Nash - Scientific Computing with PCs (523165), страница 4
Текст из файла (страница 4)
Even so, thesetasks may be a very important part of our overall scientific computing effort. We take the viewpoint thata problem is not truly "solved" until the results have been convincingly reported. Thus preparing reports,talks and papers, along with a variety of other work that supports and permits the solution of problems,is scientific computing.2.1Non-mathematical Data Processing Related to CalculationWhile this book is about scientific computing, it is rare that PCs are engaged in formal calculation workfor more than a small percentage of the time.
The primary reason for this is that PCs serve their users asdata and word-processing tools, as well as engines to perform computations. In contrast, a highperformance supercomputer will generally be fed tasks by other computers. We do not expect suchnumber-crunching boxes to occupy their expensive cycles waiting for keyboard entry or mouse clicks.Because PCs act as a doorway to calculations, we shall point out how they prepare for and controlcalculations and how they help manage and report results. PCs are also used to direct tasks on othercomputers, or gather data or bibliographic citations from remote sources. These tasks, while tangential toscientific computing, are nevertheless essential to the advancement of research.
Especially in the socialsciences and statistics, we may need a great deal of clerical effort before calculation starts.Clearly such work is not mathematical in nature. The main goal is to move, transform and put in ordervarious quantities of data so that calculations can be performed, results organized and analyzed, andreports written. In the remainder of this chapter we shall look at the capabilities of PCs to help in thiswork.2.2Communications — Access to Remote DataBecause workers must communicate with their colleagues, a large part of any researcher’s time is spentin written and oral communications.
Letters are giving way to facsimile transmissions (fax) and electronicmail (email) exchanges. Some workers are also using direct machine-to-machine linkages, either viainformal or organized attempts at group working (see for instance, the entire issue of the CommunicationsNext2: DATA PROCESSING CAPABILITIES OF PCs7of the ACM for December 1991).In sharing ideas, we must choose between direct interaction — the real-time, online, face-to-face ortelephone exchange — and indirect communication — letters, email, fax, bulletin board exchanges, voicemail, or similar offline (time-independent) information exchange. Our own preference is offline: we liketo read and sort messages at our convenience.
Call Waiting or mobile telephone service offer to keep usalways in touch. Following our policy throughout this book, we stress that each user of technology shouldcarefully analyze his or her own situation. PCs are a part of such communication choices.From our perspective, the important feature is connectivity — the fact that a PC can be linked to whateversource of data or information the user requires. There are some difficulties:•Physical obstacles exist in the form of incompatible hardware such as interface ports and connectorsat either end of the connection we wish to make, lack of suitable linkages such as transmission wires,or modems operating under different physical protocols (Derfler and Byrd, 1991).•Logical obstacles occur when we do not have suitable accounts on the email networks of interest, orthere is no gateway between our network and that of the person with whom we wish to exchangeinformation, or the information is stored or transmitted in incompatible formats.•Lack of training or understanding of the software and protocols may prevent us from establishing theconnection.
For example, we may not know how to log into a bulletin board, or if we do know, wemay not know the correct modem settings for the speed, parity and word size.Increasingly, users of PCs are connected to Local Area Networks (LANs). These should serve to simplifyand speed up the connections we are likely to make. However, they may inhibit the worker who needsto do something out of the ordinary. LANs are usually managed for the benefit of a perceived "majority"of users. As LANs get larger, or are linked together, Wide Area Networks (WANs) are beginning toappear.Increasingly, academic and other research environments are being linked together by special high-speednetworks.
These allow for rapid electronic mail, file transfer, and remote computer operation. There havebeen some unfortunate abuses of these systems in the form of unauthorized access and destructiveprograms (viruses, worms, etc.), but the benefits appear to outweigh the costs by a wide margin. Thenetworks now link most parts of the world and connections and email are increasingly reliable. This hasfostered the development of electronic newsletters and journals, special database servers to supplyparticular software or data, and a variety of electronic "conferences" (Krol, 1992; Stoll, 1990).For PC users, it is important to note that while the networks transmit binary data, there may be nodesbetween us and a source of information that insist on working with text information.
