Effective partial redundancy elimination

Abstract

Partial redundancy elimination is a code optimization with a long history of literature and implementation. In practice, its effectiveness depends on issues of naming and code shape. This paper shows that a combination of global reassociation and global value numbering can increase the effectiveness of partial redundancy elimination. By imposing a discipline on the choice of names and the shape of expressions, we are able to expose more redundancies, As part of the work, we introduce a new algorithm for global reassociation of expressions. It uses global information to reorder expressions, creating opportunities for other optimization. The new algorithm generalizes earlier work that ordered FORTRAN array address expressions to improve optimization~25].

Property-Driven Benchmark Generation

“…We therefore employ code motion techniques [8,10,11,20,12,9,21] for de-localizing the information and therefore require a global analysis for reconstructing the transition relation.…”

A Fresh Look at PRE as a Maximum Flow Problem

“…In Step 6, we find a unique minimum cut E −1 (C Λ ) on G ss by applying the "Reverse" Labelling Procedure of [10], where Λ is the smallest possible (Lemma 8). Figure 1(d) depicts G ss for our example, together with the following minimum cut: (2,5), (3,5), (4,6), (7,9), (8, 10)} C ins Λ = {(4, 6), (7, 9)} C copy…”

“…All optimal algorithms must find one way or another the unique minimum cut on a flow network G ss derived from a CFG. Classic PRE has been extended to perform other important optimisations, including strength reduction [8,12,16,18], global value numbering [3], live-range determination [21], code size reduction [24], redundant load/store elimination [21] and data speculation [20]. Its scope has also been extended by means of code restructuring [2].…”

“…As a result, PRE encompasses both global common subexpression elimination and loopinvariant code motion. Over the years, PRE has also been extended to perform other optimisations at the same time, including strength reduction [8,12,16,18], global value numbering [3] and live-range determination [21]. For these reasons, PRE is regarded as one of the most important optimisations in optimising compilers.…”

A Framework for Exploring Optimization Properties

“…First, optimizations may degrade performance in certain circumstances. For example, Briggs and Cooper reported improvements ranging from +49% to -12% for their algebraic re-association optimization [3]. Second, optimizations enable and disable other optimizations so the order of applying optimizations can have an impact on performance [1] [18][21] [10], which is known as the phase ordering problem.…”