Unified Analysis of Array and Object References in Strongly Typed Languages

Abstract

We present a simple, unified approach for the analysis and optimization of object field and array element accesses in strongly typed languages, that works in the presence of object references/pointers. This approach builds on Array SSA form [14], a uniform representation for capturing control and data flow properties at the level of array elements. The techniques presented here extend previous analyses at the array element level [15] to handle both array element and object field accesses uniformly.

In the first part of this paper, we show how SSA-based program analyses developed for scalars and arrays can be extended to operate on object references in a strongly typed language like Java. The extension uses Array SSA form as its foundation by modeling object references as indices into hypothetical heap arrays. In the second part of this paper, we present two new sparse analysis algorithms using the heap array representation; one identifies redundant loads, and the other identifies dead stores. Using strong typing to help disambiguation, these algorithms are more efficient than equivalent analyses for weakly typed languages. Using the results of these algorithms, we can perform scalar replacement transformations to change operations on object fields and array elements into operations on scalar variables.

We present preliminary experimental results using the Jalapeño optimizing compiler infrastructure. These results illustrate the benefits obtained by performing redundant load and dead store elimination on Java programs. Our results show that the (dynamic) number of memory operations arising from array-element and object-field accesses can be reduced by up to 28%, resulting in execution time speedups of up to 1.1×.