(There is a lot of information in this paper. I suggest that you focus
on the first five sections of the paper, through Section 5.2. Section 2
is actually a good, concise summary of the ideas behind SMT. Starting
at 5.3, you can just skim the paper rather than read it in great
- Q1: What structures are dynamically shared among all the threads? Which, if any, of the structures are not dynamically shared (that is, replicated or statically partitoned amoung the threads)?
- Q2: What is "IQ Clogging"? What does this paper propose to address the issue?
- Q3: There is one important structure in the processor core that isn't discussed much in the paper. It turns out, this structure is difficult to dynamically share. What is the structure and why might it be difficult to share?