User-Extensible Simplification—Type-Based Optimizer Generators

Abstract

Abstract. For abstract data types (ADTs) there are many potential optimizations of code that current compilers are unable to perform. These optimizations either depend on the functional specification of the computational task performed through an ADT or on the semantics of the objects defined. In either case the abstract properties on which optimizations would have to be based cannot be automatically inferred by the compiler. In this paper our aim is to address this level-of-abstraction barrier by showing how a compiler can be organized so that it can make use of semantic information about an ADT at its natural abstract level, before type lowering, inlining, or other traditional compiler steps obliterate the chance. We present an extended case study of one component of a C++ compiler, the simplifier; discuss the design decisions of a new simplifier (simplifier generator) and its implementation in C++; and give performance measurements. The new simplifier is connected to the Gnu C++ compiler and currently performs optimizations at very high level in the front end. When tested with the Matrix Template Library, a library already highly fine-tuned by hand, we achieved run-time improvements of up to six percent.

A Language for Specifying Compiler Optimizations for Generic Software

“…The syntactic forms provided by the Pavilion language allow relatively sophisticated analyses and transformations to be written while further hiding the details of the internal program representation. In particular, generalized matching of expression trees provides much more power than is possible using literal tree pattern matching (such as that used in Simplicissimus [125]) or simple finite-state analysis of data flow (such as that used in Broadway [64,65]). Therefore, adding extra functionality to the Pavilion language beyond those previous systems will enhance the ability of users to easily write sophisticated optimizations for their abstractions.…”

“…A similar approach that can be implemented using the current C++ language is to provide traits classes that indicate which types model each concept [75,105,125]. Such an implementation simulates concepts using C++, and provides methods to use template metaprogramming [149] to determine whether a type models a particular concept.…”

“…The main work in concept-based optimization is Simplicissimus, which provides optimizations in the form of expression rewrites guarded by concept constraints on the involved types [125]. Simplicissimus is intended for algebraic simplification of generic programs, and so the restrictions on allowed optimizations are less of a problem in that domain.…”

Design of the CodeBoost transformation system for domain-specific optimisation of C++ programs

“…Simplicissimus [21,22] shares many of the same goals as CodeBoost. It is implemented as a plug-in to a compiler (currently GCC, but other compliers may be supported as well), and supports user-specified rewriting of expressions.…”

Generic programming and high-performance libraries

“…The author of LiDIA would therefore introduce the rewrite rule 1.0 / f → f.Inverse() whenever f is a LiDIA data type. Specializing rewrite rules such as these can improve performance particularly when multiple generic libraries are used in conjunction [31].…”