Summary of Improving Parallel Program Performance with Llm Optimizers Via Agent-system Interface, by Anjiang Wei et al.
Improving Parallel Program Performance with LLM Optimizers via Agent-System Interface
by Anjiang Wei, Allen Nie, Thiago S. F. X. Teixeira, Rohan Yadav, Wonchan Lee, Ke Wang, Alex Aiken
First submitted to arxiv on: 21 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Distributed, Parallel, and Cluster Computing (cs.DC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A new framework for automating the development of high-performance mappers, which are crucial for efficient parallel programming and complex modeling/simulation in modern scientific discovery, is introduced. The framework leverages generative optimization and richer feedback beyond scalar performance metrics. It features an Agent-System Interface with a Domain-Specific Language (DSL) to abstract away low-level complexity and define a structured search space, as well as AutoGuide, which interprets raw execution output into actionable feedback. This approach outperforms traditional reinforcement learning methods like OpenTuner in fewer iterations, achieving 3.8X faster performance. The framework reduces tuning time from days to minutes while finding mappers that surpass expert-written ones by up to 1.34X speedup across nine benchmarks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new way to create high-performance computer programs that can handle complex tasks and simulations. It’s like having a super-smart assistant that helps you find the best solution quickly, without needing to be an expert in computer systems. The method uses a special language to describe what you want to achieve, and then it finds the best solution through trial and error. This approach is much faster than previous methods and can even beat human-written code in some cases. |
Keywords
* Artificial intelligence * Optimization * Reinforcement learning
