Nash - Scientific Computing with PCs, страница 12

With proper packaging and someluck in the postal system, they are a means by which data can be transferred from one machine to another.In contrast, the fixed (or "hard") disks cannot normally be exchanged from drive to drive but are sealedin a dust-proof case. Such systems offer compact, high volume storage on platters that have beendecreasing from 14 to 8 to 5 to 3.5 to 2 inches in diameter.

The major problem with such disks is backup,because the high volume of data requires a good deal of time for transfer to another medium. Indeed, asignificant part of the expense of any fixed disk system is a backup mechanism, such as tape cartridges.On mainframes, the reel-to-reel 0.5 inch tape was the primary removable data storage medium for manyyears. Drives are expensive because the sophisticated mechanical tape handling system must start and stopthe tape very rapidly without stretching it.

Someone had the good sense to recognize that the data couldbe recorded steadily at a slower speed if the controller could buffer the data. Such streaming tape drivescan be offered at quite modest prices when the same idea is applied to compact cartridge tapes.4: WILL A PC SUFFICE?29PCs use practically every imaginable printer. The most common type is the dot matrix variety. So-called"letter quality", daisywheel impact printers were initially the common method of obtaining output withan acceptable "business" look. These have almost disappeared, thanks first to competition from multi-strikedot matrix printers and later to laser or ink-jet printers. The latter achieve an amazing diversity of typefonts and graphic designs by clever use of software, either running in the PC or within the printer itself.(The printers are themselves often controlled by quite powerful microprocessors and memory.) Multicolorversions of various types of printers are challenging the pen-plotter role.In turn, pen plotters have achieved much of their development thanks to PCs.

Initially aimed at thespecialist technical market, they are now a tool for business. With the wider market base, prices havedropped considerably, but the popularity of pen-plotters is low compared to that of graphic printers. Animportant niche market for plotter technology is a cutter in which the pen of the plotter has is replacedby a knife.

This allows stencils or other masks to be cut, so that signs and displays may be created morequickly. Large plotters attached to PCs are also used industrially for preparing cutting plans, circuitdiagrams or engineering and architectural drawings.Visual displays for PCs and workstations continue to evolve. From the user point of view, it is importantto note that since the quality of the image displayed is dependent on the number of dots on the screenand the number of colors that can be displayed, a high-quality display demands much memory to storethe image. (This refers to raster displays, and does not apply to vector-graphic devices, which are muchless common.) If this memory is taken from the system’s main memory, there is less available for ourprograms.

If it is a separate memory, the speed with which data can be transferred to the video memorymay affect the speed of our whole machine. Indeed, the video memory of the machine used to write thismaterial is over 14 times "slower" than the rest of the memory. In serious computations, we must avoidwriting to the screen if we want our results quickly.There are many pointing devices for the interactive use of PCs. These include light pens, mice, trackballs,and even head motion pointers.Devices such as scanners and digital cameras are becoming common for input of graphic data.

Alsoavailable are interfaces to musical instruments or to sources of sound and video data, and laboratoryinterfaces for control of equipment or data acquisition.4.5Operational ConsiderationsThe PC has become such a common piece of equipment that one may forget that it is still a computer. Assuch it imposes various demands on its users for hardware, software and data maintenance. The hardwareof a PC is a lot easier to accommodate than that of a mainframe in that no raised floor or special powerare required. However, as we shall note in Chapter 11, the absence of specific requirements does not meanthat we should ignore common sense when choosing an operating environment.We have already noted that the software for a PC will require configuration depending on the particularperipherals to be attached. New or upgraded software may also mean that the configuration (systemgeneration) must be repeated.

This is particularly so for operating system upgrades. Worse, we must alsocheck all programs that directly use the operating system to ensure that they will still execute correctly.This can be sufficient reason for deciding to stick with an "old" version of the operating system,particularly if its deficiencies impose only minor inconvenience to our machine use. We think thatconfiguration is an onerous chore with which software developers burden the user. Several times we havehad to spend the better part of a day making configuration changes so new software would work.

