Summary STUDY HARD

CS 405
COMPUTER SYSTEM ARCHITECTURE
MODULE-1 [PART-1]

Syllabus: Parallel computer models – Evolution of Computer Architecture,
System Attributes to performance, Amdahl's law for a fixed workload,
Multiprocessors and Multi computers, Multi vector and SIMD computers,
Architectural development tracks, Conditions of parallelism.




Evolution of Computer Architecture


Started with Von Neumann architecture

 Sequential machine

 Executing scalar data

 Slow

Major Advancement...

 Look ahead

 Parallelism and pipelining

 Flynn’s classification

 Parallel/vector computer



Look ahead technique


Introduced for instruction prefetching

 Prefetch the next instruction in advance while executing the current instruction.

To overlap I/E operations (Instruction fetch/decode and execution).



Enable functional parallelism

 Multiple functional units simultaneously

,  Pipelining at various processing levels



Pipeline


¡ Pipelining is an implementation technique where multiple instructions are overlapped in
execution.

¡ The computer pipeline is divided in stages.

¡ Each stage completes a part of an instruction in parallel.




Flynn’s classification


Classification based on

 Instruction stream

 Data stream

Classifications are

 SISD (Single Instruction Stream over a Single data Stream)

▪ Conventional sequential machines

 SIMD (Single Instruction Stream over Multiple data Stream)

▪ Vector computers with scalar and vector hardware

 MIMD (Multiple Instruction Stream over Multiple data Stream)

▪ Parallel computers

 MISD (Multiple Instruction Stream over a Single data Stream)

▪ Special purpose computers