Lecture 19
Learning objectives
After this class, you should be able to:
- Explain the place of a GPU in the PC architecture.
- Explain how threads are scheduled on a GPU.
- Given an application with parameters specifying granularity (such as tile size in matrix multiplication), decide on a suitable value of the parameter based on constraints, such as the number of threads that can run concurrently and number of blocks that can run concurrently.
Reading assignment
- GPU-5, sections 5.1 and 5.3 on Blackboard, under the "course library" tab.
Exercises and review questions
- Exercises and review questions on current lecture's material
- Determine the right number of threads per block for your group project based on the criteria discussed in class. Post your answer on the discussion board.
- Preparation for the next lecture
- How much shared memory and register is available per SM on gpu.cs.fsu.edu?
Last modified: 7 Apr 2010