Optimizing Solution Quality in Synchronization Synthesis

TL;DR

This paper presents a novel method for automatically inserting synchronization primitives into multithreaded programs to ensure correctness and optimize performance based on platform-specific models, scaling well to real-world applications.

Contribution

It introduces a global constraint-based framework for optimizing synchronization placement considering correctness and performance objectives.

Findings

Scales to real-world concurrent programs like device drivers and servers.

Performance predictions closely match actual measurements.

First comprehensive approach for optimizing synthesized synchronization placement.

Abstract

Given a multithreaded program written assuming a friendly, non-preemptive scheduler, the goal of synchronization synthesis is to automatically insert synchronization primitives to ensure that the modified program behaves correctly, even with a preemptive scheduler. In this work, we focus on the quality of the synthesized solution: we aim to infer synchronization placements that not only ensure correctness, but also meet some quantitative objectives such as optimal program performance on a given computing platform. The key step that enables solution optimization is the construction of a set of global constraints over synchronization placements such that each model of the constraints set corresponds to a correctness-ensuring synchronization placement. We extract the global constraints from generalizations of counterexample traces and the control-flow graph of the program. The global constraints enable us to choose from among the encoded synchronization solutions using an objective function. We consider two types of objective functions: ones that are solely dependent on the program (e.g., minimizing the size of critical sections) and ones that are also dependent on the computing platform. For the latter, given a program and a computing platform, we construct a performance model based on measuring average contention for critical sections and the average time taken to acquire and release a lock under a given average contention. We empirically evaluated that our approach scales to typical module sizes of many real world concurrent programs such as device drivers and multithreaded servers, and that the performance predictions match reality. To the best of our knowledge, this is the first comprehensive approach for optimizing the placement of synthesized synchronization.