Finding effective compilation sequences

Abstract

Most modern compilers operate by applying a fixed, program-independent sequence of optimizations to all programs. Compiler writers choose a single "compilation sequence", or perhaps a couple of compilation sequences. In choosing a sequence, they may consider performance of benchmarks or other important codes. These sequences are intended as general-purpose tools, accessible through command-line flags such as -O2 and -O3.Specific compilation sequences make a significant difference in the quality of the generated code, whether compiling for speed, for space, or for other metrics. A single universal compilation sequence does not produce the best results over all programs [8, 10, 29, 32]. Finding an optimal program-specific compilation sequence is difficult because the space of potential sequences is huge and the interactions between optimizations are poorly understood. Moreover, there is no systematic exploration of the costs and benefits of searching for good (i.e., within a certain percentage of optimal) program-specific compilation sequences.In this paper, we perform a large experimental study of the space of compilation sequences over a set of known benchmarks, using our prototype adaptive compiler. Our goal is to characterize these spaces and to determine if it is cost-effective to construct custom compilation sequences. We report on five exhaustive enumerations which demonstrate that 80% of the local minima in the space are within 5 to 10% of the optimal solution. We describe three algorithms tailored to search such spaces and report on experiments that use these algorithms to find good compilation sequences. These experiments suggest that properties observed in the enumerations hold for larger search spaces and larger programs. Our findings indicate that for the cost of 200 to 4,550 compilations, we can find custom sequences that are 15 to 25% better than the human-designed fixed-sequence originally used in our compiler.

A Framework for Exploring Optimization Properties

“…To evaluate FOP, we compared three approaches to applying optimizations: a fixedorder approach, an empirical approach that uses a genetic algorithm (GA) to search for effective optimization sequences [1] and our approach. We run experiments on an Intel Pentium IV 2.4GHz machine, with 512MB of memory and RedHat Linux.…”

“…The empirical approach (GA) has the same configuration as in [1]. We performed a search for each function in a program with 10 generations.…”

“…Various reports from research and commercial projects indicate that the performance of software can be improved by 20% to 40% with aggressive optimization [1] [2]. However, it has long been known that there are issues with the application of optimizations.…”

“…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. Finally, optimization configurations can impact the effectiveness of optimizations (e.g., how many times to unroll a loop or the tile size) [15][4] [8].…”

“…The empirical approach has been used to determine the correctness of an optimizer [6]. It has also been used to determine profitability and interactions for finding good optimization sequences and configurations [1][8] [18] [11]. A disadvantage of the experimental approach is its cost and scalability, as the execution of the program is required.…”