...layers

Material layer, in this context, also includes local material modifications such as implantation regions.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...layout

If additional technology parameters - sheet resistances, layer thicknesses, etc. - are supplied.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...loop

Not shown in the picture are loops at equipment level used to adjust equipment settings according to process step specifications. See [SGHH91] for an example of an equipment-level process control system.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...dimensions

According to [Dav96], channel lengths are expected to reach 3#3 by 2004, gate oxide thicknesses 2.5 nm. See [Ass94] for a summary of expected trends and developments in semiconductor technology.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...specifications

Traditionally, circuit design has been at the demanding side, while process design strives for fulfilling these demands.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...parameters

A modification of the process sequence is opted for only very reluctantly, as it introduces additional degrees of freedom; existing process flows are used as long as possible.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...parameters

Calibration parameters are either explicitly provided by the simulator, or they may be introduced artificially in a pre-processing fashion.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...defined

In the sense of a combination of generic tool plus calibration.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...ranges

So do, of course, the processing results after each fabrication process steps. As these effects are, in the general case, statistically uncorrelated, only the overall impact is of interest.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...parameters

See [AAY126#12688] for the application of statistical methods to VLSI design.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...method

See [HH83] for an introduction to Monte Carlo methods.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...sequences

In the sense that some implants are only applied to one type of device.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...profiles

As of now, no measurement techniques exist for determining the two-dimensional distribution of dopants in the wafer material.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...used

See, for instance, [SC96] for a completely unstructured alternative to polynomial modeling.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...PROMIS

Initially, the PROMIS process simulator [HHW126#12690] [HPW91] was capable of simulating fabrication process sequences including ion implantation and diffusion processes. In order to make different numerical and process-modeling related components independent from each other and to foster the modularization of the simulator, it has been divided into a set of tools each modeling only a single fabrication process [Sti93] [Puc96].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...costs

In terms of CPU time, disk space, network traffic, and user patience.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...manually

I.e, manual setup of scripts for optimization, design of experiments, etc.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...applications

External applications run as separate processes on the operating system.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...gridders

VISTA/SFC uses external gridding tools to ensure data consistency after simulation steps; see Section 4.3.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...callback

See Appendix B.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...fingertips

The level of task abstraction presently supported by the VISTA/SFC GUI does not completely reflect all available functionality. Future work is required to establish concepts for the representation of complex TCAD tasks.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...processes

Processes also include daemons and servers that run completely independent of the simulation environment, whereas executables are started on behalf of it.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...clients

The distinction between client and server sometimes cannot be clearly made. In the example of an optimizer, for instance, it can be regarded as a server that provides the optimization service and delivers the optimum found, or as a client that requests evaluations from the simulation environment.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...interface

See Section 8.3 for a description of the class of Evaluable Entity (EVE) objects.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...VLISP

See [Tup96] for more information on the Vienna Base System and the Vienna LISP Interpreter. Appendix A gives a brief overview of the LISP programming language and the XLISP and VLISP implementations.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...simulation

Note that the scope of this chapter does not include device characterization or circuit simulation, although most of the concepts and solutions presented can be extended to these domains.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...characterization

In lieu of device characterization, the term parameter extraction is often used.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...tools

Trivial as it sounds, this fact is not always present when process simulation tools are being devised. Rumor has it that even at the most prestigious institutions tool developers are sometimes baffled when learning of the requirement for a tool to be able to understand its own output.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...simulation

See Section 6.2.1 on how to avoid the potential computational overhead arising from this abstraction.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...addressed

Geometry objects like points, lines, and faces are not represented as separate objects, but are organized in lists. E.g., a pointList object contains all points of a geometry, a segmentList contains all segments. To access a single point from a point list, an indexed reference is used.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...geometry

These inconsistencies are not a consequence of negligent behavior on part of the tools in question, but necessary simplifications to let tool developers focus on the core issues of the simulation task.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...segments

For a detailed analysis of grid-related tasks during merging see [Hal94].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...necessary

In a typical multi-tool process flow simulation scenario, numerous calls to the gridding utility are made to reconcile tool output with pre-existing data. We have found it to be extremely helpful to differentiate between ``new'' data generated by a simulator, and ``old'' data that already is the result of a re-gridding operation.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...modules

PIF-tool developers can rely on powerful libraries that support the internal handling of grids [Fas94].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...application

Using a standardized VLISP function pointer as interface between the run controller and external tools keeps the interface as flexible as possible, while still allowing for the automatic generation of the function code from a more abstract tool description to minimize manual coding efforts. As VLISP does not differentiate between code and data, it is ideally suited for code generation applications.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...NAME="990">

For the sake of clarity, the LISP representations of boolean data, T and NIL, are mapped to a pair of strings, e.g., "on" - "off" or "yes" - "no".

