CS 553 Midterm Review 2014
--------------------------

----------------
Outline

Data-flow analysis

    - be able to do liveness analysis
    
    - use the results of liveness analysis to do register allocation

    - difference between a must versus a may analysis


Lattice Theoretic Framework

    - Terminology like top, bottom, meet, and under.
    
    - Tuples of lattices and how to use it for reaching constants
      and available expressions in LLVM.
      
    - What is the relationship between a bitvector tuple of lattices and 
      a powerset lattice?

Dominators

    - What are the dominators and what is the dominance
      frontier for every node in the control-flow graph?
      
    - What is the relationship between dominance frontiers and SSA?


Data dependencies

    - Flow/True, Anti, and Output dependencies

    - Distance vectors

    - Direction vectors
    
    - How to determine distance or direction vectors for a single loop example.

----------------
Review Questions from reading assignments

    - What are the three main components in the typical compiler design?
    
    - What is a peephole optimization?  What are some example peephole optimizations?
    
    - Does register allocation in LLVM occur on the SSA code or after the code has been taken out of SSA form?
    
    - What NP-Complete problem is register allocation typically reduced to?
    
    - What are some ways that virtual registers can be prioritized/ordered for register allocation heuristics?
    
    - What are three special characteristics of partial order relations?
    
    - There were a number of different program transformations described in the Padua and Wolfe paper.  What were the three main goals these transformations targeted?
    
    - Define lexicographic ordering.