To allow the binarydata to cross such a node, it is common to encode the binary data as printable text characters. Thereceiving computer reverses the coding. Two common formats for such exchange are UUENCODE andBINHEX, for which programs may be found in a variety of public repositories. There are also ways totransmit binary data across modems directly (Nash J C, 1993a).2.3Automated Data CollectionResearchers are in the business of gathering data. Measurements used to be taken with simple instrumentsand recorded on paper, traditionally in bound, hard-cover laboratory notebooks.
Analysis was with penciland paper. With electronic technology has come a capacity for instruments to provide output directly indigital form, suitable for direct input to computers. There is an increasing use of computer technology forcontrol and calibration of these same instruments. PCs have frequently been used as both control anddata-gathering devices (e.g., Okamura and Aghai-Tabriz, 1985). Instruments with auto-answer modemsconnected to the telephone network allow remote measuring stations to transfer data to a central site. Wesuggest readers investigate the possibilities of automated instruments with care, since data collection isgenerally done in real-time and must necessarily interfere with conventional machine operations.8Copyright © 1984, 1994 J C & M M NashNash Information Services Inc., 1975 Bel Air Drive, Ottawa, ON K2C 0X1 CanadaSCIENTIFIC COMPUTING WITH PCsCopy for:Dr.
Dobb’s JournalOptical scanning equipment, though designed for optical character recognition and for the input of graphicmaterial, can provide data to be processed in a variety of ways. There are cameras for input of still orvideo images. Researchers studying sound or electro-magnetic signals have a variety of instruments forinputting raw data.
MIDI interfaces allow two-way transmission of musical information betweeninstruments, synthesizers and computers. We have used scanners with moderate success to process photosand other images for use in reports and also to convert a typescript of a century-old diary into machinereadable form.A variety of other interface mechanisms are used between computers and measurement devices.2.4Input, Output, Edit, and SelectionPCs are well suited to the task of entering, correcting and transforming data. It is relatively easy to writeparticular programs that provide a very friendly user interface yet carry out extensive error checking forwell-structured problems.
That is, if the problems are always of a specific type, a data entry program canbe prepared straightforwardly to allow non-specialists to enter the data.Preparation of a general data entry program is much more difficult. A priori we do not know the structure,type or precision of the data to be expected. Is it numeric or alphabetic? Whole numbers, fractions ormixed? Four digits or ten? How is each data element to be identified? Typically we, along with manyusers, employ a text editor and enter our data as simple lines of text information, though this allows noimmediate control on the elements we are keying in. Statisticians use the word "edit" for the process ofchecking data elements to see if they are "reasonable"; the word "imputation" is used for the substitutionof reasonable values for missing entries.Similarly, a general data transformation program is difficult to produce, but special ones are easilyimplemented on PCs.
Spreadsheet packages come close to being general data transformation programs.They have the great advantage that they are generally well known by non-programmers. They are easyto use for data entry. Editing individual entries to correct small errors is not difficult, but more globalediting, such as a repeated find and replace, may be more awkward. They also generally lack a way toprepare general "scripts" that can be applied to many worksheets (Nash J C, 1993b). Simple graphs canbe prepared easily. Most spreadsheet programs also incorporate some database management functions,but their strengths are the straightforward manipulation of tabular data.Most PC database management programs also offer some attractive features for data entry, edit andselection. In particular, we may quite easily "program in" checks for absurd data.
For example, atraditional concern with census data in the United States is the number of women who report multiplechildren and also report their age as under 14 (Sande I, 1982). Database management programs generallyrequire more effort to use, in our opinion, than spreadsheet programs, but they offer greater flexibility.2.5Scientific Text ProcessingSince the preparation of technical papers and reports is a primary data-processing task of PCs, it isimportant to consider the possible approaches we may take.
There are several special-purpose textprocessors designed to handle scientific material. In particular, the input and editing of mathematical and,to a lesser extent, chemical formulas are useful requirements. Such formulas are only slightly moredifficult to write down on paper than the prose that usually surrounds them. However, as far as we areaware, all computer text processing systems make formula entry much more difficult than text entry. Thisis because text is mostly linear in the way in which it is presented, that is, one word follows another, leftto right across the page in European languages. Formulas require us to recognize a variety of horizontaland vertical spacing rules, all of which can have very important meanings for the writer and reader.2: DATA PROCESSING CAPABILITIES OF PCs9It is our opinion that the learning cost is the greatest barrier to the use of special-purpose scientific textprocessors.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
