AI & MLRecommendedAIML-SYS-015
Accelerators and the Memory Hierarchy
Why GPUs are fast and when they are not: compute-bound versus memory-bound operations and arithmetic intensity.
- Difficulty
- 4 / 5
- Importance
- 4 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can place an operation on a roofline plot from its arithmetic intensity and decide whether it is compute- or memory-bound.