these extensions are taking place in the standard itself and all presented CGNS trees are CGNS compliant The elsa solver also uses the BC structure for a very specific purpose the input of partial solution fields The elsa
75. s document details the interface of the elsAxdt module the input CGNS trees the module can translate for elsA and the output CGNS trees that can be returned from the elsA computations 2 49 elsAxdt ELSA MU 02052 v2 0 Teens ONERA a O U THE FRENCH AEROSPACE LAB Forewords 1 FOREWORDS A part of the elsA CFD solver interface is available through a Python module named elsAxdt It performs solver input output using plain Python data structures and uses the CGNS data model as logical description of information We explain in this document how to use the elsAxdt interface and thus how to use elsA with CGNS 1 1 CGNS Standard The elsA solver is a Navier Stokes fluid solver it can be used on his own or integrated into a more complex multi physics simulations A way to insure interoperability between the codes involved in such a complex simulation is to use a standard The CFD community has a standard for data model definition and disk storage CGNS CFD General Notation System 1 1 1 Public standards The use of a public data standard opens a door to other software customers and providers You can invent your own standard for a given proprietary simulation but you cannot use third party tools It is more efficient for a team in terms of developpement and maintenance to learn and add a CGNS adapter for everybody than to add one translator for each possible code or framework A public standard also increases the user community and thus
76. s you pass to the tree parser myflag READ MESH myflag myflag READ CONNECT myflag READ BC Here you say you want to read the meshes and the connectivities of the argument tree All other parts of the tree will be ignored The two syntaxes for setting the flag are the same Please refer to the Python manual the flags are true Python objects If you want to know exactly what the elsAxdt takes from the CGNS tree refer to the section The load mode flags are dependant This means some of them are forced if you ask for another one For example READ_MESH is forced if you ask only for READ CONNECT or READ BC The load flag has to be set on the mode attribute xdt XdtCGNS file cgns myflag READ MESH xdt mode myflag The way to unset a flag is to make a logical AND with the existing flags and a NOT of the flag you want to unset In the example below the trace mode is set to OFF xdt mode amp elsAxdt READ TRACE 12 49 elsAxdt ELSA MU 02052 v2 0 cans ONERA a ee THE FRENCH AEROSPACE LAB Module User Interface 3 7 4 PROFILE The logical organisation of data managed by elsAxdt is CGNS compliant The Python trees you have as input output or even available during run time have a CGNS data model the node names types their value meaning and the structure of node parents and children are those recommanded by CGNS 4 1 Compliance and profile The compliance to CGNS means that all
77. se two levels in the figure below TAILBOOM defines a surface for the overlapping while FSG is a family that gathers all the zones for the body and mask definition MT 3 ZoneBC ZoneBC t BC on SF404 BC t C1 FamilySpecified PointRange IndexRange_t ta 1 61 13 5 61 2111 FamilyName FamilyName_t ii TAILBOOM YInwardNormallndex int IndexDimension 0 1 0 we BC_on_SF466 BC_t cl FamilySpecified FlowSolution EndOfRun FlowSolution_t FlowSolution AbsoluteCoordinates FlowSolution_t FamilyName_t a FEG Fig 25 Two kinds of families YFamilyName The overlap BC can be defined in two ways either at the Bc_t level or at the family level We strongly ELSA MU 02052 v2 0 elsAxdt 29 49 ONERA Sa A NS Profile ET OR Sy THE FRENCH AEROSPACE LAB recommand to use the family which would be re used for the neighbourhood Eo ZoneBC ZoneBC _t MT o moverlaps UserDefinedData_t cl FamilySpecified PointRange IndexRange_t 14 C 113 59 1 117 59 2 FamilyName FamilyName_t cl SlatOverlap Fig 26 Family specified overlapping BC The actual type of the BC can be BCoverlap which is NOT a CGNS BC type or UserDefinedData We recommand the use of UserDefinedData which is standard The Solver Overlap is a sub node of this family it defined the list of families used during neighbourhood search The list is composed of lt CGNSBase t name gt lt Family t name gt separated with
78. t returns the list of the names of the CGNS trees currently in the solver ccname returns the name of the elsA instance i e the c option value state returns the current state of the solver see section 5 2 2 State stops 0000 ranksize is used for parallel application it returns a tuple of integers rank size required for parallel distribution of tasks amongst the processors The call can be performed even in sequential mode i e without a call to the MPI library in that case the tuple 0 1 is returned The use of the arg function usually is made with the sys argv variable In the context of code coupling the pattern shown in the code snippet below is mandatory because elsA is expected command line args for the code coupling parameters import elsAxdt import sys elsAxdt args sys argv There is another function xdt used for a new xdt object creation You can create such an object using this xdt call with the required arguments or you may prefer to use the interfaces provided by the Xdt classes The next section describes these classes ELSA MU 02052 v2 0 elsAxdt 7 49 ONERA pene el ei NS Module User Interface nanas S y THE FRENCH AEROSPACE LAB 3 2 The Xdt class An Xdt tree is C tree with a CGNS tree layout i e the organization of the nodes their names etc The elsA solver can run a computation using all or a part of the data it can found in this tree The overall structure of
