CS 5431

Advanced Computer Architecture

Fall 2024

Topics

1. Available Parallelism and Parallel Architectures

Functional Parallelism
Course grain parallelism
Fine grain parallelism

Data Parallelism
Flynn's classification
Review of data and function parallel architectures.
Exploiting Parallelism using replication
Exploiting Parallelism through pipelining.

2. Fundamental Issues in Parallel Processing

Synchronization
Communication
Data dependencies and their relation to fundamental issues
Limits of data dependencies
Limits of control dependencies

3. Exploiting Fine Grain Parallelism

VLIW and EPIC Architectures
Microprogramming

          Microprogramming 21MX Computers Part I
          Microprogramming 21MX Computers Part II
          Microprogramming 21MX Computers Part III       

VLIW and Compiling for VLIW Architectures
EPIC architecture

Superscalar Architectures
Processor states

Parallel Decoding

Parallel instruction issue
Shelving and dataflow
Alternative solutions to shelving

Parallel instruction execution

Preserving sequential consistency and exception behavior
Reorder buffer
Future file
Checkpoint recovery
Renaming and its interaction with state management

The fetch problem
Prediction and its limits
Predication and its limits

Memory dataflow
Speculation of dependencies through memory

Multi-Threading
Von-Neumann approaches to multi-threading
Dataflow approaches to multi-threading

4. Exploiting Coarse Grain Parallelism

Reference books :

1. Advanced Computer Architectures
A Design Space Approach
Dezso Sima, Terence Fountain and Peter Kacsuk

2. Superscalar Microprocessor Design
Mike Johnson

3.Processor Architecture From Dataflow to Superscalar and Beyond
  Kurij Silc
  Borut Robic
  Theo Ungerer

4. MODERN PROCESSOR DESIGN:
Fundamentals of Superscalar Processors, 
John Shen & Mikko Lipasti

Grading

• Term paper and Presentation 40 % 
  Midterm 1 30 % 
  Midterm 2 30 % 
  
  The final grade will be calculated using a sliding scale using the percentages above.

Collaboration Policy: 

• You are NOT allowed to work in groups. 

Exam policy:
I trust you will always work on your own during the exams. If you happen to study for the exam together with other people, make sure that you sit at the opposite corners of the classroom during the exam. 

Disclaimer: In the case of a conflict, information given out in class supersedes any contained on this page.

• Michigan Technological University complies with all federal and state laws and regulations regarding discrimination, including the Americans with Disabilities Act of 1990. If you have a disability and need a reasonable accommodation for equal access to education or services at Michigan Tech, please call the Dean of Students Office, at 487-2212. For other concerns about discrimination, you may contact your advisor, Chair/Dean of your academic unit or the Affirmative Programs Office, at 487-3310