370L12(2) - 12 A Gentle Introduction to Pipelining EECS 370 Introduction to Computer Organization Fall 2007 Prof Valeria Bertacco Prof Don Winsor

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon
The material in this presentation cannot be copied in any form without our written permission EECS 370 – Introduction to Computer Organization – Fall 2007 EECS Department University of Michigan in Ann Arbor, USA 12. A Gentle Introduction to Pipelining
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2/38 Single-cycle LC2Kx Datapath PC Instruction memory Register file Data memory Control ROM M U X M U X M U X M U X Sign extend + 1 + A L U 3x8 decoder R/W En En Instruction bits 15-0 21-19 18-16 24-22 18-16 2-0
Background image of page 2
EECS 370: Introduction to Computer Organization 3/38 The University of Michigan PC Memory Register file M U X M U X M U X Sign extend A L U R/W En En M U X M U X Instruction Reg Control Multi-cycle LC2Kx Datapath addr data En En 1 0 ALU result 2 cycles min. 5 cycles max.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
EECS 370: Introduction to Computer Organization 4/38 The University of Michigan Performance Metrics 1. Response time : when is my job done (time)? When will my books arrive from amazon.com? How long will this program/instruction take? 2. Throughput : how much work can get done within a specified time (work/time)? How many books will amazon.com sell this week? How many programs/instructions complete per hour? Improved relatively easily by using multiprocessors.
Background image of page 4
EECS 370: Introduction to Computer Organization 5/38 The University of Michigan Relating response time to throughput Number of users/programs/etc. Performance metric Throughput Response time 1 n worse better More throughput ! lower response time ( Little’s law ) Higher throughput is sending out more data in one cycle.
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
EECS 370: Introduction to Computer Organization 6/38 The University of Michigan Performance Metrics – execution time ! Response time for a program is its execution time Execution time (for an application): = total instructions executed x CPI x clock period ! CPI = avg number of clock cycles per instruction for an application ! For multi-cycle processor implementations we need: Cycles necessary for each type of instruction Mix of instructions executed in the application (dynamic instruction execution profile) Execution time formula is VERY IMPT! On a multi cycle data path, the CPI di ! ers per instruction.////On exam, we have to Fnd out the CPI ourselves.//// Nowadays, clock period has been maxed out because of heat. Ppl are working on minimizing CPI.
Background image of page 6
EECS 370: Introduction to Computer Organization 7/38 The University of Michigan How far have we come? ! Single-cycle processor implementations CPI = ? clock period = ?
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 8
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 04/02/2008 for the course EECS 370 taught by Professor Bertacco during the Winter '08 term at University of Michigan.

Page1 / 38

370L12(2) - 12 A Gentle Introduction to Pipelining EECS 370 Introduction to Computer Organization Fall 2007 Prof Valeria Bertacco Prof Don Winsor

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online