1
Basic and Historical Definitions

This chapter, introduction to the field, presents after the basic definitions, a history. In this context, the key characteristics of the microprocessors representative of these five decades are given.

1.1. Basic definitions

A microprocessor or MPU (MicroProcessor1 Unit) is the integrated version of the central processing unit2 (CPU) of computers from the 1960s, the prefix "micro" referring to miniaturization3. The central unit was made up of several electronic cards at the time where many integrated circuits in DIP (Dual-In-Line Package; cf. § 3.3 of Darche (2004)) were installed. These cards communicated with each other using a backplane bus (cf. § V2-4.2.7). The architecture of the microprocessor therefore inherits that of the central processing units of previous computers, themselves resulting from the von Neumann model (cf. § V1-3.2.2) for almost all. Of course, all of the current features, such as virtual memory (VM) or cache (this will be covered in a future book), were only gradually integrated due to the integration limitations of the technologies of the time.

This component is in the form of a single Integrated Circuit (IC) in monolithic version, originally belonging to the LSI category (Large-Scale Integration; cf. § V1-1.2) and developed in PMOS technology (positive (channel) metal-oxide semiconductor; cf. § 2.1.3 of Darche (2004)). Today, the microprocessor, essentially CMOS technology (Complementary MOS; cf. § 2.4 of Darche (2004)), offers parallel hardware execution. Exceptions included where the microprocessor was a set of integrated circuits, an example being the iAPX 432 (Witten and Cleary 1983) of the company Intel made up of three integrated circuits but that remained marginal.

From a logical point of view, a microprocessor is a sequential system (cf. Chapter 3 of Darche (2002)) programmable, synchronous4 most of the time, which executes a program, that is, an ordered sequence of instructions, stored in a so-called central, main or primary memory, volatile or not. We can see it as a complex form of state machine (FSM for Finite-State Machine; cf. § 3.7.3 of Darche (2002)).

Its basic instructions make it possible to carry out the logical and arithmetic processing of information and to carry out a transfer of control. To do this, it has three functions, calculation, storage and transfer of information.

When it integrates memory and Input/Output (I/O) controllers, it becomes a microcontroller or MCU (MicroController or MicroComputer Unit; cf. § 5.3) including the TMS 1000 from the company Texas Instruments (TI - 1974) was the first representative. The microprocessor can be general purpose (multipurpose or general-purpose MPU, GPP for General-Purpose Processor) or specialized in an application domain (ASP for Application-Specific Processor or ASIP for Application- Specific Instruction set Processor). The Digital Signal Processor (cf. 5.2) is a microprocessor specializing in digital signal processing, which is found, for example, in home video amplifiers. A subcategory of ASP is the bespoke processor (Moore 2017), a proposal by Cherupalli (2017) where the custom architecture is automatically generated to meet the needs of applications with high energy constraints such as the Internet of Things (IoT). The fields of use of the microprocessor are obviously IT, micro or classic, but industrial embedded systems are the real market. Since 2000, the trend has always been towards more integration. This has led to the emergence of systems-on-chip (SoCs, cf. § V1-1.2 and § V2-4.2.9), which involves the integration of an entire digital system, that is, multiple processors, memory and I/O controllers, into a single chip.

By extension, a processor is a reduction of a microprocessor in the sense that it contains at least one control unit, a calculation unit and registers. Today we also talk of core. The latter term appeared with the integration on the same chip of several cores in 2001 (multicore approach) with, today (2018), a dozen cores integrated on industrial chips. This trend was a response to the powerwall (cf. § V1-1.5), which materializes the physical limit.

