Using Process Calculus for Optimizing Data and Computation Sharing in Complex Stateful Parallel Computations
🤖 AI Summary
To address inefficient data and computation sharing in complex stateful parallel computing—such as agent-based simulation—this paper introduces Behavioral Equations, a novel formal framework extending the π-calculus to uniformly model both code and state, enabling automatic composition and equivalence-preserving transformations. We unify key optimizations—including program merging, message synthesis, local communication elimination, communication-to-local-computation conversion, and aggregation pushdown—into a set of formally verified equation-based rewrite rules. Based on this foundation, we design and implement OptiFusion, a compiler and runtime system supporting end-to-end optimization. Experimental evaluation on representative stateful workloads demonstrates that OptiFusion achieves over 10× speedup compared to state-of-the-art systems, up to 2× improvement over hand-optimized implementations, and significantly outperforms existing automated optimization approaches.
Application Category
📝 Abstract
We propose novel techniques that exploit data and computation sharing to improve the performance of complex stateful parallel computations, like agent-based simulations. Parallel computations are translated into behavioral equations, a novel formalism layered on top of the foundational process calculus $pi$-calculus. Behavioral equations blend code and data, allowing a system to easily compose and transform parallel programs into specialized programs. We show how optimizations like merging programs, synthesizing efficient message data structures, eliminating local messaging, rewriting communication instructions into local computations, and {aggregation pushdown} can be expressed as transformations of behavioral equations. We have also built a system called OptiFusion that implements behavioral equations and the aforementioned optimizations. Our experiments showed that OptiFusion is over 10$ imes$ faster than state-of-the-art stateful systems benchmarked via complex stateful workloads. Generating specialized instructions that are impractical to write by hand allows OptiFusion to outperform even the hand-optimized implementations by up to 2$ imes$.
Problem
Research questions and friction points this paper is trying to address.
Optimizing data and computation sharing in stateful parallel computations
Translating parallel computations into behavioral equations for optimization
Specializing programs via transformations to outperform hand-optimized implementations
Innovation
Methods, ideas, or system contributions that make the work stand out.
Behavioral equations blend code and data
Transformations optimize parallel computations
OptiFusion system implements novel optimizations
🔎 Similar Papers
No similar papers found.
