Summary of Mirage: a Multi-level Superoptimizer For Tensor Programs, by Mengdi Wu and Xinhao Cheng and Shengyu Liu and Chunan Shi and Jianan Ji and Kit Ao and Praveen Velliengiri and Xupeng Miao and Oded Padon and Zhihao Jia
Mirage: A Multi-Level Superoptimizer for Tensor Programs
by Mengdi Wu, Xinhao Cheng, Shengyu Liu, Chunan Shi, Jianan Ji, Kit Ao, Praveen Velliengiri, Xupeng Miao, Oded Padon, Zhihao Jia
First submitted to arxiv on: 9 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Programming Languages (cs.PL)
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 The proposed Mirage system is a multi-level superoptimizer designed for tensor programs, aiming to discover novel optimizations by combining algebraic transformations, schedule transformations, and kernel generation. The core idea behind Mirage is the representation of tensor programs at various GPU compute levels, allowing for pruning and probabilistic equivalence verification procedures to ensure optimality. Experimental results demonstrate Mirage’s superiority over existing approaches, achieving 1.1-2.9x speedups even on heavily optimized DNNs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Mirage is a new tool that helps make computer programs run faster. It works by looking at the program in different ways and finding new ways to make it run quickly. Mirage uses a special way of representing the program, called , which lets it find optimizations that combine different techniques. To make sure the optimized program is correct, Mirage checks if it’s equivalent to the original program. The results show that Mirage can make programs run 1-3 times faster than other methods. |
Keywords
» Artificial intelligence » Pruning
