Dsmake Overview

Dsmake (Object Data Dependent Code Generator) is Altair's proprietary preprocessor, which transforms the Object Definition contained in the source files into the C source files. Objects can be treated as simple C struct's equipped with specialized private functions, and generic capabilities. These generic capabilities are different from typical properties of objects in C++ or Smalltalk: there is no data encapsulation, nor inheritance, althoug limited cabability for derived classes and virtual methods is provided using programmatic methodology (similiar to the one used in most GUI packages written in C, like Xt or Motif)

The basic scheme is as follows;

An ASCII text file, called the Object Definition file, (ObjDefs.defs ), contains a description of Object types or a given module of a program. This file is then passed through a program generator, named dsmake, to produce a library of Object data dependent code.  Each Object definition creates one header file, which contains the definition of the Object's structure, all required constants,  and necessary #include's to other Objects. Developers then use a standard generic interface to access this data dependent code (i.e. to manipulate Objects).  In C/C++ it is efficient to use direct access to the content of each struct, but formalized API's to access each Object from C, Fortran,Tcl and interactively from GUI are provided. At this level, no attempt is made to provide organizing data structures to these Objects. This is provided for, on request, by an automatic Object organizer package (ObjArrays, LinkLists, etc.).

The generated Object data dependent C code consists of numerous C-functions, and supporting header files. Presently dsmake creates C functions, which for any Object allows the developer to;

dsmake organizes these generated C functions, by Object and field, into arrays of function pointers. It also provides a unique symbolic representation for each Object, and for each of its respective fields, which corresponds to the indices of these arrays. These arrays, then, coupled with the symbolic representations provide the programmatic link between the Object data dependent and the Object data independent code.

After generating the Object data dependent C code, the developer then compiles and binds this code with the generic Object data independent kernel. resulting in a working database. The data independent code contains functions which operate on Objects in a generic sense, making no assumptions about the form of the data of any particular Object.
 

Object Organizers

All Objects can have automatic iterators maintained consistently with Object's creation and deletion. User can as for number of objects, for the Object at the position  n (roughly equivalent to the order of their creation) or for the next object of given type). Developer can create arbitrary organizer of any type of Objects, (ObjArray, LinkedList, Stack, etc.), add and remove Objects to/from this organizer, and iterate through the content of organizer.

Object Set  module provides specialized type of organizers, which is very powerfull and can be used almost everywhere in the code. Set module allows to define various membership criteria (e.g. all elements with given material, or nodes within given bounding box) and create derived sets using Boolean arithmetic (inside the code or from the GUI form). All resulting sets can be used in the solution process (e.g. solve the stresses inside the bounding box), adaptivity (estimate errors and adapt all elements of a given set) and postprocessing (display temperature and pressure on the boundary of subset of elements with given rock properties). Set Editor Form allows to inspect all current sets, create the new ones using standar boolean operators, and Bounding Box Editor Form  is used to create interactively element sets based on the bounding box. Set module is very robust: Objects in the set can be deleted, Finite Elements refined or unrefined, new Objects with given property (e.g. within the bounding box) can be created, and the content of the set is automatically maintained. Set definitions are included in the save/restart capability, so they are automatically recreated when the restart file is loaded into the code.
 

User Data Objects

Gridfile module extends the internal definition of the Objects to the input data file (gridfile - roughly equivalent to the BULKDATA file used in Nastran codes). This module implements form of Object Class inheriitance, where typical user data objects (material/property data, forces, boundary conditions, coordinate systems, etc.) are all descendants of the generic Object Class. This generic class allows automatical interpretation of the input data in the free format, with each Object and Object's data (e.g. Viscosity, Poisson's ratio, etc) can be independently named in the input file. Developer has to provide simple creator function, which can simply associate input fields with object fields, but it may be more complex, allowing to test validity of data, providing for alternate ways to input values, compute default missing components, interpret textual flags, and use functions instead of constants for various data values. As an example, 'Viscosity' property in the Material Object may be defined as a function of temperature at the current point. The actual definition of the function will be provided in the gridfile, either as a table of values (uniform or non-uniform spacing, random nodal values for multidimensional functions, linear or quadratic interpolation), or as a procedural definition in the form of simple mathematical expression or complex procedure (Tcl language is used as underlying expression evaluator).