Thisis time lost to scientific work.Finally, our own data must be maintained in good order and there is usually no one we can rely on todo this but ourselves. Thus, all backup and house-cleaning are the responsibility of the user, verydifferent from the automatic file maintenance of most centrally supported machines.Operationally, the pros and cons of PCs devolve from the power of decision granted to the user. We have30Copyright © 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 Journalcontrol of where and how the machine is to be used, and must assume sole responsibility for thosedecisions. In the rest of this chapter, we turn to strategies for coping with the difficulties that attend theuse of PC for scientific calculation. Readers can choose among the ideas discussed — the PC is a privatecalculating and data processing engine. Why heed the advice of others if we can discover ways to meetour goals in a manner convenient to our own habits and tastes? We hope the sections which follow aidthat discovery.4.6Issues to be AddressedIf we are fortunate enough to be able to choose between several computing platforms, then we first needto know which of them can carry out our tasks. This book deals directly with PCs, but the sameapproaches can be used for any computer.

We will want to know if an available computer can carry outour (scientific) work by means of suitable organization of the machine, the tasks, or both.The next seven chapters look at the practical, non-scientific details related to deciding if our PC can, orcan be made to, perform required tasks. Chapter 5 considers the management of data, of files and ofprograms, probably the thorniest problem for users of PCs. While not all scientific computing requires thatwe program, Chapter 6 discusses ways to make our efforts effective when it is required.

Since memorysize is critical to our ability to fit our problems in the PC, Chapter 7 considers this matter. If size questionscan be answered positively, we then need to decide if the speed of computation is sufficient, and timingis investigated in Chapter 8. The question of programming errors and their detection and repair, eitherin our own programs or in those we acquire, is the subject of Chapter 9.

Chapter 10 suggests some utilitysoftware for the support of our work. The physical care of PCs and their accessories is the topic of Chapter11, which also touches on configuration management.Following the housekeeping matters — that inhibit effective computing to an extent that we devote abouthalf this book to them — we turn to problems of a type likely to be encountered in scientific computationand some approaches we may make to their solution. The general steps in problem solving are the meatof Chapter 12.

Chapter 13 looks at the important matter of formulating problems in a manner suitable forcorrect and efficient solution on our PC. The rest of the book, apart from a final chapter on choosing one’sown strategy, consists of case studies in problem solving. Our goal there is to suggest ideas for thesolution of the user’s own problem — your problem. As a warning that there may be very greatdifferences in the time required to carry out scientific computations across PC hardware and software,Chapter 18 presents a study of the execution times for three almost identical algorithms in a selection oftheir implementations.

Chapter 19 shows some ways performance of different software may be comparedusing graphical data analysis tools. The same tools may be used in analyzing other sets of data.Since our message is that it is critical to try out ideas if there is a question of PC capability orperformance, much of the book mentions the kinds of tests and tools appropriate to answering suchquestions. We conclude in Chapter 20 by summarizing some of the ideas we have developed in a quartercentury of working with limited-capability computers.Previous Home Next5: FILE MANAGEMENT31Chapter 5File management5.15.25.35.45.55.65.7File creation and namingTypes and distributions of filesCataloguing and organizing filesBack-upArchivingVersion controlLoss versus securityThis chapter considers the thorny problem of file management - how to keep track of our program anddata resources.

We think this is the biggest problem for all computer users, no matter what the size oftheir machine. The task does not depend on the computer type or the application involved. The onlyareas in which PC practice is different from that of a mainframe are the mechanisms for file security andthe utilities available to aid the user.We note that we want:•To have data protected from destruction and loss (backup and archiving);•To be able to find data (i.e., files) easily (name recognition and avoiding duplicate names).In the following sections we will consider various aspects of file management, be they physical, logicalor human.

Some suggestions for coping with the problems will be made, but the major burdenunfortunately rests squarely on the shoulders of the user. Knowing what to do and how to do it countsfor nothing in file management — we still have to do it!5.1File Creation and NamingThe main unit of "permanent" memory on computers is the file. This is simply a sequence of words (bytes)of data that has a name and some access mechanism so that the computer can read the data. Presentationor use of the data is handled by programs. Data may represent counts, text characters, picture elements(pixels), program code in source, object or executable form, or variously coded forms of any of the aboveitems.A file created on one computer may not be usable on another, so it is important for the user to be awareof the compatibility of the file with his or her machinery.