CODE OPTIMIZATION IN MODERN COMPILERS

Krishna V. Palem (Courant Institute) and Vivek Sarkar (IBM)

The primary goal of this tutorial is to provide an overview of state-of-
the-art code optimization techniques used in compilers for modern
processors.  The performance gap between optimized and unoptimized code
continues to widen as modern processors have evolved to include hardware
features such as superscalar and pipelined functional units, and
sophisticated memory hierarchies.  Compile-time code optimization is
critical to automatically harnessing the high performance offered by
modern processors.

The tutorial is self-contained and begins with an overview of the design
of optimizing compilers.  Selected state-of-the-art optimizations such
as instruction scheduling, register allocation and optimizations for
memory hierarchies are then discussed in detail, using a quantitative
framework to explain the key principles.  Issues related to interactions
among individual optimizations are also discussed.  The optimizations
covered in this tutorial are most relevant for RISC processors, and
third-generation programming languages such as C and Fortran.