Variant-based competitive parallel execution of sequential programs

Abstract

Competitive parallel execution (CPE) is a simple yet attractive technique to improve the performance of sequential programs on multi-core and multi-processor systems. A sequential program is transformed into a CPE-enabled program by introducing multiple variants for parts of the program. The performance of different variants depends on runtime conditions, such as program input or the execution platform, and the execution time of a CPE-enabled program is the sum of the shortest variants. Variants compete at run-time under the control of a CPE-aware run-time system. The run-time system ensures that the behavior and outcome of a CPE-enabled program is not distinguishable from the one of its original sequential counterpart. We present and evaluate a run-time system that is implemented as a user-space library and that closely interacts with the operating system. The paper discusses two strategies for the generation of variants and investigates the applicability of CPE for two usage scenarios: i) computation-driven CPE: a simple and straightforward parallelization of heuristic algorithms, and ii) compiler-driven CPE: generation of CPE-enabled programs as part of the compilation process using different optimization strategies. Using a state-of-the-art SAT solver as an illustrative example, we report for compiler-based CPE speedups of 4-6% for many data sets, with a maximum slowdown of 2%. Computation-driven CPE provides super-linear speedups for 5 out of 31 data sets (with a maximum speedup of 7.4) and at most a slow-down of 1% for two data sets.