mlsanityData to 13 Oct 2026

Generating GPU Compiler Heuristics using Reinforcement Learning

arXiv 2111.12055 ↗First posted 23 Nov 2021cs.LG cs.AI cs.CLcited by 3

Not on the map. Why?

Abstract (arXiv)

GPU compilers are complex software programs with many optimizations specific to target hardware. These optimizations are often controlled by heuristics hand-designed by compiler experts using time- and resource-intensive processes. In this paper, we developed a GPU compiler autotuning framework that uses off-policy deep reinforcement learning to generate heuristics that improve the frame rates of graphics applications. Furthermore, we demonstrate the resilience of these learned heuristics to frequent compiler updates by analyzing their stability across a year of code check-ins without retraining. We show that our machine learning-based compiler autotuning framework matches or surpasses the frame rates for 98% of graphics benchmarks with an average uplift of 1.6% up to 15.8%.

Key facts. Cited by 3 papers in the map (data through 13 Oct 2026). Builds on 15 of its 32 references that are in the map. First posted on arXiv 23 Nov 2021.

Builds on

15 of 32 references.

References · 15

  1. 2013Playing Atari with Deep Reinforcement Learningcited by 3,221
  2. 2020Conservative Q-Learning for Offline Reinforcement Learningcited by 1,782
  3. 2016Understanding deep learning requires rethinking generalizationcited by 1,666
  4. 2020Learning from Noisy Labels with Deep Neural Networks: A Surveycited by 581
  5. 2018code2vec: Learning Distributed Representations of Codecited by 250
  6. 2017Device Placement Optimization with Reinforcement Learningcited by 85
  7. 2020DisCor: Corrective Feedback in Reinforcement Learning via Distribution Correctioncited by 71
  8. 2018Neural Code Comprehension: A Learnable Representation of Code Semanticscited by 59
  9. 2018A Survey on Compiler Autotuning using Machine Learningcited by 40
  10. 2021MLGO: a Machine Learning Guided Compiler Optimizations Frameworkcited by 40
  11. 2019NeuroVectorizer: End-to-End Vectorization with Deep Reinforcement Learningcited by 37
  12. 2019Deep Reinforcement Learning for Imbalanced Classificationcited by 24
3 more
  1. 2020ProGraML: Graph-based Deep Learning for Program Optimization and Analysiscited by 23
  2. 2019AutoPhase: Compiler Phase-Ordering for High Level Synthesis with Deep Reinforcement Learningcited by 12
  3. 2021RL-Scope: Cross-Stack Profiling for Deep Reinforcement Learning Workloadscited by 5

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 1; 2024: 0; 2025: 1.

3 papers.

  1. 2025Discovering Software Parallelization Points Using Deep Neural Networkscited by 1
  2. 2022Learning to generate Reliable Broadcast Algorithmscited by 0
  3. 2023Path Planning using Reinforcement Learning: A Policy Iteration Approachcited by 0