1.2. History

The microprocessor was created in California in Silicon Valley. It was originally designed to meet the needs of a Japanese manufacturer of desktop calculators, the Nippon Calculating Machine Corporation. Intel5 announced it commercially in an advertisement in the November 15, 1971, issue of Electronic News. In fact, this component was part of a family of four integrated circuits called MCS-4 (Intel 1973), MCS for Micro Computer Set. It consisted of the ROM with I/O port of 2 Kib, the RAM with output port of 320 bits, the SIPO register (Serial-In Parallel-Out) of 10 bits and the CPU referenced respectively 4001, 4002, 4003 and 4004. Originally, this company required the design of 12 integrated circuits for the Busicom 141-P calculating machine (Figure 1.1(a)). Marcian6 E. Hoff, Jr., engineer in charge of the project at Intel, had the idea of integrating a programmable logic system. Subsequently, the rights were transferred to the company Intel for $60,000 at the time. The MPU included 2,300 transistors (Figure 1.1(b)), and it had a 4-bit architecture with single internal bus (cf. § V1-3.4.1). Ironically, the idea of the microprocessor was patented by the company Texas Instruments (Boone 1973). Additional historical information can be found in Noyce and Hoff (1981), Mazor (1995) and Halfhill (2006).

Before the microprocessor, there were several offers for the integration of logical subsets. Let us cite, in 1967, the company Fairchild which introduced a calculation unit in 8-bit format with an accumulator under the reference AL1. In 1969, the company Four Phased founded by Lee Boysel introduced a terminal using this component, perhaps the first bit-slice microprocessor (cf. § 5.1). In 1971, the offer revolved around families of components allowing realizing the function of MPU in discrete components. Let us take as an example of the MAPS family (Microprogrammable Arithmetic Processor System) from the company National Semiconductor (NS), which was a family of five components composed among others of an Arithmetic and Logic Unit (ALU) (MM5700), a register unit (MM5701), a control unit with a clock circuit (MM5702) and a control ROM (MM5705). In addition, industrialists working for the American military worked on this type of family long before. Let us cite the D200 from Autonetics/North American Aviation (Booher 1968; Shirriff 2016).

Figure 1.1. (a) The Busicom 141-P calculator and (b) photomicrograph of the 4004 (respective sources: Wikipedia.org and the Intel company). For a color version of this figure, see www.iste.co.uk/darche/microprocessor3.zip

The next generation was the 8008, a component originally designed for the Datapoint 2200 terminal from Computer Terminal Corporation. The latter did not use it for performance reasons. It should be noted that the first MPU was the TMX 1795 (X for eXperimental) from the company TI, prototyped but never sold. The reference patent is Boone (1973). It was designed to replace the CPU of the Datapoint 2200 terminal. In 1972, the 8008 was used by the French company R2E to manufacture the first7 microcomputer (MICRAL N released in 1973). This circuit manufactured in PMOS technology like its predecessor included 3,500 transistors. Its 8-bit architecture was single internal bus. The calculations despite its format were performed in n = 1 bit format. It was later followed by PPS-4 from Rockwell, PPS-25 from Fairchild, IMP-168 from NS and 5065 from Mostek.

To increase the speed of calculation, one approach has been to increase the format n of calculation. We also speak of bit-level parallelism. Thus, microprocessors can be classified into five generations by taking as a criterion the format of the integers that it handles. The first generation was that of 4 bits, then that of 8 bits, 16 bits and 8 bits improved, 32 bits and the fifth 64 bits. From the 16-bit generation, the components were called (at the time) "super microprocessor" (Vajda 1986) because they become able to execute programs written in High-Level programming Languages (HLL) and to support a multitasking Operating System (OS). Suzuoki et al. (1999) proposed a 128-bit version.

Table 1.1 presents two key characteristics of microprocessor development. The microprocessor now processes integers in 64-bit format when their format was originally 4 bits. The number of integrated transistors has increased from a few thousand to 10 billion. It thus doubled approximately every 18 months according to Moore's "law" allowing the integration of new functional units. Knowing that the acronyms CISC (Complex Instruction Set Computer) and RISC (Reduced Instruction Set Computer; this will be covered in a future book by the author on microprocessors) refer to models of ISA (Instruction Set Architecture, cf. § V1-3.5), it should be noted the downward difference in the number of transistors of this second architecture for a higher computing power. The economic aspect of the development of integrated circuits can be found in Moore (2003). In particular, it was found that the price of the transistor went from $1 in 1968 to $10-6 in 1998, a slope of about...