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...signature

In analogy with the terminology of many programming language, the signature of a tool refers to the names, positions, and respective ranges of its input parameters.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...types

In most cases, the data type can be derived from the default value.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...file

It is not required to define an external tool before calling the STI. By default, an unknown name is used as-is on the command line.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...call

VISTA/SFC is fully operational on both UNIX and VMS systems.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...(NFS)

As a matter of fact, NFS is based on RPC. See [Ste90b] for a detailed introduction to UNIX network programming.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...data

This might not cause trouble with sequential files; for random access files, it is disastrous.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...given

Etch and deposition rates are again an abstraction from the treatment of the wafer, e.g., a CVD or RIE process taking place in a given piece of equipment.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...views

Equipment recipes, process steps, and process flows contain a great deal of information, and many different kinds of users generate or consume that information, including unit process developers, facility operators, technicians, and managers, as well as process flow and device developer [DCB93].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...parameters

As a matter of fact, uncertainties and ambiguities still exist in terms of the executables invoked on the system level and their respective versions. Two identical system calls on the same system need not produce the same result, if, e.g., the search path (on UNIX) or the definition of a symbol or logical (on VMS) get changed between the calls. Furthermore, installing a new version of a tool may also lead to changes in the results computed. Some of these pitfalls may be circumvented by specifying complete path names, checking the file modification time of executables invoked, etc., but such procedures are rather cumbersome and therefore not widespread.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...established

In many TCAD groups, only one dedicated simulation tool exists for a given process step, rendering the tool selection stage trivial. In general, though, a number of choices exist for each process step.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...unchanged

This approach is also aimed at sharing process flow files among several users, with each of them running different experiments and performing different analyses without need to create private versions of the complete file.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...SiO

Possible ambiguities in material names are resolved by a material data base [Ins96b] that uses an inheritance-based strategy to identify material specifications.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...passing

Non-positional parameter passing   does not enforce a strict order for passing parameter values, but allows each value to be specified together with the parameter's name. A default value may be specified for each parameter that is not required to be set explicitly.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...[#B0307##1#]

The editor supports different modes of operation, one providing editing capabilities for layout data as well as for cut-line and sub-domain definition.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...NAME="1544">

See Section 9.2 for an overview, [Wol95b] for a concise treatment.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...processing

See Chapter 8 for more details.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...points

While a split point denotes the location the first difference between two runs, a split branch represents the part of a run that differs from its split parent.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...started

See Chapter 4 for more details.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...undisturbed

An invaluable advantage of subdividing numerical modules and controlling instances into separate executables lies in the fact that segmentation faults, floating point exceptions, and so forth, which otherwise would have a detrimental effect on the framework, affect only the tool they occur in.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...time-consuming

In terms of time spent by the user. CPU time, in general, is available in plenty, although shortages occur during peak hours.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...development

A tendency exists with some companies to delegate some of these aspects, e.g., the tool development part, to external institutions, arguing that the expert status of such dedicated providers increases the quality of the service rendered and decreases its costs. Nevertheless, we feel that the close interaction between technology and TCAD groups is of vital interest for both parties, because it minimizes reaction time and fosters each side's insight into the other one's needs and concerns.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...settings

Variations of the device geometry by changing lithography mask dimensions exert the most harmful influence on the stability of the final simulation results as they induce changes in the grids used by process and device simulation tools.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...values

Note that code modifications can be quickly applied to the VISTA/SFC environment by loading appropriate LISP files. Being read by the interpreter, such modifications have immediate effect; there is no need to leave a session for this purpose.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...operation

A major complication in building graphical user interfaces atop any application arises from the inherently parallel nature of access through the GUI. Strict and rigorous state checking is indispensable to ensure robust and reliable operation.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...possible

See [KGMB94] for a detailed account of import design aspects of graphical user interfaces for CAD frameworks.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...chosen

As a matter of fact, the actual value of the colors used for displaying states is specified in the user's X resource file, e.g., .Xresources. The environment uses names like "active" or "busy" to modify the display state of visible objects. The colors presented here have proven to convey most concisely the intended meaning.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...tool

The actual program to invoke is determined by the user as a configuration variable of the framework. Available plotting tools on UNIX include gnuplot, xmgr, xpif1d, and a built-in visualization module based on the SVG library [Hal94]; additional plotting tools can be added to this list easily.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...xpif1d

