Design and implementation

Main article: CPU design

The basic concept of a CPU is as follows:

Hardwired into a CPU's design is a list of basic operations it can perform, called an instruction set. Such operations may include adding or subtracting two numbers, comparing numbers, or jumping to a different part of a program. Each of these basic operations is represented by a particular sequence of bits; this sequence is called the opcode for that particular operation. Sending a particular opcode to a CPU will cause it to perform the operation represented by that opcode. To execute an instruction in a computer program, the CPU uses the opcode for that instruction as well as its arguments (for instance the two numbers to be added, in the case of an addition operation). A computer program is therefore a sequence of instructions, with each instruction including an opcode and that operation's arguments.

The actual mathematical operation for each instruction is performed by a subunit of the CPU known as the arithmetic logic unit or ALU. In addition to using its ALU to perform operations, a CPU is also responsible for reading the next instruction from memory, reading data specified in arguments from memory, and writing results to memory.

In many CPU designs, an instruction set will clearly differentiate between operations that load data from memory, and those that perform math. In this case the data loaded from memory is stored in registers, and a mathematical operation takes no arguments but simply performs the math on the data in the registers and writes it to a new register, whose value a separate operation may then write to memory.

Control unit Main article: Control unit

The control unit of the CPU contains circuitry that uses electrical signals to direct the entire computer system to carry out stored program instructions. The control unit does not execute program instructions; rather, it directs other parts of the system to do so. The control unit must communicate with both the arithmetic/logic unit and memory.

Integer range

The way a CPU represents numbers is a design choice that affects the most basic ways in which the device functions. Some early digital computers used an electrical model of the common decimal (base ten) numeral system to represent numbers internally. A few other computers have used more exotic numeral systems like ternary (base three). Nearly all modern CPUs represent numbers in binary form, with each digit being represented by some two-valued physical quantity such as a "high" or "low" voltage.

MOS 6502 microprocessor in a dual in-line package, an extremely popular 8-bit design.

Related to number representation is the size and precision of numbers that a CPU can represent. In the case of a binary CPU, a bit refers to one significant place in the numbers a CPU deals with. The number of bits (or numeral places) a CPU uses to represent numbers is often called "word size", "bit width", "data path width", or "integer precision" when dealing with strictly integer numbers (as opposed to floating point). This number differs between architectures, and often within different parts of the very same CPU. For example, an 8-bit CPU deals with a range of numbers that can be represented by eight binary digits (each digit having two possible values), that is, 28 or 256 discrete numbers. In effect, integer size sets a hardware limit on the range of integers the software run by the CPU can utilize.

Integer range can also affect the number of locations in memory the CPU can address (locate). For example, if a binary CPU uses 32 bits to represent a memory address, and each memory address represents one octet (8 bits), the maximum quantity of memory that CPU can address is 232 octets, or 4 GiB. This is a very simple view of CPU address space, and many designs use more complex addressing methods like paging to locate more memory than their integer range would allow with a flat address space.

Higher levels of integer range require more structures to deal with the additional digits, and therefore more complexity, size, power usage, and general expense. It is not at all uncommon, therefore, to see 4- or 8-bit microcontrollers used in modern applications, even though CPUs with much higher range (such as 16, 32, 64, even 128-bit) are available. The simpler microcontrollers are usually cheaper, use less power, and therefore generate less heat, all of which can be major design considerations for electronic devices. However, in higher-end applications, the benefits afforded by the extra range (most often the additional address space) are more significant and often affect design choices. To gain some of the advantages afforded by both lower and higher bit lengths, many CPUs are designed with different bit widths for different portions of the device. For example, the IBM System/370 used a CPU that was primarily 32 bit, but it used 128-bit precision inside its floating point units to facilitate greater accuracy and range in floating point numbers. Many later CPU designs use similar mixed bit width, especially when the processor is meant for general-purpose usage where a reasonable balance of integer and floating point capability is required.

Clock rate Main article: Clock rate

The clock rate is the speed at which a microprocessor executes instructions. Every computer contains an internal clock that regulates the rate at which instructions are executed and synchronizes all the various computer components. The CPU requires a fixed number of clock ticks (or clock cycles) to execute each instruction. The faster the clock, the more instructions the CPU can execute per second.

Most CPUs, and indeed most sequential logic devices, are synchronous in nature. That is, they are designed and operate on assumptions about a synchronization signal. This signal, known as a clock signal, usually takes the form of a periodic square wave. By calculating the maximum time that electrical signals can move in various branches of a CPU's many circuits, the designers can select an appropriate period for the clock signal.

