This course will focus on compilation techniques for optimizing programs for single core and multicore processors. We will begin with classic topics such as intermediate program representations, interprocedural and intraprocedural dataflow analysis, register allocation, and scheduling for single core processors. Then we will focus on topics pertaining to optimizing parallel applications running on multicore processors. We will study compiler optimization techniques to improve the performance of regular and irregular programs on multicore processors. We will conclude with a discussion of current research directions in compilers, e.g., autotuning, iterative compilation, and intelligent compilers.

A previous course in compilers (e.g., CISC672) is useful, although not required. Students should have a familiarity with C, C++, or Java and the ability to work with large applications in one of these languages. A significant portion of the course will be a large project which includes substantial software development, as well as a written project report. Students will also present a paper (which they choose) based on a relevant topic to the course. There is no required book for the course. Instead, topics will be synthesized from a variety of sources, including papers, information on the internet, and sections of books *all* on reserve at the library.