Lecture 14
Learning objectives
After this class, you should be able to:
- Given data and an associative operator, show the steps involved in the work-efficient parallel prefix algorithm.
- Discuss the issue of thread divergence in the work-efficient algorithm.
Reading assignment
- UIUC Lecture 14. (Skip inclusive scan in the slides, but read the idea behind a different approach to it in fig. 9.3 of the text.)
- Chapter 9 of text, sections 9.4 - 9.6.
Exercises and review questions
- Exercises and review questions on current lecture's material
- Can you redesign the work-efficient algorithm so that in the reduction phase, even-indexed locations are updated?
- Preparation for the next lecture
- None.
Last modified: 24 Feb 2015