DARE-III(ID:7447/)Continuous simulation languagesRelated languages
References: Extract: Introduction Computer simulation is vital for engineering development and partial system tests. ~ Interactive digital simulation systems2 let an experimenter enter and modify programs from a keyboard and, with a minimum of delay, present graphic- display results as the simulation proceeds. Since experiments will be modified and rerun many times, close interaction is necessary between the experimenter and the model. Such interaction requires convenient facilities for program and data entry and modification (editing), fast program translation and solution, prompt error reporting, selection of different integration routines without retranslation, runtime graphic displays and CRT-terminal listings, self-documenting hard-copy output of programs and data and report preparation, interactive file manipulation (storage, retrieval, and combination of programs and data), and provision for interactive simulation studies involving multi-run experiments (crossplots, statistics, optimization, model matching). While much digital continuous-system simulation is still batch processed, interactive simulation is usually preferable. Interactive simulation is necessary for partial system tests involving, say, real missile-guidance components on a flight table Positioned by a Computer simulating the vehicle. Interactive digital simulation has been realized by the various Dare differential analyzer replacement simulation systems developed at the University of Arizona since 1968. In particular, Dare/Eleven runs on DEC PDP-I I minicomputers, 2 and the newer Dare P/l system is portable machine-independent) and suitable for timesharing as well as for VAX 32-bit superminicomputers. Dare programs translate into Fortran and can access all the facilities of that language. Fortran-based simulation languages like CSSL IV, ACSSL, or Dare require translating of system equations into Fortran, compilation of the resulting Fortran program, and linking of the binary program with a set of library routines for integration, display, etc. Dare systems implement this translation sequence automatically on a COMPILE command. The complete translation process requires from 35 seconds for short programs to several minutes for longer programs. This period is tolerable if we have many interactive runs with the same model, since parameter changes do not require recompilation. Research simulation, however, often involves frequent changes in the model (modified system equations) and/or changes in the experiment, and thus in the job- control program that calls the simulation runs. Under these conditions, even a one-minute recompilation becomes annoying, especially if the translator or compiler rejects a program change because of an error. Translation delays interrupt the experimenter's train of thought. For this reason, immediate simulation-language translation with a line-by-line interpreter is attractive, even though the resulting code will necessarily execute more slowly than compiled code. Interpreter systems permit at least primitive line-by-line editing without a special editor, and they report many syntax errors as soon as a program line is typed (interactive programming). The best-known interpreter language is Basic, which can serve modest simulation applications. As an example, the University of Arizona s BDare simulation package, written in ANSI minimal Basic, runs even on eight-bit personal computers. Basic, though, is inadequate for larger simulations. Interpreter execution is necessarily slow; in particular, the entire derivative computation and integration loop must be translated again and again at each integration step DT. Few Basic systems, moreover, permit double-precision integral accumulation, which is absolutely necessary to prevent roundoff errors in simulation. Basic can also be compiled, but its compilation produces less efficient code than Fortran. A different solution is needed. Extract: Microdare I, II, III The Microdare I and II systems, developed at the University of Arizona, were the first to combine a job-control interpreter and a simple compiler language, which emulated analog-computer block diagrams with threaded-code routines. Microdare III added a real inline expression compiler. Microdare, designed primarily for fast digital signal processing and control rather than for simulation, has only two integration rules and compiles fixed point operations to gain speed. The simulations are, then, very fast but must be scaled. The University of Arizona s new language, Desire (Direct Executing Simulation in Real Time), on the other hand, employs a scale-factor- free floating point language and permits the user to overlay any number of different integration rules without recompilation. Extract: Summary Summary Early Desire is a new floating point equation-language system for interactive dynamic system simulation. It runs 1.3 to five times faster than threaded Fortran and executes immediately on a RUN command without any external compiler or linker. An experimenter thus obtains results of model changes at once. An interpreted job-control language serves for interactive program entry, editing, and file manipulation and for programming multi-run simulation studies. The dynamic program segment containing differential equations in first- order form is entered just like the job-control statements and accesses variables with the same name. An efficient, extra-fast mini-compiler translates the dynamic segment practically instantaneously. Different precompiled integration routines can be overlaid from mass storage while the program runs. Early Desire runs on any 28K-word PDP-I I or LSI-I I processor with FIS or FPU instructions under the RT-I I operating system and is configured for up to 40 state variables. The system provides a true CRT screen editor with cursor-control and number-pad keys and produces runtime graphics even on VT-52 alphanumeric terminals and on VT-I I displays. Future Desire systems will run on 32- bit super-minicomputers and, for extra speed, on minicomputers with attached array processors. in COMPUTER 16(5) 1983 view details |