CMSC 714 Midterm (Fall 2002)

(1) This exam is closed book, closed notes, and closed neighbor.

(2) You have 70 minutes to complete this exam. If you finish early, you may turn in your exam at the front of the room and leave. However if you finish during the last ten minutes of the exam please remain seated until the end of the exam so you don't disturb others.

(3) Write all answers in the supplied exam booklet. Start each new problem (but not sub-problem) on a new page.

(4) Partial credit will be given for most questions assuming I can figure out what you were doing.

(5) Please write neatly. Print your answers if your handwriting is hard to read. If you write something, and wish to cross it out, simply put an X through it.

1. (20 points) Define and explain the following terms:

A. Spin Lock

B. Directory based cache coherency

C. Reduction Operation

D. Data Decomposition

2. (20 points) The Earth Simulator employs parallelism on at least three levels (vector, SMP, cluster).

A. Explain the granularity of what is being done in parallel at each level.

B. Explain how it can use each of these levels as part of a single application (e.g., explain how the HPF climate code was parallelized).

3. (20 points) Consider a hardware counter that increments by one every cycle the processor is stalled waiting for memory operations, it interrupts the processor when it overflows, and it is writeable (but not readable). How could you implement the memory hierarchy metric of Mtool using this counter?

4. (20 points) Amdahl’s law provides a limit on the speedup possible by using parallel computing. Describe the limit implied by Amdahl’s law. People sometimes report speedups that would appear to exceed the limits of Amdahl’s law. Describe two different reasons such speedups might be seen and explain why the don’t violate Amdahl’s law.

5. (20 points) Conducting experimental measurements for papers is difficult. Explain three different flaws in the methodology from any of the papers read so far, and how they could have been corrected. You may use as many (or as few) as you like any one paper.