Project 1: Research Project

• Pick a topic related to CPU design.
  
• Research the topic in the book, in the library, and on the net.  Write up a short topic paper, due February 24th, where you state what you want to talk about, give a brief outline of your talk, and give at least five useful references (websites, books, etc) on your topic.
  
• Present your results in class on March 3rd.  You'll be giving fairly short 25-minutes-plus-questions talks.
• You don't have to prepare powerpoint slides, but you MUST study hard, think beforehand about what you'll say, and have several interesting graphs/pictures/examples/demos.
  

Possible Topics (Choose one!)

• Hardware pipelining:
• Latency vs throughput
  
• Pipeline stages used by modern CPUs
  
• Operand forwarding
• Branch prediction, in theory or as implemented on modern CPUs.
  
• Data dependencies, the big limit on all parallel execution of serial code:
• Register renaming (i.e., Tomasulo's Algorithm), a way to remove 2 of the 3 types of dependencies.
• Superscalar execution:
• Functional units
• Speculation
• Hazard detection
• Cache design, cache hit rates, cache replacement algorithms, etc.
  
• Semiconductor fabrication technology: copper interconnects, deep-UV photolithography, or the recent 45nm process shrink (and upcoming 32nm and 22nm).
  
• Alternate programming models: Field-programmable gate arrays (FPGA), general-purpose Graphical Processing Units (GPGPU), etc.
  
• Alternate physical computers: relay computers, DNA computers, etc.
• Or, pick your own topic!  As long as it's architecture-related, it's good!