Nash - Scientific Computing with PCs, страница 11

An example concerns communication ports. On our own machines, we have had to limit theways in which we use Microsoft Windows because there are not enough IRQ lines to fully supportinternal modems that are set up as the third serial communication port on two of our machines.Because PC systems have fewer discrete components than mainframes, they ought to have an intrinsicallyhigher reliability. Counterbalancing this is the fact that they tend to be used in less than idealenvironments — unfiltered electrical power, dust, heat or cold, humidity and general rough handling alladd to the potential sources of faults. However, the use of interchangeable circuit boards and/or socketedcomponents makes servicing usually quite easy.In summary, PC users must carefully assess the processing, disk and memory resources available forproblem solving rather than the raw or total resources implied by the PC hardware.4.2SoftwareThe software for PCs is, we believe, a much greater obstacle to their use in scientific problem solving thanthe hardware.

We may arbitrarily divide software into the following four categories:•The operating system, which enables us to execute all other software and to send data to or receiveit from peripherals;•Utilities, which allow the user to do certain housekeeping functions such as copying or listing files;•Programming language translators, to compile or interpret programs written in other than machinecode;•Application programs, that is, all the rest.We will look at these categories in reverse order. Programs for a given application need not be drasticallydifferent just because they are to be run on a PC. Primarily, we want to be sure our task is feasible interms of the actual working memory or disk space available within the application program.

Morepositively, PC application software may access low-level machine functions for speed. Also, since we arethe manager of the PC, we can also choose to let inefficient programs run for hours or days at a time ifthis is the least-cost method of achieving our goals. On a practical level, however, we need to know that4: WILL A PC SUFFICE?27the execution time to get a task done is not so long that it will inhibit use of the PC.We have already seen in Section 2.9 that practically all main stream programming languages can be usedwith PCs.It is usually in the area of housekeeping support that PC systems are at a disadvantage.

At nearly allmainframe or workstation sites there are systems programmers to help handle housekeeping tasks. OnPCs, utility programs that are not provided in or with the operating system must be acquired separatelyby the user. There is no shortage of choice of tools, as almost any PC trade magazine will illustrate. In fact,the wide selection is a source of difficulty, since the user must choose between several attractive choices.However, with each selection costing money, buying each set of utilities will draw resources away fromthe scientific software budget.

Freeware or public-domain utility programs exist, but the chore ofevaluation requires much time. Since we are usually dealing with programs that move, modify or deletefiles, program errors can have serious consequences, so untested utilities are worse than none at all.Operating system software controls all the activity in the PC. Since many PC systems are configured witha unique combination of processor, memory and peripherals, the user must usually make at least minormodifications to the operating system to allow the processor to use the resources available to it.Configuration management (Section 11.8) extends to most major applications packages such as wordprocessors, desktop publishing software, CAD programs, compilers, file managers, or drawing programs.When we add a new resource (printer, plotter, scanner, mouse, video adapter, network adapter) to oursystem, we have to adjust all affected software.

An advantage of the newer operating environments suchas Microsoft Windows or IBM OS/2 for Intel-based PCs, or Finder for Apple Macintosh, is that theoperating software is nominally the only place where such adjustments are made.With increased memory sizes in PCs, operating system (OS) software has tended to become bloated.Sometimes users have had to reinstall "old" versions of their OS because critical applications programswould no longer operate under the "new" system. Our view is that operating software should be as smallas possible.

First, the operating system takes time to load; as each program terminates, the commandprocessor is generally reloaded to memory, with load time proportional to its size. Extra OS features thatwe do not use still require some computing effort to bypass and the memory and disk they occupy canbe better used for other purposes. Also, the larger the code, the easier it is for accidental errors or virusesto corrupt it.

Finally, the more features there are, the larger the manual through which we have to sift tofind a particular item of detailed information.4.3Interfacing, Communications and Data SharingPCs are well suited to being connected to other computers or data handling devices. PCs can bereconfigured to a specific task, whereas a mainframe must usually serve many users and cannot beradically altered to suit just one application. Moreover, some components of a PC system areprogrammable; that is, their properties may be changed by sending special control instructions. Anobvious example is a dot-matrix or laser printer to which font descriptions can be down-loaded.The disadvantages are that the user may have to do the work of interfacing or pay quite heavy fees tohave someone else do it. On the hardware side, simple cables may cost over $200 for a custom-made itemthat has parts worth only a few dollars.

The most commonly needed cables needed cost less than $20.Miniaturization has presented some scientifically trivial but administratively unpleasant obstacles suchas connectors that do not match. For example, the serial connector on the original IBM PC is a male DB25 plug.

This was and is a commonly available component. To reduce size, and noting that not all of the25 pins were used, manufacturers sometimes switched to a DB 9 plug. An added level of confusion existedwith early IBM PC-compatibles by using a DB 25 female connector for the serial port. Gender changersand adapters are fortunately available inexpensively from many suppliers, though we have on occasionsbeen told by ignorant employees of some stores that "that kind isn’t made."28Copyright © 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 JournalAnother problem is that the transmit and receive lines may need to be swapped using a device called anull modem. To simplify our own work, especially in travelling, we have found it useful to have a cablewith both male and female DB 25 connectors at each end of a 6 to 8 foot length of ribbon cable.

This, witha null modem and some gender changers and 9—pin to 25—pin adapters allows us to make neededconnections. Once we have established the right connections, a simpler cable can be obtained. We havealso found the cables supplied with the Laplink file transfer software to be very useful.PCs handle data communications quite easily. The main mechanisms commercially available are eitherLocal Area Network (LAN) products or modems designed for use with the standard telephone network.We will not deal here with special equipment for use with private telephone lines or satellite channels.There are so many forms of LAN hardware that we will not attempt more than a cursory description ofthe possibilities. The important aspects are that simple cabling — either coaxial cable or telephone-likemultiple wire cable — allows very high rates of data movement between computers.

The applications ofsuch high-speed transmission are that a user can work with data on another computer as if it were presenton the local fixed disk. Software operating the LAN may have strong influences on PC configuration,performance and user practices.Modems, though much slower, also allow for the transport of data between machines. Ironically, this isoften via the intermediary of a mainframe computer. We use this technique to make class notes availableto students via a shared disk area on a university mainframe. Such mainframes are commonly nodes ofthe international academic and commercial networks, and provide access to many electronic mail and filetransfer facilities (see Krol, 1992).4.4PeripheralsPractically every type of input, output or storage device can be attached to PCs, though certain types arefavored. The main mass storage mechanism for PCs has evolved from the flexible disk to the high-capacityfixed disk.

Flexible (or floppy) disks can store data in a bewildering variety of sizes, densities, recordingmechanisms and formats. We have found that it is helpful to describe these by the type of machine,capacity, and physical disk size rather than an adjective preceding the word "density". We tend to use IBMPC 360KB 5.25 inch disks for distribution of the relatively small files associated with our own work andIBM PC/Macintosh 1.44 MB 3.5 inch disks for archiving data, for the work of individual projects andwhen travelling.

Apple chose a variable speed drive for its computers so that more data can be packedinto the outer tracks of the diskette surface than the inner ones. This contrasts with fixed rotation speedsin the MS-DOS world, but many Macintosh machines now read and write the 1.44 MB fixed-speed format.Capacities have been evolving in both Macintosh and PC families of 3.5 inch disks. This can be a nuisancefor users, but manufacturers seem aware of the need for backward capability so "old" diskettes can stillbe read. Despite the difficulties of standardization, flexible disks are cheap, quite reliable, and offer asatisfactory method of storing and moving data and programs for PCs.