Home > Articles

Software

  • Print
  • + Share This
This chapter is from the book

Hardware Interfaces

Linking hardware to software has always been one of the biggest challenges facing those charged with designing computer systems—not just individual computers but whole families of computers, those able to run the same programs. The solution has almost invariably been to wed the two together by layering on software, so much so that the fastest processors struggle when confronted by relatively simple tasks. An almost unbelievable amount of computer power gets devoted to moving bytes from software to hardware through a maze of program code, all in the name of making the widest variety of hardware imaginable work with a universal (at least if Microsoft has its way) operating system.

The underlying problem is the same as in any mating ritual. Software is from Venus, and hardware is from Mars (or, to ruin the allusion for sake of accuracy, Vulcan). Software is the programmer's labor of love, an ephemeral spirit that can only be represented. Hardware is the physical reality, the stuff pounded out in Vulcan's forge—enduring, unchanging, and often priced like gold. (And, yes, for all you Trekkers out there, it is always logical.) Bringing the two together is a challenge that even self-help books would find hard to manage. Yet every computer not only faces that formidable task but also tackles it with aplomb, although maybe not as fast as you'd like.

Here's the challenge: In the basic computer, every instruction in a program gets targeted on the microprocessor. Consequently, the instructions can control only the microprocessor and don't themselves reach beyond. The circuitry of the rest of the computer and the peripherals connected to it all must get their commands and data relayed to them by the microprocessor. Somehow the microprocessor must be able to send signals to these devices. Today the pathway of command is rarely direct. Rather, the chain of command is a hierarchy, one that often takes on aspects of a bureaucracy.

Control Hierarchy

Perhaps the best way to get to know how software controls hardware is to look at how your system executes a simple command. Let's start with a common situation.

Because your patience is so sorely tested by reading the electronic version of this book using your Web browser, Internet Scapegoat, you decide to quit and go on to do something really useful, such as playing FreeCell. Your hand wraps around your mouse, you scoot it up to the big black x at the upper-right corner of the screen, and you click your left mouse button. Your hardware has made a link with software.

Your click is a physical act, one that closes the contacts of a switch inside the mouse, squirting a brief pulse of electricity through its circuitry. The mouse hardware reacts by sending a message out the mouse wire to your computer. The mouse port of your computer detects the message and warns your computer by sending a special attention signal called a hardware interrupt squarely at your microprocessor.

At the same time, the mouse driver has been counting pulses sent out by your mouse that indicate its motion. The mouse driver counts each pulse and puts the results into the memory of your computer. It uses these values to find where to put the mouse pointer on the screen.

The interrupt causes your computer to run a software interrupt routine contained in the mouse driver software. The driver, in turn, signals to Windows that you've clicked the button. Windows checks in memory for the value the mouse driver has stored there for the position of the mouse pointer. This value tells Windows where the mouse pointer is to determine how to react to your button press. When it discovers you've targeted the x, Windows sends a message to the program associated with it—in this case, your browser.

The browsers reacts, muttering to itself the digital equivalent of "He must be insane," and it immediately decides to pop up a dialog box that asks whether you really, really want to quit such a quality program as the browser. The dialog box routine is part of the program itself, but it builds the box and its contents from graphics subroutines that are part of the Windows operating system. The browser activates the subroutines through the Windows application interface.

Windows itself does not draw the box. Rather it acts as a translator, converting the box request into a series of commands to draw it. It then sends these commands to its graphics driver. The driver determines what commands to use so that the video board will understand what to do. The driver then passes those commands to another driver, the one associated with the video board.

The video board's driver routes the command to the proper hardware ports through which the video board accepts instructions. The driver sends a series of commands that causes the processor on the video board (a graphics accelerator) to compute where it must change pixels to make the lines constituting the box. Once the graphic accelerator finishes the computations, it changes the bytes that correspond to the areas on the screen where the line will appear in a special memory area on the video board called the frame buffer.

Another part of the video board, the rasterizer, scans through the frame buffer and sends the data it finds there to the port leading to your monitor, converting it into a serial data stream for the journey. Using synchronizing signals sent from the video board as a guide, the monitor takes the data from the data stream and illuminates the proper pixels on the screen to form the pattern of the dialog box.

When you awake from your boredom-inspired daze, you finally see the warning box pop up on the screen and react.

The journey is tortuous, but when all goes right, it takes a smaller fraction of a second than it does for you to become aware of what you've done. And when all doesn't go right, another tortuous chain of events will likely result, often one involving flying computers, picture windows, and shards of glass on the front lawn.

Keys to making this chain of command function are the application program interface, the driver software, and the device interfaces of the hardware itself. Sometimes a special part of your computer, the BIOS, also gets involved. Let's take a deeper look at each of these links.

Application Program Interface

A quick recap: An interface is where two distinct entities come together. The most important of the software interfaces in the Windows environment is the application program interface or API.

