You can join the Facebook group for this module here.
New: marks
For the Extended version of this module: Continuous assessment marks
Slides
Slides from the lectures.
lab exercises
New: Lab 4 challenge
Lab 3 on tiled matrix multiplication and profiling.
Lab exercise sheet 2. Please use lab LG04, where CUDA 5.5 has been installed.
Lab times are Thursday afternoon.
The demonstrator for this module is Asiri Rathnayake.
New: Peerwise homework 3
Marks are given up to a maximum of 4 points (= percent of module mark). For each new question on material from Week 5 onwards, 0, 1 or 2 points may be given, depending on quality. These marks are intended to be harder to achieve than the previous (participation) marks on Peerwise. For example, trivial questions about definitions of terms or the like will not receive any marks.
Deadline: 12 March, 12 noon. Late submissions are given 0 marks.
Homework 2 on Peerwise
New: Marks for Peerwise homework 1 in percent of final module mark, so 2 points = full marks
The second Peerwise homework is as follows.
- 2 points for authoring at least five questions (in total) on Peerwise relevant to the module content. A partial mark of 1 point will be given for authoring at least 3 questions in total.
- 1 point for making at least five comments.
- 1 point for an answer score on Peerwise of at least 400.
- Deadline: 12 noon, Monday 17 February.
- Weighting: 20% of continuous assessment mark = 4% of total module mark.
- Discretionary bonus points may be awarded for particularly good questions or comments (such as comments that correct mistakes in a question and/or help to improve it).
The more questions you create, answer, comment on, etc, the more you will learn and the more you will help other students on the module. If a good number of students contribute good questions, there will be plenty of revision material for everyone before the exams. You may also enjoy winning badges and trying to get on the leaderboards.
Here are some guidelines on creating good multiple-choice questions.
You need to register on the Peerwise server.
- The course ID for this module is 8487.
- Your "identifier" on Peerwise is the same as your Bham student ID number.
- You can choose any username and password when registering.
- Please tag your questions with appropriate topics, so that others can find them more easily.
Parallel Programming (Extended)
For the Extended version of this module, see here.
Textbook
David B. Kirk and Wen-mei W. Hwu: Programming Massively Parallel Processors: A Hands-on Approach (second edition), 2013. Available on Kindle.
Links to recommended reading
- Optimizing Parallel Reduction in CUDA
- Prefix Sums and Their Applications by Guy Blelloch
- Data Parallel Algorithms by Daniel Hillis and Guy Steele
- Parallel Prefix Sum (Scan) with CUDA by Mark Harris, nVidia
- Thrust from nVidia