This period must be longer than the amount of time it takes for a signal to move, or propagate, in the worst-case scenario. In setting the clock period to a value well above the worst-case propagation delay, it is possible to design the entire CPU and the way it moves data around the "edges" of the rising and falling clock signal. This has the advantage of simplifying the CPU significantly, both from a design perspective and a component-count perspective. However, it also carries the disadvantage that the entire CPU must wait on its slowest elements, even though some portions of it are much faster. This limitation has largely been compensated for by various methods of increasing CPU parallelism. (see below)

However, architectural improvements alone do not solve all of the drawbacks of globally synchronous CPUs. For example, a clock signal is subject to the delays of any other electrical signal. Higher clock rates in increasingly complex CPUs make it more difficult to keep the clock signal in phase (synchronized) throughout the entire unit. This has led many modern CPUs to require multiple identical clock signals to be provided to avoid delaying a single signal significantly enough to cause the CPU to malfunction. Another major issue as clock rates increase dramatically is the amount of heat that is dissipated by the CPU. The constantly changing clock causes many components to switch regardless of whether they are being used at that time. In general, a component that is switching uses more energy than an element in a static state. Therefore, as clock rate increases, so does energy consumption, causing the CPU to require more heat dissipation in the form of CPU cooling solutions.

One method of dealing with the switching of unneeded components is called clock gating, which involves turning off the clock signal to unneeded components (effectively disabling them). However, this is often regarded as difficult to implement and therefore does not see common usage outside of very low-power designs. One notable late CPU design that uses clock gating to reduce the power requirements of the videogame console is that of the IBM PowerPC-based Xbox 360. It utilizes extensive clock gating in which it is used. Another method of addressing some of the problems with a global clock signal is the removal of the clock signal altogether. While removing the global clock signal makes the design process considerably more complex in many ways, asynchronous (or clockless) designs carry marked advantages in power consumption and heat dissipation in comparison with similar synchronous designs. While somewhat uncommon, entire asynchronous CPUs have been built without utilizing a global clock signal. Two notable examples of this are the ARM compliant AMULET and the MIPS R3000 compatible MiniMIPS. Rather than totally removing the clock signal, some CPU designs allow certain portions of the device to be asynchronous, such as using asynchronous ALUs in conjunction with superscalar pipelining to achieve some arithmetic performance gains. While it is not altogether clear whether totally asynchronous designs can perform at a comparable or better level than their synchronous counterparts, it is evident that they do at least excel in simpler math operations. This, combined with their excellent power consumption and heat dissipation properties, makes them very suitable for embedded computers.

Parallelism Main article: Parallel computing Model of a subscalar CPU. Notice that it takes fifteen cycles to complete three instructions.

The description of the basic operation of a CPU offered in the previous section describes the simplest form that a CPU can take. This type of CPU, usually referred to as subscalar, operates on and executes one instruction on one or two pieces of data at a time.

This process gives rise to an inherent inefficiency in subscalar CPUs. Since only one instruction is executed at a time, the entire CPU must wait for that instruction to complete before proceeding to the next instruction. As a result, the subscalar CPU gets "hung up" on instructions which take more than one clock cycle to complete execution. Even adding a second execution unit (see below) does not improve performance much; rather than one pathway being hung up, now two pathways are hung up and the number of unused transistors is increased. This design, wherein the CPU's execution resources can operate on only one instruction at a time, can only possibly reach scalar performance (one instruction per clock). However, the performance is nearly always subscalar (less than one instruction per cycle).

Attempts to achieve scalar and better performance have resulted in a variety of design methodologies that cause the CPU to behave less linearly and more in parallel. When referring to parallelism in CPUs, two terms are generally used to classify these design techniques. Instruction level parallelism (ILP) seeks to increase the rate at which instructions are executed within a CPU (that is, to increase the utilization of on-die execution resources), and thread level parallelism (TLP) purposes to increase the number of threads (effectively individual programs) that a CPU can execute simultaneously. Each methodology differs both in the ways in which they are implemented, as well as the relative effectiveness they afford in increasing the CPU's performance for an application.

Instruction level parallelism Main articles: Instruction pipelining and Superscalar Basic five-stage pipeline. In the best case scenario, this pipeline can sustain a completion rate of one instruction per cycle.

