Nash - Scientific Computing with PCs (523165), страница 35
Текст из файла (страница 35)
To a large extent, thesedifferences show up in the configuration of the PC. In the configuration we include the set of physicaldevices that make up the PC system, the connections and interfaces to outside devices and networks, thephysical installation, the partitioning of disk and memory storage, the structure of directories on fixed diskstorage, the optional choices that are set for all operational hardware and software, and the selection ofand order of loading of any resident software.Many important aspects of the configuration of MS-DOS PCs are embedded in three particular places:•The continuous configuration memory (often called CMOS memory) of the computer that holdsphysical startup information;•The CONFIG.SYS file in the startup directory (root directory of the "bootup" disk) of the PC containsinstructions that set the maximum number of files allowed, the number of buffers, the partitioningof memory for different uses, and the loading of device drivers for special disks (includingcompression of the existing disk), tapes, keyboard features, or other tasks;•The AUTOEXEC.BAT file that is executed when the operating system command processor is activeand that loads memory resident software, sets the command environment, file path and commandprompts, and that launches programs we want active immediately, for example, a windowingenvironment, word processing shell, or integrated programming environment.Change to one configuration component may affect others, forcing us to test all regularly used softwareafter significant changes.
For example, during the writing of this book, we needed a mouse with a 9-pinconnector for our notebook computer. The only one available was on our 80386 tower machine, whichcould take either a 9-pin or 25-pin connector. We "borrowed" a different mouse from our old XT machinewhere it was little used, then had to change the configuration files of the tower machine to use the correctsoftware associated with this different mouse.
The AUTOEXEC.BAT file, Microsoft Windows and severalother programs had to be "set up" again. Unfortunately, after several uneventful weeks, we found theeditor BRIEF was unable to use the "new" mouse driver, and we had to reverse all our efforts.Updated programs require installation steps that also affect the configuration. Indeed, version numbersof programs in use form a part of the configuration information.
Unfortunately, upgrades that affectoperating software or features may make the user memory too small to run our favorite programs. Manypackages are now very memory-hungry. We have seen systems employing overlays appear to "run" aftera configuration change, then fail due to lack of memory when a menu choice is made. This is poorsoftware design, since the main routine should perform memory checks to allow us to avoid a "crash".Keeping a record of the configuration is essential if we need to restore a PC after a failure or disastroususer error, such as accidental reformatting of a fixed disk. Programs exist that can save the contents of theCMOS memory, though it is generally possible to edit this information at startup time if one has a listingof the appropriate settings.
Otherwise, if the PC can be started from a diskette, this is what we do. Datais then copied from the diskette to the fixed disk and CMOS. We also have to load backup software sowe can restore software and data from backup volumes.On 80386 and later MS-DOS machines, the configuration files, especially CONFIG.SYS, allow options thatcan markedly alter the amount of user-accessible memory for programs.
Setting the correct options formaximum memory is not a trivial task, especially if there are several special drivers for CD-ROM, networkor other functions as well as a set of resident programs to load. Overcoming perverse interactions betweendifferent functions we want to include can waste many hours of our time.A corollary to our configuration is the set of manuals describing the software and hardware we use.
Wehave three "collections": the hardware manuals, the software manuals we use regularly, and the less usedor obsolete software manuals that we may refer to occasionally. In the hardware manuals we try to keeppencil notes of switch settings or other important information. Discard manuals that you will never useagain.11: HARDWARE AND OPERATING PRACTICE101While we have talked here of MS-DOS PCs, Macintosh configurations have entirely equivalent functions,but the mechanisms are different.
Configuration information is held in the system folder. The user mayselect preferences for the operating environment; some of these may be changed for each folder. Selectedprograms may be made rapidly available for execution, and others may be made permanently active. Anoverly "busy" configuration is detrimental to performance. Serious pruning may give better performance.The configuration is important to how we work with our PC. Vanilla configurations can be very tiresome.A "travelling disk" of everyday utilities can save time when visiting colleagues.
We can set up a directoryfor temporary use and invoke familiar tools that aid productivity. Before leaving, the temporary directoryshould be purged.We always configure our machines with a RAM disk, since we find it important to have a fast area ofscratch space on "disk".
Print buffers, which are built into most networks, avoid waits for printing tocomplete. There are occasional difficulties with such buffers if a PC is interrupted in the middle ofprinting or a printer is not ready and data is lost.If we have more than one computer type, the potential configuration problems multiply, especially if wemust share data or peripherals between the different platforms.
With some care to standardize on diskformats, printer interfaces and characters sets (e.g., PostScript), we may be able to reduce the level ofannoyance. It is useful to localize certain tasks on one platform or another. In our own case, for example,we use an IBM mainframe almost exclusively for electronic mail, and IBM RS /6000 for Internetconnections, a Macintosh to help edit a professional journal and MS-DOS machines for everything else.Previous Home Next102Copyright © 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 JournalPart III:The problem and its solution methodThis part of the book presents our viewpoint on the actual solution of problems inscientific computation with PCs.
In previous parts of the book we have dealt with themachinery and the software, with their attendant difficulties and details. Now we willlook at problems and how their solution. First we will look at formulation, then considersome cases, simplified for expository purposes. Throughout we stress empiricism — ifsomething works, use it! If it doesn’t, find out why.Chapter 12Steps in Problem Solving12.112.212.312.412.512.612.712.812.9Problem formulation exampleAlgorithm choiceAlgorithm implementation — programmingDocumentationTesting — data entryTesting — sub-programsTesting — complete programsProduction runsModifications and improvementsFaced with a computational problem, one explicitly or implicitly follows a sequence of steps in passingfrom the statement of the problem to the number, set of numbers, formula or graph that provides asolution.This chapter outlines these steps and partially illustrates them with an example of linear least squaresfitting of data to a model.
This problem arises throughout the physical, biological and social sciences, inapplications ranging from chemical kinetics to real estate valuation. Readers with experience in leastsquares modelling may disagree with some choices made at various phases of the problem solution.Let us consider the steps in developing a solution. The steps are not necessarily sequential, though somenaturally precede others.1.2.3.4.5.Problem formulation or statement;Mathematical formulation or restatement;Choice of technique or method;Choice of algorithm;Selection of computing environment, namely computing system and programming language orapplication package;12: STEPS IN PROBLEM SOLVING6.7.8.9.10.11.12.103Coding of program;Testing of program;Documentation of program;Revision and modification, with retesting;Further documentation of program;Regular or "production" use of program.Repeat as necessary!12.1 Problem Formulation ExampleThere is rarely a unique formulation to a problem.
Without considerable practical experience, we areunlikely to have a good appreciation of the relative merits of different formulations. A correct formulationof a problem is vital to the choice and development of a satisfactory solution method. Moreover, theproblem statement may not show the formulation. For example, we could be asked:Find the relationship between the earnings of computer programmers and their formal education andexperience.This problem presupposes a relationship, but does not give a form to it.
Thus, our first choice is the formof the model. We will simply propose a model in the form:(12.1.1)Earnings = r * Education + s * Experience + tThis form, a linear model, is well-known and widely used. Before, we can apply it however, we needappropriate data for the variables Earnings, Education and Experience.Earnings could be measured by gross annual salaries, Experience by years of paid employment as aprogrammer.
For Education we propose years of post-secondary education in a computer-related subject,though this ignores other post-secondary studies that may have some impact on earnings. Our task istherefore to estimate r, s, and t, the three linear coefficients in this model. In certain circumstances we mayalso be required to provide measures of precision of these coefficients.Mathematically, it is convenient to rewrite the problem in a more compact way.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
