Lecture 8

Learning objectives

After this class, you should be able to:

1. Describe the following CUDA feature: cudaMemcpyToSymbol.
2. Use the above feature to write code that uses constant memory efficiently.

Reading assignment

1. Read UIUC Lecture 8 slides.
2. Chapter 8 of text, until (and including) section 8.3.

Exercises and review questions

• Exercises and review questions on current lecture's material
  1. Implement the 1-D convolution kernel and compare the performance with and without shared memory.
  2. Consider performing 1-D convolution on array {4, 1, 3, 2, 3} with with mask {2, 1, 4}. What is the resulting output array?
  3. How many halo cells are there when we perform 2-D convolution on an n x n matrix with an m x m mask?

• Preparation for the next lecture
  1. How can the algorithm discussed in class be improved?.

---

Last modified: 31 Jan 2013