One of the simplest methods used to accomplish increased parallelism is to begin the first steps of instruction fetching and decoding before the prior instruction finishes executing. This is the simplest form of a technique known as instruction pipelining, and is utilized in almost all modern general-purpose CPUs. Pipelining allows more than one instruction to be executed at any given time by breaking down the execution pathway into discrete stages. This separation can be compared to an assembly line, in which an instruction is made more complete at each stage until it exits the execution pipeline and is retired.

Pipelining does, however, introduce the possibility for a situation where the result of the previous operation is needed to complete the next operation; a condition often termed data dependency conflict. To cope with this, additional care must be taken to check for these sorts of conditions and delay a portion of the instruction pipeline if this occurs. Naturally, accomplishing this requires additional circuitry, so pipelined processors are more complex than subscalar ones (though not very significantly so). A pipelined processor can become very nearly scalar, inhibited only by pipeline stalls (an instruction spending more than one clock cycle in a stage).

Simple superscalar pipeline. By fetching and dispatching two instructions at a time, a maximum of two instructions per cycle can be completed.

Further improvement upon the idea of instruction pipelining led to the development of a method that decreases the idle time of CPU components even further. Designs that are said to be superscalar include a long instruction pipeline and multiple identical execution units. In a superscalar pipeline, multiple instructions are read and passed to a dispatcher, which decides whether or not the instructions can be executed in parallel (simultaneously). If so they are dispatched to available execution units, resulting in the ability for several instructions to be executed simultaneously. In general, the more instructions a superscalar CPU is able to dispatch simultaneously to waiting execution units, the more instructions will be completed in a given cycle.

Most of the difficulty in the design of a superscalar CPU architecture lies in creating an effective dispatcher. The dispatcher needs to be able to quickly and correctly determine whether instructions can be executed in parallel, as well as dispatch them in such a way as to keep as many execution units busy as possible. This requires that the instruction pipeline is filled as often as possible and gives rise to the need in superscalar architectures for significant amounts of CPU cache. It also makes hazard-avoiding techniques like branch prediction, speculative execution, and out-of-order execution crucial to maintaining high levels of performance. By attempting to predict which branch (or path) a conditional instruction will take, the CPU can minimize the number of times that the entire pipeline must wait until a conditional instruction is completed. Speculative execution often provides modest performance increases by executing portions of code that may not be needed after a conditional operation completes. Out-of-order execution somewhat rearranges the order in which instructions are executed to reduce delays due to data dependencies. Also in case of Single Instructions Multiple Data — a case when a lot of data from the same type has to be processed, modern processors can disable parts of the pipeline so that when a single instruction is executed many times, the CPU skips the fetch and decode phases and thus greatly increases performance on certain occasions, especially in highly monotonous program engines such as video creation software and photo processing.

In the case where a portion of the CPU is superscalar and part is not, the part which is not suffers a performance penalty due to scheduling stalls. The Intel P5 Pentium had two superscalar ALUs which could accept one instruction per clock each, but its FPU could not accept one instruction per clock. Thus the P5 was integer superscalar but not floating point superscalar. Intel's successor to the P5 architecture, P6, added superscalar capabilities to its floating point features, and therefore afforded a significant increase in floating point instruction performance.

Both simple pipelining and superscalar design increase a CPU's ILP by allowing a single processor to complete execution of instructions at rates surpassing one instruction per cycle (IPC). Most modern CPU designs are at least somewhat superscalar, and nearly all general purpose CPUs designed in the last decade are superscalar. In later years some of the emphasis in designing high-ILP computers has been moved out of the CPU's hardware and into its software interface, or ISA. The strategy of the very long instruction word (VLIW) causes some ILP to become implied directly by the software, reducing the amount of work the CPU must perform to boost ILP and thereby reducing the design's complexity.

Thread-level parallelism

Another strategy of achieving performance is to execute multiple programs or threads in parallel. This area of research is known as parallel computing. In Flynn's taxonomy, this strategy is known as Multiple Instructions-Multiple Data or MIMD.

One technology used for this purpose was multiprocessing (MP). The initial flavor of this technology is known as symmetric multiprocessing (SMP), where a small number of CPUs share a coherent view of their memory system. In this scheme, each CPU has additional hardware to maintain a constantly up-to-date view of memory. By avoiding stale views of memory, the CPUs can cooperate on the same program and programs can migrate from one CPU to another. To increase the number of cooperating CPUs beyond a handful, schemes such as non-uniform memory access (NUMA) and directory-based coherence protocols were introduced in the 1990s. SMP systems are limited to a small number of CPUs while NUMA systems have been built with thousands of processors. Initially, multiprocessing was built using multiple discrete CPUs and boards to implement the interconnect between the processors. When the processors and their interconnect are all implemented on a single silicon chip, the technology is known as a multi-core processor.