79. the elsAxdt parser A if variable name is not allowed for such an output elsAxdt ignores the entry A In the current elsAxdt versions the PointList is always interpreted as a PointRange See next section for details on where the partial field will be located in the output CGNS tree gt FlatPlate CGNSBase t E Fluid Zone _t Zonelype ZoneType GB ZoneBC ZoneBC_t El Wall Heat BC t PointRange IndexRange_t GridLocation GridLocation_t Er DataSet 01 BCDataSet_t Gl NeumannData BCData_t NormalHeatFlux DataArray_t PaintRange IndexRange_t GridLocation GridLocation_t amp FlowSolution EndOfRun FlowSolution_t E GlobalConvergenceHistory ConvergenceHistory_t Fig 15 Surface output result 4 5 Grid connectivity SIDS 8 1 The mesh connectivity should be a multizone connectivity in the ZoneGridConnectivity_t node The nodes taken into account are of type GridConnectivityltol_t and GridConnectivity t 4 5 1 Matching connectivity The connectivity with type GridConnectivityltol t is reserved to zone interface with matching patches The connectivity sub tree should declare one interface per zone Then if you have a zone with more than one interface to another zone you have to define one connectivity per patch The onor zone that is the opposite zone of the patch would also have the definition of the interface Each sub tree contains a local patch PointRange or PointList and the other z
80. thods The class has the following attributes some of them are automatically updated but you can overwrite the values O mode flags to drive the parts of the tree to read O selection list of Zones to read list O distribution association Zone processor number dictionary O debug level of debug trace 3 3 2 XdtPython The Python adapter is a specialized XdtParser that can use a CGNS Python tree as entry This adapter is used when you want to create or modify CGNS trees during run time or depending on specific parameters you may only know at run time This is the required adapter for code coupling with the solver since we do not use the file storage for data exchange but rather the actual data in memory ELSA MU 02052 v2 0 elsAxdt 9 49 m WNE ONERA Pa A NS Module User Interface PRET Oa S THE FRENCH AEROSPACE LAB XdtPython creates a new object argument is the Python object i e a list to use as tree The pyCGNS module uses the Python in memory CGNS tree You can load store trees on CGNS files and use many useful methods for your Python scripts This pyCGNS module is not included in elsAxdt it is a stand alone Open Source Python module A This class interface is not stable and we do not recommand to use this class in your operational scripts 3 3 3 XdtCGNS This adapter is used when you use a CGNS file You should give the complete file name with the extension the adapter does not care if you have an adf
81. thon applications 5 1 1 Overview of Python external script If you want elsA to exchange Python CGNS trees so called SCOPE trees you have to run elsA with the C X and P command line arguments We will see these options later in this section These options are setting elsA in the code coupling mode The code coupling mode forces the solver to suspend a computation and to delegate the control to an external Python script This external script or the callback script has to be written by the application developper as well as the main script of the computation Once the script is performed the solver resume its work You have to set when the solver has to suspend and what kind of run time data it should read write elsAscript py couplingscript py x XdtCGNS file print a a 4 cfdpb compute V The elsAscript is executed then at each iteration the coupling script is executed The two scripts are run by the same elsA if you define a varaible a in the main script the variable would be available in the couplingscript If you change its value this value is kept from one iteration to the next one 5 1 2 Example The code coupling command line gives the connection parameters These are the states where the solver is expected to suspend itself the name of the connection and the script to run while the solver waits The name of your code coupling connection would be used by the solver for its internal use For example
