Computer Organization

for us in computer science, this class is a back stage tour of the place we work (the computer)
we'll study both the physical layout as well as the functional view
architecture can be applied in a variety of ways in the computer arena
- microarchitecture (chip design)
- system/computer architecture (chipset design)
- machine architecture (package design)
- software architecture (how large software systems are designed)
- enterprise architecture (how organizations establish a computing infrastructure)
from our (programmer's) vantage point, think in layers
- high-level language (e.g. java, c/c++)
- assembly language and compilers
- operating system
- instruction set architecture
- microarchitecture
- digital logic and circuits
- material science and signal processing
it can get complicated fast, but there is a relatively simple story at its heart

programming in the 1940s: switches and plugs
it was a glorified calculator that took a fleet of programmers to run
John von Neumann's idea: give the calculator memory and a vocabulary
the von Neumann architecture
measurements: milli-, nano-, Kilo-, Mega-, Giga-, Tera-, bps/baud, 48X, Hz
understanding parts and pieces somewhat follows a biological analogy
basic anatomy in seven easy pieces:
1) the central processor unit (CPU): the brain and an implied nervous system
2) short-term memory (RAM)
3) long-term memory (e.g. hard drives, flash, DVD, CD)
4) delegated processors (e.g. video cards, sound cards)
5) peripherals (e.g. mouse, keyboard, monitor)
6) power supply: the heart and an implied circulatory system
7) software (e.g. Windows/MacOS/Linux and browsers/spreadsheets)
connecting tissue: motherboard, bus, a gazillion other cables/wires, and a cabinet
basic processing (metabolic) pathways (consider how the anatomy is involved in each of these)
A) the instruction cycle: fetch > decode > execute > store [check out this simulator]
B) the storage game: concentric circles of memory
C) delegation and arbitration: layers of software, the virtual machine
D) first things: booting
E) doing something useful: starting and using an application

part of it is just so you know what's underneath you, your context
this knowledge is power: better performance and better diagnostics
understanding what the computer does with our programs can lead to code that is tuned to take advantage of what's underneath
it can also help us when things don't work
but the story from the system side of things is mostly about improving performance whether or not the programmer knows about it
consider how we've packed more stuff into the same square inch
the implication is we can do more per clock tick because we can do stuff in parallel
and so, it's not just about the clock frequency; it's about finding what we can do at the same time
by the way, some have a different view of this trend
there are limits to what can be done in parallel
consider Amdahl's Law for measuring the speedup by doing some things in parallel:
speedup = (run time on a single processor) / (run time on N processors) = 1 / [ (1 - f) + f / N ]
where f is the code (or work) that can run in parallel
so how do we know if we're computing faster?
one way is to measure how fast the computer runs instructions
this sometimes involves computing the Cycles Per Instruction or CPI
if you have this, you can figure out how many instructions per second the computer runs
measures such as MIPS and MFLOPS do this sort of measuring
however there is a bit of apples versus oranges problem with this
so some punt on the more pure approach and settle for empirical results via benchmarks