Specialized Processor Extensions

In the preceding chapters, we discussed the features of general-purpose computer architectures and some architectural specializations intended to address domain-specific requirements. This chapter focuses on extensions commonly implemented at the processor instruction set level and in other computer system hardware to provide additional system capabilities beyond generic computing needs.

After reading this chapter, you will understand the purpose of privileged processor modes and how they operate in multiprocessing and multiuser contexts. You will be familiar with the concepts of floating-point processing units and instruction sets, techniques for power management in battery-powered devices, and processor and system features that enhance system security.

We will discuss the following processor extensions and system features in this chapter:

• Privileged processor modes
• Floating-point mathematics
• Power management
• System...

Technical requirements

The files for this chapter, including solutions to the exercises, are available at https://github.com/PacktPublishing/Modern-Computer-Architecture-and-Organization-Second-Edition.

Privileged processor modes

Most operating systems running on 32-bit and 64-bit processors control access to system resources using the concept of privilege levels. The primary reasons for managing access in this manner are to enhance system stability, prevent unauthorized interactions with system hardware, and prevent unauthorized access to data.

Privileged execution improves system stability by ensuring only trusted code is allowed to execute instructions that provide unrestricted access to resources such as processor configuration registers and I/O devices. The operating system kernel and related modules, including device drivers, require privileged access to perform their functions. Because any crash of a kernel process or a device driver is likely to halt the entire system immediately, these software components generally undergo a careful design process and rigorous testing before being released for general use.

Running the operating system in a privileged context prevents...

Floating-point arithmetic

Modern processors usually support integer data types in widths of 8, 16, 32, and 64 bits. Some smaller embedded processors may not directly support 64-bit or even 32-bit integers, while more sophisticated devices may support 128-bit integers. Integer data types are appropriate for use in a wide range of applications, but many areas of computing, particularly in the fields of science, engineering, and navigation, require the ability to represent fractional numbers with a high degree of accuracy.

As a simple example of the limitations of integer mathematics, suppose you need to divide 5 by 3. On a computer restricted to using integers, you can perform an integer calculation of this expression as follows, in C++:

#include <iostream>
int main(void)
{
    int a = 5;
    int b = 3;
    int c = a / b;
    std::cout << "c = " << c << std::endl;
    return 0;
}

This program produces the following output:

c = 1

...

Power management

For users of portable battery-powered devices such as smartphones, tablets, and laptop computers, the ability to operate for long time periods without recharging is an important feature. Designers of portable systems place great emphasis on ensuring battery power consumption is minimized under all operating conditions.

Some techniques designers use to reduce power consumption are:

• Placing idle subsystems in a low-power state or turning them off completely when they are not needed. This technique may not be possible for peripherals that must be available to respond to incoming requests, such as network interfaces.
• Reducing integrated circuit supply voltages and clock speeds during periods when execution speed is not critical.
• When possible, saving system state information and turning the processor power off. Users of laptop computers are familiar with the two options for reducing power consumption when the system is not in use: standby and...

System security management

We have seen how the separation of privilege levels between kernel and user modes supports the effective separation of applications started by one user from those of other users and from system processes. This represents security at the level of executing software.

This is fine as far as it goes, but what about systems that must remain secure even when untrusted users have unrestricted physical access to them? Additional measures must be implemented at the hardware level to prevent curious or malicious users from accessing protected code, data, and hardware resources.

Before getting into the details of hardware-level security features, it is helpful to list some of the categories of information and other resources that must be protected in digital systems:

• Personal information: Information such as government identification numbers, passwords for accessing bank accounts, contact lists, emails, and text messages must be protected even if...

Summary

Building upon the preceding chapters, this chapter introduced computer architecture features addressing domain-specific functional requirements. We focused on extensions commonly implemented at the processor instruction set level and in components external to the processor that provide additional system capabilities beyond generic computing requirements.

You should now have a good understanding of privileged processor modes and how they are used in multiprocessing and multiuser contexts, the concepts of floating-point processors and instruction sets, techniques for power management in battery-powered devices, and the processor and system features intended to enhance computer system security.

This background prepares us for the next chapter, where we will examine the most popular processor architectures and instruction sets currently used in personal computing, business computing, and smart portable devices. These architectures are the x86, the x64, and the 32-bit and...

Exercises

Using a programming language that allows access to the byte representation of floating-point data types (such as C or C++), write a function that accepts a 32-bit single-precision value as input. Extract the sign, exponent, and mantissa from the bytes of the floating-point value and display them. Remove the bias term from the exponent before displaying its value and display the mantissa as a decimal number. Test the program with the values 0, -0, 1, -1, 6.674e-11, 1.0e38, 1.0e39, 1.0e-38, and 1.0e-39. The numeric values listed here containing e are using the C/C++ text representation of floating-point numbers. For example, 6.674e-11 means 6.674 x 10^-11.

1. Modify the program from Exercise 1 to also accept a double-precision floating-point value, and print the sign, exponent (with the bias removed), and mantissa from the value. Test it with the same input values as in Exercise 1, and with the values 1.0e308, 1.0e309, 1.0e-308, and 1.0e-309.
2. Search the internet...