Download Syllabus Schedule for Computer Microarchitecture | ECE 512 and more Study notes Computer Architecture and Organization in PDF only on Docsity!
Lecture 1: Intro to the course
Objectives of Lecture
• Overview • Logistics • Workload • Why performance? • How much performance?
Who Should Take this Course
-^
Anyone who is currently a research student in architecture and who has not yet beenexposed to the material.
-^
Anyone who thinks they want to be in category 1.
-^
Anyone who wants to be an architect of a future generation chip.
-^
I would discourage people who are casually interested because the class is beyond full.
-^
Prerequisite Knowledge
Basic CMOS circuits
Basic compiler optimization and code generation
Architecture: Tomasulo’s Algorithm, 2-level branch prediction, caching, pipelining,
Logistics
• Topical overview
– 6 major sections, 3 processor case studies – Heavy on the state of the art, some on the future
• Lectures
– Two days a week, discussion oriented, centered
around the readings
• Readings
– Should do readings prior to corresponding lecture
Why?
• Why Performance? • Why General-Purpose? • What are the constraints?
Advanced Computing Systems Group© 2003 Sanjay J. Patel
A decade of Intel x86 performance
10 1 100 Sep- Jan- Jun- Oct- Mar- Jul- Dec- Apr- Sep- Jan- Relative Processor Performance Based on SPEC 100 10 1 Relative FO4 Inverter Speed Pentium Pentium Pro Pentium II Pentium III Pentium IV FO4 Inverter Perform ance 70x performance in 10 years 800nm 600nm 350nm 250nm 180nm 130nm Relative Speed of FO4 Inverter
A Decade of x
• Pentium had 3.1M Transistors • Pentium 4 has 45 M • Memory speeds got better too: but not as quickly
Assignment
• What are the factors that contribute to the
remaining 10x after we account for the speed ofthe faster devices?
• Why does a microarchitectural family increase in
performance even within the same processtechnology?
• Who are you? Write a 1-2 paragraph bio like
you would find in an IEEE Journal.