82. tion a request to a conservative variable implicitely forces the output of the corresponding convergence A As the init solution and the output solution may have the same dimensions it is possible to use the init solution as output skeletton again using the links mechanisms A special name is reserved in the case your solution starts with FlowSolution AbsoluteCoordinates the parser would force the coordinates output to be absolute These coordinates are put in the flow solution sub node not in the grid coordinates sub node You can set the elsa specific parameters using the Solver Output node You add a child with the elsa parameter name as node name and its value with a DataArray_t node type 5 FlowSolution EndOfRun FlowSolution_t GridLocation GridLocation_t Density UserDefinedData_t Momentum UserDefinedData_t Momentum UserDefinedData_t MomentumZ UserDefinedData_t Pressure UserDefinedData_t Mach UserDefinedData_t E Solver Output _ UserDefinedData_t loc DataArray_t period DataArray_t Fig 8 Proprietary settings for a solution output For example for a given flow solution the grid location of the data is fixed If you want vertex solutions and a CellCenter solutions you should have two separate solutions or per grid location Use the loc attribute defined below The default value is CellCenter The Solver Output node is a UserDefinedData_t node with the def
83. ts floating point numbers is a physical view of the same point In this document the data specifications or models are represented as trees That is a node with a set of nodes so called children and so on Each node has a type a name a value and a list of children We detail in this manual all the trees elsA can read write or even update during run time These trees are logical views of data and you need an actual physical mapping to use them in your application For this mapping the elsAxdt provides a Python representation of the tree data which is the default in this package In that case the data are kept in memory as 100 Python objects with pointers to actual elsA arrays in memory The elsAxdt interface provides means to store this Python CGNS tree on disk The physical representation is then the one provided by your own CGNS library In other words if you have a CGNS HDF5 library and you use it in the elsA production the actual physical files will be HDF5 files If you have a CGNS ADF library the elsA files will be ADF files All these translations are easy to use these are not time nor memory consuming and insure a good consistency for all simulation co workers 1 3 Python arrays implementation You should take care of the actual Python array implementation you are using The elsAxdt module can be used with one of the three usual arrays modules Numeric numarray and numpy The numpy module is the default and strongly recommended
84. ur Python script for any purpose your application would like to 2 4 CGNS trees layout The main script looks very easy to write Actually this script only contains the read and the write of the CGNS tree As you may note the output tree is not defined in any way There is a default output tree If you do not specify a specific output the returned tree contains conservative values for all zone fields with the associated convergence per zone You have to refer to next sections if you want a complete description of input output CGNS tree for elsAxdt The input tree is never modified it is taken as read only by e sAxdt the output tree is generated from scratch with the same base zone structure but it does not contain the grid coordinates As explained in the next sections you have to make links to the grid coordinates by yourself One should understand that the tree in memory once you have performed a read is not a Python tree but rather an xdt tree This means you cannot use the mytree object with any kind of Python functions you can only use the xdt class interface You can have the same tree as a 100 Python objects tree you use the hollow method on your xdt tree instance import elsAxdt mytree elsAxdt XdtCGNS 001Disk cgns scopetree mytree hollow print scopetree The variable scopetree is a list of lists representing a Python tree or CGNS Python tree see section 5 5 4 SCOPE This Python tree is a plain Python ob
85. verwritten at the next call to the coupling script Then it would be up to Python to make the tree variable point to the newly created Python list coming from the get call The previous Python list created in the rpevious coupling script call will see its reference counter decrease and the list will be freed if no other Python object is using it import elsAxdt as X tree X get xdt output tree print tree X free xdt output tree del tree Now you understand that if you call free before the print the result will be a memory error when you reach the numpy arrays that have no more allocated memory The second example shows a classical MPI send receive The send data can be freed but we really want to keep the same tree structure we had as output 5 4 7 Overview of parallel The elsA solver can be run on several processors using the parallel features brought by MPI A parallel computation is a distribution of the zones on the processors It is up to the user application to indicate on which processor a zone has to be taken into account The input tree for a given processor should contain at least the zones taken into account by this processor The other zones are not required However an easy way to pass a correct tree to all processors is to give them the whole CGNS tree and a list of selected zones The elsAxdt processing would only create the zones related to the target processor The output tree is restricted to the loc

elsAxdt User Manual Python/CGNS interface for elsA

Contents

Download Pdf Manuals

Related Search

Related Contents