It was later recognized that finer-grain parallelism existed with a single program. A single program might have several threads (or functions) that could be executed separately or in parallel. Some of the earliest examples of this technology implemented input/output processing such as direct memory access as a separate thread from the computation thread. A more general approach to this technology was introduced in the 1970s when systems were designed to run multiple computation threads in parallel. This technology is known as multi-threading (MT). This approach is considered more cost-effective than multiprocessing, as only a small number of components within a CPU is replicated to support MT as opposed to the entire CPU in the case of MP. In MT, the execution units and the memory system including the caches are shared among multiple threads. The downside of MT is that the hardware support for multithreading is more visible to software than that of MP and thus supervisor software like operating systems have to undergo larger changes to support MT. One type of MT that was implemented is known as block multithreading, where one thread is executed until it is stalled waiting for data to return from external memory. In this scheme, the CPU would then quickly switch to another thread which is ready to run, the switch often done in one CPU clock cycle, such as the UltraSPARC Technology. Another type of MT is known as simultaneous multithreading, where instructions of multiple threads are executed in parallel within one CPU clock cycle.

For several decades from the 1970s to early 2000s, the focus in designing high performance general purpose CPUs was largely on achieving high ILP through technologies such as pipelining, caches, superscalar execution, out-of-order execution, etc. This trend culminated in large, power-hungry CPUs such as the Intel Pentium 4. By the early 2000s, CPU designers were thwarted from achieving higher performance from ILP techniques due to the growing disparity between CPU operating frequencies and main memory operating frequencies as well as escalating CPU power dissipation owing to more esoteric ILP techniques.

CPU designers then borrowed ideas from commercial computing markets such as transaction processing, where the aggregate performance of multiple programs, also known as throughput computing, was more important than the performance of a single thread or program.

This reversal of emphasis is evidenced by the proliferation of dual and multiple core CMP (chip-level multiprocessing) designs and notably, Intel's newer designs resembling its less superscalar P6 architecture. Late designs in several processor families exhibit CMP, including the x86-64 Opteron and Athlon 64 X2, the SPARC UltraSPARC T1, IBM POWER4 and POWER5, as well as several video game console CPUs like the Xbox 360's triple-core PowerPC design, and the PS3's 7-core Cell microprocessor.

Data parallelism Main articles: Vector processor and SIMD

A less common but increasingly important paradigm of CPUs (and indeed, computing in general) deals with data parallelism. The processors discussed earlier are all referred to as some type of scalar device. As the name implies, vector processors deal with multiple pieces of data in the context of one instruction. This contrasts with scalar processors, which deal with one piece of data for every instruction. Using Flynn's taxonomy, these two schemes of dealing with data are generally referred to as SIMD (single instruction, multiple data) and SISD (single instruction, single data), respectively. The great utility in creating CPUs that deal with vectors of data lies in optimizing tasks that tend to require the same operation (for example, a sum or a dot product) to be performed on a large set of data. Some classic examples of these types of tasks are multimedia applications (images, video, and sound), as well as many types of scientific and engineering tasks. Whereas a scalar CPU must complete the entire process of fetching, decoding, and executing each instruction and value in a set of data, a vector CPU can perform a single operation on a comparatively large set of data with one instruction. Of course, this is only possible when the application tends to require many steps which apply one operation to a large set of data.

Most early vector CPUs, such as the Cray-1, were associated almost exclusively with scientific research and cryptography applications. However, as multimedia has largely shifted to digital media, the need for some form of SIMD in general-purpose CPUs has become significant. Shortly after inclusion of floating point execution units started to become commonplace in general-purpose processors, specifications for and implementations of SIMD execution units also began to appear for general-purpose CPUs. Some of these early SIMD specifications like HP's Multimedia Acceleration eXtensions (MAX) and Intel's MMX were integer-only. This proved to be a significant impediment for some software developers, since many of the applications that benefit from SIMD primarily deal with floating point numbers. Progressively, these early designs were refined and remade into some of the common, modern SIMD specifications, which are usually associated with one ISA. Some notable modern examples are Intel's SSE and the PowerPC-related AltiVec (also known as VMX).