Computer Systems Architectures

Classically uniprocessor^[Uniprocessor architectures are characterised by having one ALU.] architectures include:
- [[Von Neumann Architecture]]
- [[Harvard Architecture]]
Multiprocessor (or parallel) architectures are potentially faster but harder to understand, control and predict.
- There is no single dominant parallel architecture, but a set of alternatives that depend upon different types of parallelisms (either on task or data level) described by the [[Flynn's Taxonomy]].
- There is a limit on how much a process can be parallelized described by the [[Amdahl's Law]].
There are different types of [[Memory Architectures | memory architectures]] — shared and distributed.
[[IO]].
[[Boolean Logic and Circuits]].

Operating Systems

Shared memory MIMD is becoming dominant in general purpose computing, due to the diminishing returns on building larger/faster uniprocessors.
Multi-core processors — multiple CPUs on the same chip — are now standard.
- They are typically used for multitasking.
- However, it can be challenging to design software that effectively utilises multithreading capabilities.

A numeration system is a writing system for expressing numbers.
The radix (or the base) determines the number of unique digits, including zero, that a numeration system uses to represent numbers.
To convert a number in base $n$ to decimal, sum up the products of each digit and $n$ to the power of its position, starting with 0, from left to right. For example: $$ 134_8 = 1 \times 8^2 + 3 \times 8^1 + 4 \times 8^0 = 92_{10} $$
To convert decimal to binary, divide the given number by two^[If the number is odd, first subtract one and then divide it by two.], recording each remainder. Then write down the remainders starting from the bottom; this is the result. For example, converting 167 to binary:

Starting from the bottom and omitting the preceding zero, the result is $167_{10} = 10100111_2$.

For reference, the first 16 powers of 2 are: $$ \begin{align*} 2^1 & = 2 \ 2^2 & = 4 \ 2^3 & = 8 \ 2^4 & = 16 \ 2^5 & = 32 \ 2^6 & = 64 \ 2^7 & = 128 \ 2^8 & = 256 \ 2^9 & = 512 \ 2^{10} & = 1024 \ 2^{11} & = 2048 \ 2^{12} & = 4096 \ 2^{13} & = 8192 \ 2^{14} & = 16384 \ 2^{15} & = 32768 \ 2^{16} &= 65536 \end{align*} $$

Principle of abstraction, used to build systems as layers
5 Classic components of a Computer (von Neumann)
Stored program concept: instructions just data (von Neumann)
Data can be anything (integers, floating point, characters): a program determines what it is
Principle of Locality, exploited via a memory hierarchy (cache)
Greater performance by exploiting parallelism

The elements of the numeration system are that: as many digit symbols as the base are needed; the decimal system’s base or radix of ten, the cardinal number of standards in the basic set, is denoted by ‘10’; place values increase from right to left in successive powers of the base (a positional system); addition is used to make up a number consisting of combinations of digits and multiplication is used to make up the number represented by a digit in a specific place; there is an agreed starting point (the ‘unit’ place); and a point is used to denote this place. ↩

Week	Presentations	Comments
01	4/4
02	2/2	C basics are noted very briefly [[./C Basics.md \| here]].
03	3/3
04	4/4	Note this¹ in [[Computer Systems Architectures.md]].
05	0/5	Data representation.
06	0/0	Consolidation week.
07	1/3	L01_P01 and P02 instructions.
08	3/3
09	2/3	L03_P01 interrupt driven IO and DMA IO.
10	3/3
11	3/3
19	3/3
20	1/2	C `malloc()` stuff, wasn't covered in the lecture, a recording should be posted.
21	2.5/2.5	Slides 1, 2, and the beginning of 3.
22	2/2	Slides 3 and 4.
23	2/2	Slides 5 and 6.
24	1/2	Slides 7 and 8.