COMPUTERS SYSTEM ORGANIZATION

DESIGN PRINCIPLE FOR MODERN COMPUTERS

Machine designers should always keep an eyes out for technological changes that may effect the balance among the component.

ALL INSTRUCTIONS ARE DIRECTLY EXECUTED BY HARDWARE

Eliminating a level of interpretation provides high speed for most instruction.

MAXIMIZE THE RATE AT WHICH INSTRUCTION ARE ISSUED

Modern computers resort to many tricks to maximize their performance, chief among which is trying to start as many instructions per second as possible.this principle suggest that parallelism can play a major role in improving performance, since issuing large numbers of slow instruction in short time interval is possible only if multiple instructions can execute at once.

INSTRUCTION SHOULD BE EASY TO DECODE

ONLY LOADS AND STORES SHOULD REFERENCE MEMORY

One of the simplest ways to break operation into separate steps is to require that operands for most instructions come from - and return to - CPU registers.
access memory will take a long time and the delay is unpredictable.
Conclusion, that  only LOAD and STORED instructions should reference memory. All other instructions should operate only on registers.

PROVIDE PLENTY OF REGISTERS

Instruction level- parallelism

Most computer architects look to parallelism (doing two or more thing at once) as a way to get even more performance for given clock speed. Parallelism comes in two general forms, instruction level parallelism and processor level parallelism.

PIPELINING

Pipelining is doing doing two thing at the same time. after one instruction is fetching, the other instruction will follow.

eg.

S1: Instruction fetch unit

S2: Instruction decode unit

S3: Operand fetch unit

S4: Instruction execution unit

S5: Write back unit

Explanation:

During clock cycle 1, stage S1 is working on instruction 1, fetching it from memory.

during clock cycle 2, stage S2 decode instruction 1, while stage S1 fetching instruction 2.

and so on..

This process will take a short time than wait one by one instruction to finish execute.

SUPERSCALAR ARCHITECTURES\

if one pipelining is good two pipelining are better. 

in reality, most functional units in stage 4 take appreciable longer than one clock cycle to execute, certainly the one that access memory or do floating point arithmetic.

MULTIPROCESSORS

the processing element in data parallel processor are not independent CPUs, since there only one control unit there. Multiprocessors is a system with more than one CPU sharing a common memory. They must coordinate to avoid getting in each other,s way. 

MULTICOMPUTERS

Multicomputer is a system consisting of large numbers of interconnected computers, each having its own private memory, but no common memory.

Since multiprocessor are easier to program and multicomputers are easier to build , there is much research on designing hybrid system that combine the good properties of each other.

amni dalilah(B031310284)