Permission is granted for noncommercial reproduction of the work for educational or research purposes.

Heterogeneous parallel systems incorporate different models of parallelism within a single machine or across machines and are well suited for many diverse applications [1, 2, 3]. These systems offer a wealth of underutilized resources for achieving high performance, but are more difficult to program than homogeneous systems. Compilers can and should assist in handling this complexity. Previous research has not considered the impact of architectural heterogeneity on the compiler structure [4]. A compiler for heterogeneous systems needs to order transformations based on component architectures, represent diverse architectural features, allow extension to cover new machines, and reuse its transformations and intermediate representation (IR) to target different architectures. A compiler's IR plays a central role in supporting these capabilities. This paper discusses each of these implications in detail and presents an intermediate representation, Score, that meets the needs of heterogeneit...

In modern optimizing compilers, linear human-readable text representation of a program is first transformed into an abstract syntax tree that represents the structure of that program. Abstract syntax tree is then transformed into intermediate representation (IR), based on which compiler optimizations are accomplished. The optimized IR is sent to the code generator and finally translated into assembly or machine code. Research on IRs has been focused on how they can be designed to facilitate compiler optimizations or more effective code generation on specific architecture. This thesis presents a mid-level intermediate language, called Floey. In a Floey program, control flowgraphs are separated into different tree-like structures called control expressions. Different control expressions are connected by entries. On Floey, a machine independent optimization, called the reduction algorithm, is implemented. By comparing the reduction algorithm to various conventional optimizations, we argue that not only Floey facilities compiler optimization design, it also provides a cleaner and uniform perspective on compiler optimizations in general.