Rather than a physical thing, the API is a standard set of rules for exchanging commands and data. The Windows API comprises a set of word-like commands termed program calls. Each of these causes Windows to take a particular action. For example, the command DrawBox could tell Windows to draw a box on the screen, as in the preceding example.

To pass along data associated with the command—in the example, how large the box should be and where to put it on the screen—many calls require your program to send along parameters. Each parameter is ordinary data that is strictly formatted to meet the expectation of Windows. That is, the order of the parameters is predefined, and the range permitted for the data is similarly constrained.

Each software interface in the Windows system uses a similar system of calls and parameter passing. Gaining familiarity with the full repertory of the API is one of the biggest challenges facing programmers.

Device Drivers

Device drivers are matchmakers. A device driver takes a set of standardized commands from the operating system and matches them to the capabilities of the device that the driver controls. Typically the device that gets controlled is a piece of hardware, but, as our example shows, one driver may control another driver that controls the hardware.

Just about every class of peripheral has some special function shared with no other device. Printers need to switch ribbon colors; graphics boards need to put dots onscreen at high resolution; sound boards need to blast fortissimo arpeggios; video capture boards must grab frames; and mice have to do whatever mice do. Different manufacturers often have widely different ideas about the best way to handle even the most fundamental functions. No programmer or even collaborative program can ever hope to know all the possibilities. It's even unlikely that you could fit all the possibilities into an operating system written in code that would fit onto a stack of disks you could carry. There are just too many possibilities.

Drivers make the connection, translating generalized commands made for any hardware into those used by the specific device in your computer. Instead of packing every control or command you might potentially need, the driver implements only those that are appropriate for a specific type, brand, and model of product that you actually connect up to your computer. Without the driver, your operating system could not communicate with your computer.

Device drivers give you a further advantage. You can change them almost as often as you change your mind. If you discover a bug in one driver—say sending an uppercase F to your printer causes it to form-feed through a full ream of paper before coming to a panting stop—you can slide in an updated driver that fixes the problem. You don't have to replace the device or alter your software. In some cases, new drivers extend the features of your existing peripherals because programmer didn't have enough time or inspiration to add everything to the initial release.

The way you and your system handles drivers depends on your operating system. Older operating systems (such as DOS and old versions of Windows) load all their drivers when they start and stick with them all while you use your computer. Windows 95 and newer versions treat drivers dynamically, loading them only when they are needed. Not only does this design save memory, because you only need to load the drivers that are actually in use, it also lets you add and remove devices while you're using your computer. For example, when you plug in a USB scanner, Windows can determine what make and model of scanner you have and then load the driver appropriate for it.

Although today drivers load invisibly and automatically, things were not always so easy. The driver needs to know what system resources your hardware uses for its communications. These resources are the values you set through the Windows Add Hardware Wizard (click the Resources tab in Device Manager to see them). With old-fashioned drivers, hardware, and operating systems, you had to physically adjust settings on the hardware to assign resources and then configure the driver to match what you configured. It wasn't pretty, often wasn't easy, and was the most common reason people could not get hardware to work—usually because they created the problem themselves. Now Windows creates the problems automatically.

BIOS

The Basic Input/Output System, mentioned earlier as giving a personality to your computer, has many other functions. One of the original design intentions of the BIOS was to help match your computer's hardware to its software. To that end, the BIOS was meant to act as a special driver software included with your computer so that it can boot up right after you take it out of the box.

Part of the BIOS is program code for drivers that's permanently or semi-permanently recorded in special memory chips. The code acts like the hardware interface of an operating system but at a lower level—it is a hardware interface that's independent of the operating system.

Programs or operating systems send commands to the BIOS, and the BIOS sends out the instructions to the hardware using the proper resource values. It lies waiting in your computer, ready for use.

The original goal of the BIOS was to make computer software "platform independent," meaning that programs don't care what kind of computer they're running on. Although that seems a trivial concern in these days of dynamic-loading drivers, it was only a dream decades ago when computers were invented. In those dark-old days, programmers had to write commands aimed specifically at the hardware of the computer. Change the hardware—plug in a different printer—and the software wouldn't work. The BIOS, like today's driver software, was meant to wallpaper over the difference in hardware.

The idea, although a good one, didn't work. Programmers avoided the BIOS in older computers because it added an extra software layer that slowed things down. But the BIOS really fell from favor when Microsoft introduced modern Windows. The old hardware-based BIOS couldn't keep up with changes in technology. Moreover, using software drivers allowed hardware engineers to use more memory for the needed program code.

The BIOS persists in modern computers as a common means of accessing hardware before the operating system loads. The BIOS code of every computer today still includes the equivalent of driver software to handle accessing floppy disk drives, the keyboard, printers, video, and parallel and serial port operation.

Device Interfaces

Regardless of whether software uses device drivers, looks through the BIOS, or accesses hardware directly, the final link to hardware may be made in one of two ways, set by the hardware design: input/output mapping and memory mapping. Input/output mapping relies on sending instructions and data through ports. Memory mapping requires passing data through memory addresses. Ports and addresses are similar in concept but different in operation.

