1
The Memory Function

Memory is the second function of the computer, between computation and communication. Since the invention of the modern computer in the last century, memory has received the full attention of researchers, and later, of industry. It takes physical form as a component or a system, such as a memory subsystem or a mass storage device.

This chapter begins with a presentation of the characteristics of a generic memory. The most essential is likely its size, and the standardized multiples of its unit of measure are detailed in this context. The technological and historical aspects provide an opportunity to present the different memory families. These different memory families form a hierarchy in the computer, a concept that is then defined.

1.1. Main characteristics

Memory is mainly characterized by its medium, its total storage1 capacity C and its organization, the memory access method or policy, the type of access or operation, and the permanence (or non-volatility) of the information.

1.1.1. Memory medium

The type of medium characterizes the technology used to store the information. Information storage can be done through the presence or absence of a material (mechanical technology), such as a hole in a punched card (see section 5.6 of Darche (2003) and section 3.3 of Darche (2000)). Today, it involves a semiconductor (SC) or a magnetic, optical or magneto-optical material. In electronics, storage is done, for example, by moving or storing electric charges. In mass storage memory, it uses the change in orientation of a ferromagnetic material's magnetic field (originally iron oxide) on a flexible (floppy disk2 or FD) or rigid (hard disk or HD) support, or, as in rewritable optical memory, a phase change.

Table 1.1. Vocabulary used to describe a group of bits

French term English term Format n (bits) Chiffre binaire Bit (b) 1 Quartet ou demi-octet Nibble 4 Octet Bytea (B) 8 Mot Word
Dualoct 16 Double mot Double word
Quadlet quadoct 32 Quadruple mot Quad(ruple) word
Octlet (IEEE 1996)
Octbyte 64

a. A contraction of the English word "bite", meaning "a mouthful", created to follow the word "bit".

1.1.2. Storage capacity and units of measurement

The binary elements, 0 and 1, were named "bits" by statistician John Wilder Tukey, as a contraction of binary digit, although other terms such as bigit and binit were also suggested. The term was adopted by the community following its appearance in his famous article on information theory (Shannon 1948). The unit of measurement for memory capacity C is therefore the bit. Table 1.1 reminds us of the vocabulary used to describe a set of bits. In the era of 16-bit microprocessors (µP, microprocessor unit or MPU) in the 1980s, a word represented 16 bits. With the advent of 32-bit microprocessors, the term double word was introduced. Then came the next generation with the term quad word. But be careful - the word "word" can sometimes refer to any format depending on context, and not necessarily a multiple of the byte. We also speak of memory density, which refers to the number of bits stored per unit area (in2 or mm2).

Table 1.2. New standardized prefixes for capacity units

IEC symbols and prefixes Origin Factors Examples SI symbols and prefixes (reminder) Ki, kibi Kilobinary 210 (= 1 024) 1 Kib (formerly 1 KB) k (kilo) = 103 Mi, mebi Megabinary 220 (= 1 048 576) 1 Mib (formerly 1 Mb) M (mega) = 106 Gi, gibi Gigabinary 230 1 Gib (formerly 1 Gb) G (giga) = 109 Ti, tebi Terabinary 240 1 Tib (formerly 1 Tb) T (tera) = 1012 Pi, pebi Petabinary 250 1 Pib (formerly 1 Pb) P (peta) = 1015 Ei, exbi Exabinary 260 - E (exa) = 1018 Zi, zebi Zettabinary 270 - Z (zetta) = 1021 Yi, yobi Yottabinary 280 - Y (yotta) = 1024

Table 1.2 shows the symbol and prefix (or name) for the multiples of the bit. To distinguish the SI prefix kilo (10³) from the kilo used in memory sizing, the uppercase letter K is commonly used in computing. For example, we have 1 Kb (kilobit) meaning 1,024 bits and 1 KB (kilobyte) meaning 1,024 bytes. A technical term used is computer kilo. To eliminate ambiguity, the International Electrotechnical Commission (IEC), an international standards organization, approved a new naming convention for powers of 2 with the kilobinary symbol Ki, named kibi (IEC 2000). Self (1999) explains this aspect in detail. The Institute of Electrical and Electronics Engineers (IEEE), a US-based technical professional standards body, later standardized it (IEEE 2002a, 2002b).

Figure 1.1 shows a graphic reminder of these new measurement units for semiconductor memory and the relationships between them.

Figure 1.1. Base unit and standardized prefixes for a semiconductor memory

Currently, not all professionals use these two new standards, which causes some confusion. Manufacturers of semiconductor memory continue to use the old prefixes. This confusion persists with the sizes of hard disk mass memory units, where the ratio or factor between this unit and its multiples is equal to 103 × k (k ? N*). For examples and details, see section 7.2.2 and the associated exercise 7.3 of Darche (2003).

Figure 1.2 recalls the unit of measurement for mass memory capacity and the ratios between it and its multiples, specifying its first multiples. The shaded area defines a nonexistent capacity, since the first industrial disk, the IBM 350 disk storage unit (1956) by IBM, had a capacity of 3.75 MB (5 × 106 characters or 6-bit words). Its presence is justified only to recall the kilo ratio of 103.

Figure 1.2. Unit of measurement of mass memory capacity and its first multiples

And to be complete, operating systems (OS) convert mass storage capacity into a power of 2. The Linux OS even uses the new standardized prefixes. Universal serial bus (USB) keys, which use semiconductor memories, still follow powers of 2!

1.1.3. Organization

The memory cell is the smallest subdivision (atomic entity) of memory in which it is possible to read or write information. A cell or memory word has a format or width3 l or n or a size. We speak, for example, of the width of a register (concept detailed in section 3.1 of Darche (2021c)). This format mainly depends on the memory system format and the data bus width or mechanisms such as interleaving (see section 5.2.4). It can be a byte or a word multiple of the byte or another format (n = 9 bits, for example, in the case of error detection and correction, see section 6.4). Each bit bi of the cell of format n has a position i numbered, from right to left, from 0 to n-1 (i ? [0, n-1]; see Figure 1.3).

The organization refers to the physical arrangement of the cells in the memory. It allows us, from the total size, to specify the distribution between the format n of the memory cell and the number L of cells or depth D. We speak of input/output (I/O) organization length L × width n, for example, 16 Ki × 16 bits. In the case where there are several internal banks (see section 4.3), we speak of bank organization B × length L × width n, where B is the number of banks, for example, 8 × 2 Ki × 16 bits. The total capacity C of the memory is then equal to:

[1.1]

1.1.4. Access policies

The memory cell is called by Cragon the "memory-addressable unit" (AU) (Cragon 1996). It is the atomic element (i.e. the smallest addressable word) that the MPU can address in the main memory (or primary memory, acronym MP), i.e. a byte or a word of n bits.

Figure 1.3. Organization and addressing of a memory

The memory access method or policy specifies how the memory is accessed. In the case of location-based or coordinate addressing, the memory cell is accessed using an address (address-based storage and retrieval) which is an integer. This integer generally belongs to the subset of natural numbers, a counterexample being the address used to move to increasing or...