Книга: Real-Time Concepts for Embedded Systems

1.1.5 Embedded Processor and Application Awareness

1.1.5 Embedded Processor and Application Awareness

The processors found in common personal computers (PC) are general-purpose or universal processors. They are complex in design because these processors provide a full scale of features and a wide spectrum of functionalities. They are designed to be suitable for a variety of applications. The systems using these universal processors are programmed with a multitude of applications. For example, modern processors have a built-in memory management unit (MMU) to provide memory protection and virtual memory for multitasking-capable, general-purpose operating systems. These universal processors have advanced cache logic. Many of these processors have a built-in math co-processor capable of performing fast floating-point operations. These processors provide interfaces to support a variety of external peripheral devices. These processors result in large power consumption, heat production, and size. The complexity means these processors are also expensive to fabricate. In the early days, embedded systems were commonly built using general-purpose processors.

Because of the quantum leap in advancements made in microprocessor technology in recent years, embedded systems are increasingly being built using embedded processors instead of general-purpose processors. These embedded processors are special-purpose processors designed for a specific class of applications. The key is application awareness, i.e., knowing the nature of the applications and meeting the requirement for those applications that it is designed to run.

One class of embedded processors focuses on size, power consumption, and price. Therefore, some embedded processors are limited in functionality, i.e., a processor is good enough for the class of applications for which it was designed but is likely inadequate for other classes of applications. This is one reason why many embedded processors do not have fast CPU speeds. For example, the processor chosen for a personal digital assistant (PDA) device does not have a floating-point co-processor because floating-point operations are either not needed or software emulation is sufficient. The processor might have a 16-bit addressing architecture instead of 32-bit, due to its limited memory storage capacity. It might have a 200MHz CPU speed because the majority of the applications are interactive and display-intensive, rather than computation-intensive. This class of embedded processors is small because the overall PDA device is slim and fits in the palm of your hand. The limited functionality means reduced power consumption and long-lasting battery life. The smaller size reduces the overall cost of processor fabrication.

On the other hand, another class of embedded processors focuses on performance. These embedded processors are powerful and packed with advanced chip-design technologies, such as advanced pipeline and parallel processing architecture. These processors are designed to satisfy those applications with intensive computing requirements not achievable with general-purpose processors. An emerging class of highly specialized and high-performance embedded processors includes network processors developed for the network equipment and telecommunications industry. Overall, system and application speeds are the main concerns.

Yet another class of embedded processors focuses on all four requirements-performance, size, power consumption, and price. Take, for example, the embedded digital signal processor (DSP) used in cell phones. Real-time voice communication involves digital signal processing and cannot tolerate delays. A DSP has specialized arithmetic units, optimized design in the memory, and addressing and bus architectures with multiprocessing capability that allow the DSP to perform complex calculations extremely fast in real time. A DSP outperforms a general-purpose processor running at the same clock speed many times over comes to digital signal processing. These reasons are why DSPs, instead of general-purpose processors, are chosen for cell phone designs. Even though DSPs are incredibly fast and powerful embedded processors, they are reasonably priced, which keeps the overall prices of cell phones competitive. The battery from which the DSP draws power lasts for hours and hours. A cell phone under $100 fits in half the palm-size of an average person at the time this book was written.

System-on-a-chip (SoC) processors are especially attractive for embedded systems. The SoC processor is comprised of a CPU core with built-in peripheral modules, such as a programmable general-purpose timer, programmable interrupt controller, DMA controller, and possibly Ethernet interfaces. Such a self-contained design allows these embedded processors to be used to build a variety of embedded applications without needing additional external peripheral devices, again reducing the overall cost and size of the final product.

Sometimes a gray area exists when using processor type to differentiate between embedded and non-embedded systems. It is worth noting that, in large-scale, high-performance embedded systems, the choice between embedded processors and universal microprocessors is a difficult one.

In high-end embedded systems, system performance in a predefined context outweighs power consumption and cost. The choice of a high-end, general purpose processor is as good as the choice of a high-end, specialized embedded processor in some designs. Therefore, using processor type alone to classify embedded systems may result in wrong classifications.

Оглавление книги