Hw & Exam


Graduate Computer Architecture

UC Berkeley CS252

Summer 2006


Midterm exam:

This is a take-home test, or maybe more like a mini-project that requires you to define a problem, conduct literature research, compare the solutions and think independently into the future.

Under the Papers link, every paper you read in this class is important in the history of computer architecture evolution, especially those that proposed a creative solution, which usually solved a bottleneck problem and became the standard in our computer design industry.

Pick a paper from our class collection, be prepared to read many papers and industry news in the related field. After your study, create a report that thoroughly analyzes this area. Note that you can also select a paper outside our class collection if interested or needed. Just explain in the report why you think this paper is important and deserves to be the subject of your study.

To start thinking:

  • What was the background of this paper in terms of computer architecture evolution? What was the problem this paper addressed (explain in brief words)? And why was the problem important?
  • Find 2~3 other papers that proposed different solutions to the same problem. Compare these solutions from all aspects you can think of – approach, performance, cost, compatibility, popularity, etc. Use a table or any other of presentation to organize your findings.
  • Among these solutions, which ones are more effective and what was the industry reaction? Were there industry designs that incorporated the architecture or techniques proposed by the papers you analyzed?
  • If the paper under study was written many years ago, what are the industry and academic trends today? Did the problem disappear or changed? Is there a new problem now? If there is, what are the efforts and results today trying to solve this new problem?
  • Besides these, use your own creativity to explore this topic and tell us your findings. For example, what do you think the computer will look like 10 years later, in terms of this part of design we have been discussing? Or what might have happened today if the previous solution was not developed?


  • Note that the above questions are not ‘one size fit all’, and it is NOT our objective to make you answer questions as in an exam. So do not turn in a list of answers. Instead it is about learning a topic of your own interest, conducting research on it and sharing the findings with your audiences. Try to tell a story that attracts and motivates the reader to learn with you through your study.
  • The report should be in an IEEE conference proceedings format, with up to 4 pages in length. Here is a sample manuscript.  You don’t have to stick to every detail of the manuscript formatting specifications, but just make sure that the report reads well.
  • Do not copy lines or paragraphs from the papers to make your report look like ‘long enough’. Write in your own words. It can be short but it has to be original and interesting.
  • For every paper or industry news cited, put a reference at the end of the paper – just like when you write a research paper. In this way you back up all your statements on solid ground.