Rules

1. Due next Mon 2021-02-08, noon.

2. Submit the answers to Gradescope.

3. You may do homeworks in teams of 2 students. Create a gradescope team and make one submission with both names.

4. Short but informative answers are fine.

Questions

Each is 5 points.

1. Why is one CUDA core less powerful than one Intel Xeon core?

2. If 10% of a program cannot be parallelized, then what is the max speedup even with infinite parallelism?

3. "For short running parallel programs, there can actually be a decrease in performance compared to a similar serial implementation.". Why?

4. What is the difference between strong scaling and weak scaling?

5. What is cache coherency?

6. Define "embarrassingly parallel".

7. How many CUDA cores does the RTX 8000 have?

8. Why have machine cycle speeds stopped increasing?

9. Give an advantage and a disadvantage of shared memory versus distributed memory for parallel computation.

10. In OpenMP, what's the difference between ATOMIC and CRITICAL?

Total: 50 pts.