Lecture 11

Learning objectives

After this class, you should be able to:

1. Profile CUDA code using the CUDA Visual Profiler.
2. Determine the occupancy of CUDA code using the GPU occupancy calculator.
3. Describe the tools available to profile code on the Cell processor.

Reading assignment

1. Refer to the CUDA Visual Profiler and GPU Occupancy Calculator documents on blackboard.

Exercises and review questions

• Exercises and review questions on current lecture's material

  1. Profile both your matrix multiplication codes of assignment 2, and determine the reason for the difference in performance.

  2. Use the GPU occupancy calculator to determine the occupancy of the above code.

• Preparation for the next lecture
  1. None.

---

Last modified: 15 Feb 2010