Embedded Files
CS510: Advanced Computer Architecture
CS510: Advanced Computer Architecture
Objective of the course:
Objective of the course:
A) Understand advanced topics in Computer architecture: Out of order architecture (prefetch, branch prediction, multi-issue, speculation, etc)
B) Understand memory hierarchy for advanced architecture - caches, and main memory
C) Understand GPU and machine learning architectures
Higher order objectives
D) Given an architecture, one should be able to analyse impact on hardware cost, performance, power, etc.
E) Given application characteristics, one should be able to design a new architecture, feature or modify an existing artefact.
F) Should be able to understand and criticise ISCA, Micro and HPCA papers.
Prerequisite:
Computer architecture (design or processor, pipeline, basic caches) - with any ISA
Syllabus
Syllabus
Module 1 Introduction: Introduction to advanced architecture
Module 2 Out of order architectures (chapter 2, 3 and 4 of text book)
Module 3 Memory system (chapter 7 and 10 of textbook)
Module 4 GPU and Machine learning accelerator architectures (chapter 6 and 14)
Logistics:
Logistics:
Class timings:
Class timings:
Monday 11 am to 11:50 am
Tuesday 11 am to 11:50 am
Wednesday 11 am to 11:50 am
TA:
TA:
To be announced
Office hours:
Office hours:
Send an email to get the appointment
Digital platforms
Digital platforms
Google classroom for assignment submission, discussions and distribution of notes/videos
This website for general information/announcements
Class Schedule
Class Schedule
Week 1 :Jan 6, 2025
Class starts on Jan 3
L1 (Jan 6) - No class
L2 (Jan 7) -No class
L3 (Jan 8) - No class)
L1 (Jan 10) - Introduction to course, advanced architecture
L2 (Jan 11) - Section 2.2.2 - In-order multiple issue, Section 2.3 overview of out of order, VLIW
Week 2 : Jan 13, 2025
L3 and L4 (Jan 14) - Branch prediction (section 3.1 to 3.3.4)
L5 (Jan 16) - Branch prediction (3.3.5 to 3.8)
Week 3 : Jan 20, 2025
L6 and L7 (Jan 21) - TAGE based Branch prediction -(lecture readings)
L8 (Jan 24) - Register renaming (Section 4.1 to 4.1.4)
Week 4: Jan 27, 2025
L9 , L10 - Jan 28
L11 - Jan 31
Week 5 : Feb 3, 2025
L12, L13: Feb 4
L14: Feb 7
Week 6 : Feb 10, 2025
L15, L16 (Feb 11)
L17 (Feb 14)
Week 7 : Feb 17, 2025
L18, L19
L20
L21
Week 8 : Feb 24, 2025
Mid term week
Week 9 : March 3, 2025
L22 (March 7) - GPU - Matrix Multiply
Week 10 : March 10, 2025
L23, L24 - ML accelerator
March 14 (Friday) - Holi - Public holiday
Week 11 : March 17, 2022
CCM II
L25 and L26 (March 18): ML accelerator
L27 - March 21 - reading 1 - presentation
Week 12 : March 24, 2025
L28 : March 25 - Reading 1 - discussion
L29: March 25 - Reading 2 - presentation
Friday - no class day
Week 13: March 31, 2025
L30 - April 1 - Reading 2 - discussion
L31 - April 1 - Reading 3 - presentation
L32 - April 4 - Reading 3 - Discussion
Week 14: April 7, 2025
April 8 (Friday timedtable) - Reading 4 - presentation
April 11 - Reading 4 - discussion
Week 15: April 14, 2025
April 15 : Reading 5 - presentation, discussion
April 16 - Reading 6 - presentation (Friday timetable)
April 18 - Holiday - Good Friday
Week 16: April 21, 2025
April 22 : Reading 6 discussion, Reading 7 presentation
April 25 - Reading 8 - presentation
Week 17: April 28, 2025
April 29 - Discussion - Reading 7 and 8.
Week 18: May 4, 2022
End term exam
Grading scheme
Grading scheme
Mid term (30%)
End term (30%)
Assignments (10%)
Two paper reading, presentations, peer assessment (20%)
Each student need to read, present atleast two papers from top conferences.
Class-participation/In-class Quizzes (10%)
75% of in-class quizzes will be considered
Alternate grading scheme:
Mid term (20%)
End term (20%)
Assignments (10%)
Project (20%)
Two paper reading, presentations, peer assessment (20%)
Each student need to read, present atleast two papers from top conferences.
Class-participation/In-class Quizzes (10%)
75% of in-class quizzes will be considered
Student can opt for any of these two schemes.
General guidelines:
General guidelines:
- All the submissions should be on time. Delayed submission will attract penalty at the rate of 10% per day. After delay of 10 days, no marks will be given to a assignment.
- Plagiarism of any sort (from internet resource or from friends) will be heavily penalised. In case, the case of copy is from other students of the course, both parties will get negative marks equal to double the weightage of assignment/submission. In second instance (across your studentship at IIT Ropar) will be given F.
Page updated
Google Sites
Report abuse