ECE 786 Topics in Advanced Computer Architecture: Data Parallel Processors

3 Credit Hours

In-depth study of processor architectures to exploit data-level parallelism, including general computation on graphics processing units (GPGPU, aka GPU computing architecture) and vector processors; memory subsystems; advantages and disadvantages of various architectures; technology shifts, trends, and constraints. Students undertake major course projects.

Prerequisite

Instruction set architecture, Pipelined processor architecture, caches, and C++ programming.

Course Requirements

The course requirement includes:

~4 Homeworks

~2 CUDA Programming Assignments

2 Exams (Midterm and Final)

1 Term Project

Course Topics

Introduction, Background, and Overview (1 week)

Vector Processors (1 week)

General Purpose Computation on Graphics Processing Units (GPGPU) aka SIMT Processors (7 weeks)

3.1 Programming Model & System Architecture (1 week)
3.2 Microarchitecture (3 weeks)
3.3 Memory Subsystem (1.5 week)
3.4 Performance Analysis (1.5 week)

Research Topics on GPUs and Vector Processors & Research Paper Presentation (5.5 weeks)

Midterm and Final Exams (1 week)

Textbook

None

Computer and Software Requirements

Please review the minimum computer specifications that NC State University and Engineering Online recommend.

Updated: 10/31/2022