Abstract:To ensure the correctness and efficiency of floating-point programs, automatic optimizing techniques for floating-point programs have been attached attention from academia in recent years. The core idea is to summarize transformation rules from classical numerical analysis theory and rewrite floating-point programs automatically with more stable algorithms, which produces more stable, efficient and maintainable programs. However, efficiency becomes the bottleneck of these techniques. As more and more transformation rules are discovered and summarized, the rule base used by the optimizing framework becomes larger and larger. It is more difficult to scan the whole rule base as usual. In this paper, we propose an experience-guided acceleration strategy for floatingpoint optimization. Based on the principle of similar causes of floating-point errors, the strategy extracts the corresponding symbolic and structural features of successfully optimized floating-point programs, and saves the rule sequences involved in the optimization of the programs into a hash-based optimization experience database. When optimizing a floating-point program, the optimizer first calculates the similarity between symbolic and structural features of the program and that of records in the experience base. Rule sequences associated with records having higher similarity are preferred to be used in rewriting and optimizing the program. With more and more programs to be optimized, the experience base increases gradually and a hashed-partition design ensures the efficiency of record-searching. The above approach has been evaluated on the FPBench benchmark and the open source engine OpenRelativity. The results show that compared to the traditional optimizing approach, it can achieve a 6.04x speedup in average. In addition, it improves accuracy for floating-point programs effectively as well as the traditional optimizing approach does. Therefore, our approach improves the usability of automatic optimizing techniques significantly.

Mnemonic Enhancement Real-Time Optimization with Modified Barycentric Interpolation for Process Systems

Mnemonic Enhancement Optimization (MEO) for Real-Time Optimization of Industrial Processes

Real-time Optimization Method MEO Based on Mnemonic Enhancement

Sensitivity-Based Mnemonic Enhancement Optimization (S-Meo) For Real-Time Optimization Of Chemical Process

A Modified Mnemonic Enhancement Optimization Method For Solving Parametric Nonlinear Programming Problems

An Improved Quasi-Sequential Approach To Large-Scale Dynamic Process Optimization

Domain decomposition approach for parallel improvement of tetrahedral meshes.

An Optimization Method Based on Mnemonic Enhancement for Solving Data Reconciliation Problems

Internal-growth Strategies for Dynamic Process Optimization with Differential-algebraic Equations

A Trust-Region Framework for Real-Time Optimization with Structural Process-Model Mismatch

Dynamic Optimization of Multivariable Endothermic Reaction in Cascade CSTR

Model-Free Quality Optimization Strategy for a Batch Process with Short Cycle Time and Low Operational Cost

Convergence Depth Control for Process System Optimization

A Hybrid Numerical‐symbolic Solving Strategy for Equation‐oriented Process Simulation and Optimization

Real-time Optimization for Chemical Processes Based on On-Line Modeling of Controlled Variables

Identification-based Real-Time Optimization and Its Application to Power Plants

A Parallel Function Evaluation Approach for Solution to Large-Scale Equation-Oriented Models.

Structure based model reduction in real-time optimization

Dynamic Optimization for Grade Transition Processes Using Orthogonal Collocation on Molecular Weight Distribution

Robustness Multi-Objective Optimization for Parallel Robot Based on Subregional Meta-Heuristic Iteration

An Experience-Guided Acceleration Strategy for Floating-Point Optimization