Xpif1d is a one-dimensional visualization tool based on Sander's XgPlot [San92].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...accessible

One invaluable advantage of the PIF data format lies in the random access it provides. Therefore, names and types of all data items present in a PIF file can be easily inquired. The handles returned may then be used for reading the data at a later time.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...two-dimensional

Although the operation of the VISTA/SFC environment is independent of the geometrical dimension of the simulation, two-dimensional applications are the most common ones. One-dimensional simulations of modern VLSI technology cannot accurately enough model the complex topological situation found in state-of-the-art devices, three-dimensional simulations are still too slow and produce too much data to be applied routinely in an automatic fashion.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...[#N0019##1#]

Other implementations exist apart from [FFS126#12691]. See [WN91] [BHW91] for details. .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...VISTA/SFC

The Semiconductor Wafer Representation (SWR) [SWR92a] [SWR92b] offers an alternative that has not been used too widely yet.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...step

A type specifier not only determines the data format of a data item, but may also contain information on its meaning.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...domain

Agreement exists as to the meaning of the word itself in the English language. In [AHD83], a task means a difficult or tedious undertaking, [Web81] defines it as something hard or unpleasant that has to be done.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...Simulation

Almost all process and device simulators expect exact values as inputs and generate exact values as outputs, in stark contrast with physical reality. See Section 2.1.8 for methodologies employed to represent nature's fuzziness in the virtual world of TCAD.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...sequences

These fluctuations, in general, are far below the tolerances and measurement limits of all factory settings and measurement procedures. Nevertheless, as simulators produce exact numbers, grid alterations or other numerical phenomena may introduce artificial effects that lead to erroneous results.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...NAME="2042">

Experience shows that it is a good idea to refrain from using case-sensitive  names at all, considering the difficulty in communicating verbally the cases of letters. As a matter of fact, the advantage of a greater number of available names due to case-sensitivity is more than compensated for by the increase in typing and reading efforts. Within the TCAD world, only chemical formulas seem to justify a case-sensitive interpretation of names. Wherever a significant distinction between the cases of letters is made, ambiguity and confusion arise. Furthermore, I would go as far as to claim that significance should only be attributed to the sequence of letters and digits in a name, disregarding their cases. In other words, two names should be considered equal if they agree in the sequence of their constituting letters and digits, skipping all other characters such as spaces, dashes, underscores, etc. The VLISP functions name and name=     implement this philosophy.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...system

In principal, the inputs need not be explicitly dealt with but could be modeled as additional system outputs. They appear here for the sake of clarity only.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...update.

See [GMW95] for optimization techniques and algorithms.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...algorithm

The Levenberg-Marquardt algorithm includes the minimization task as well as the fit error calculation in Figure 8.12.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...program

Access to an external plot tool is again encapsulated by an agent object; therefore, any number of different programs can be easily integrated using the standardized agent interface.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...device

A transistor which can be sufficiently described by one-dimensional equations.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...NAME="2581">

A concise treatment of short channel effects can be found in [Wol95b].

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...NAME="2584">

The threshold voltage itself represents a rather elusive term, as no universal agreement exists as to a single - physically motivated - definition of such voltage. In this example, it is pragmatically defined as the value of the gate voltage at a drain current of 1 nA.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...directly

Given the high degree of autonomy, robust operation, efficient disk space usage, and automatic load balancing as provided by VISTA/SFC, running an optimizer directly on the simulators is not an infeasible task and avoids many drawbacks associated with derived models.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...operation

All parameter definition, experiment generation and management, response-surface modeling, and optimization operations are also supported by the GUI.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...sequence

Both the sequence  and parallel  statements create objects that provide basic control capabilities for executing a set of asynchronous VLISP statements - user-defined or built-in - sequentially or in parallel, respectively. They liberate the user from explicitly taking care of connecting the statements by means of callbacks. The sequence and parallel statements can be arbitrarily nested.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...generated

Simulations are only started if no results exist in the run data base for the given process flow and parameter settings.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...form

As opposed to solve problems that are hard-coded in the program code and can only be altered by re-compilation.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...routines

The VLISP mechanisms to automatically interface with C or C++ code, making it available to the LISP user, seems to provide a sound basis for translating to C or C++ much of the LISP-code presently used in VISTA/SFC.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...paradigm

As VLISP intrinsically supports multiple parallel processes and contexts, the server need not necessarily be implemented as a program separate from the framework.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.