Input/Output Mapping

A port is an address but not a physical location. The port is a logical construct that operates as an addressing system separate from the address bus of the microprocessor, even though it uses the same address lines. If you imagine normal memory addresses as a set of pigeon holes for holding bytes, input/output ports act like a second set of pigeon holes on the other side of the room. To distinguish which set of holes to use, the microprocessor controls a flag signal on its bus called memory I/O. In one condition, it tells the rest of the computer that the signals on the address bus indicate a memory location; in its other state, the signals indicate an input/output port.

The microprocessor's internal mechanism for sending data to a port also differs from memory access. One instruction, move, allows the microprocessor to move bytes from any of its registers to any memory location. Some microprocessor operations can even be performed in immediate mode, directly on the values stored at memory locations.

Ports, however, use a pair of instructions: In to read from a port, and Out to write to a port. The values read can only be transferred into one specific register of the microprocessor (called the accumulator) and can only be written from that register. The accumulator has other functions as well. Immediate operations on values held at port locations is impossible, which means a value stored in a port cannot be changed by the microprocessor. It must load the port value into the accumulator, alter it, and then reload the new value back into the port.

Memory Mapping

The essence of memory mapping is sharing. The microprocessor and the hardware device it controls share access to a specific range of memory addresses. To send data to the device, your microprocessor simply moves the information into the memory locations exactly as if it were storing something for later recall. The hardware device can then read those same locations to obtain the data.

Memory-mapped devices, of course, need direct access to your computer's memory bus. Through this connection, they can gain speed and operate as fast as the memory system and its bus connection allow. In addition, the microprocessor can directly manipulate the data at the memory location used by the connection, thus eliminating the multistep load/change/reload process required by I/O mapping.

The most familiar memory-mapped device is your computer's display. Most graphic systems allow the microprocessor to directly address the frame buffer that holds the image that appears on your monitor screen. This design allows the video system to operate at the highest possible speed.

The addresses used for memory mapping must be off limits to the range in which the operating system loads your programs. If a program should transgress on the area used for the hardware connection, it can inadvertently change the data there—nearly always with bad results. Moreover, the addresses used by the interface cannot serve any other function, so they take away from the maximum memory addressable by a computer. Although such deductions are insignificant with today's computers, it was a significant shortcoming for old systems that were limited to a maximum of 1 to 16 megabytes. Because these shortcomings have diminished over the years, memory mapping has gained popularity in interfacing.

Addressing

To the microprocessor, the difference between ports and memory is one of perception: Memory is a direct extension of the chip. Ports are the external world. Writing to I/O ports is consequently more cumbersome and usually requires more time and microprocessor cycles.

I/O ports give the microprocessor and computer designer greater flexibility, and they give you a headache when you want to install multimedia accessories.

Implicit in the concept of addressing, whether memory or port addresses, is proper delivery. You expect a letter carrier to bring your mail to your address and not deliver it to someone else's mailbox. Similarly, computers and their software assume that deliveries of data and instructions will always go where they are supposed to. To ensure proper delivery, addresses must be correct and unambiguous. If someone types a wrong digit on a mailing label, it will likely get lost in the postal system.

In order to use port or memory addresses properly, your software needs to know the proper addresses used by your peripherals. Many hardware functions have fixed or standardized addresses that are the same in every computer. For example, the memory addresses used by video boards are standardized (at least in basic operating modes), and the ports used by most hard disk drives are similarly standardized. Programmers can write the addresses used by this fixed-address hardware into their programs and not worry whether their data will get where it's going.

The layered BIOS approach was originally designed to eliminate the need for writing explicit hardware addresses in programs. Drivers accomplish a similar function. They are written with the necessary hardware addresses built in.

Resource Allocation

The basic hardware devices got assigned addresses and memory ranges early in the history of the computer and for compatibility reasons have never changed. These fixed values include those of serial and parallel ports, keyboards, disk drives, and the frame buffer that stores the monitor image. Add-in devices and more recent enhancements to the traditional devices require their own assignments of system resources. Unfortunately, beyond the original hardware assignments there are no standards for the rest of the resources. Manufacturers consequently pick values of their own choices for new products. More often than you'd like, several products may use the same address values.

Manufacturers attempt to avoid conflicts by allowing a number of options for the addresses used by their equipment. In days gone by, you had to select among the choices offered by manufacturers using switches or jumpers. Modern expansion products still require resource allocation but use a software-controlled scheme called Plug-and-Play to set the values, usually without your intervention or knowledge.

Although a useful computer requires all three—software, hardware, and the glue to hold them together—hardware plays the enabling role. Without it, the computer doesn't exist. It is the computer. The rest of this book will examine the individual components of that hardware as well as related technologies that enable your computer to be the powerful tool it is.

  • + Share This
  • 🔖 Save To Your Account