Swedish web site
General-Purpose computing on Graphics Processing Units (GPGPU): Fundamentals and Beyond2013HT
|
|
Course plan
No of lectures
This course will be offered in partnership with Udacity, one of the start-ups
providing so called Massive Open Online Courses or MOOCs. Essentially, we aim
to bring their course cs344 to our own graduate students. Instead of lectures,
there will be approximately 7 sessions of classroom discussions apart from
significant online material as self-study & assignments.
The url of the course from Udacity is: www.udacity.com/course/cs344
Recommended for
Students from all research areas and departments are welcome. Anyone who is interested in learning how to program a GPU (Graphics Processing Units) to accelerate non-graphics applications will find this course useful.
The course was last given
It is a new course.
Goals
The course introduces the student to the fundamentals of parallel computing from the perspective of General-Purpose computing on Graphics Processing Units (GPGPU). The topics covered are GPU programming model and architecture, key GPGPU algorithms and programming patterns, and optimization techniques. The student will complete the assignments with CUDA C/C++. Beyond this, students will also be introduced to latest research trends in GPGPU.
Prerequisites
Knowledge of fundamental data structures. Experience with C/C++ is highly
desirable.
Background in Graphics or Image Processing is NOT required.
Organization
Contents
Syllabus (Reference: www.udacity.com/course/cs344 )
Lesson 1: GPU Programming Model
Project 1: Greyscale Conversion
Lesson 2 - GPU Hardware and Parallel Communication
Project 2: Smart Blurring
Lesson 3 - Fundamental Parallel Algorithms
Project 3: HDR Tonemapping
Lesson 4 - Using Sort and Scan
Project 4: Red Eye Removal
Lesson 5 - Optimizing GPU Programs
Project 5: Accelerating Histograms
Lesson 6 - Parallel Computing Patterns
Project 6: Seamless Image Compositing
Lesson 7 - The Frontiers and Future of GPU Computing
Literature
Based on the progress of the students in each assignment, selected papers on GPGPU might be announced.
Lecturers
Examiner
Unmesh D. Bordoloi
Examination
Programming assignments, quizzes and home-works.
Credit
7.5
Comments
Page responsible: Anne Moe
