mlsanityData to 14 Oct 2026

Ansor: Generating High-Performance Tensor Programs for Deep Learning

arXiv 2006.06762 ↗First posted 11 Jun 2020cs.LG cs.NE cs.PFcited by 201

Not on the map. Why?

Abstract (arXiv)

High-performance tensor programs are crucial to guarantee efficient execution of deep neural networks. However, obtaining performant tensor programs for different operators on various hardware platforms is notoriously challenging. Currently, deep learning systems rely on vendor-provided kernel libraries or various search strategies to get performant tensor programs. These approaches either require significant engineering effort to develop platform-specific optimization code or fall short of finding high-performance programs due to restricted search space and ineffective exploration strategy. We present Ansor, a tensor program generation framework for deep learning applications. Compared with existing search strategies, Ansor explores many more optimization combinations by sampling programs from a hierarchical representation of the search space. Ansor then fine-tunes the sampled programs with evolutionary search and a learned cost model to identify the best programs. Ansor can find high-performance programs that are outside the search space of existing state-of-the-art approaches. In addition, Ansor utilizes a task scheduler to simultaneously optimize multiple subgraphs in deep neural networks. We show that Ansor improves the execution performance of deep neural networks relative to the state-of-the-art on the Intel CPU, ARM CPU, and NVIDIA GPU by up to 3.8×, 2.6×, and 1.7×, respectively.

Key facts. Cited by 201 papers in the map (data through 14 Oct 2026). First posted on arXiv 11 Jun 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 15; 2022: 33; 2023: 28; 2024: 24; 2025: 53; 2026: 48.

201 papers.

Sort:

By area:

  1. 2023SpecInfer: Accelerating Generative Large Language Model Serving with Tree-based Speculative Inference and Verificationcited by 365
  2. 2023OliVe: Accelerating Large Language Models via Hardware-friendly Outlier-Victim Pair Quantizationcited by 92
  3. 2023Full Stack Optimization of Transformer Inference: a Surveycited by 80
  4. 2022Zeus: Understanding and Optimizing GPU Energy Consumption of DNN Trainingcited by 79
  5. 2022ANT: Exploiting Adaptive Numerical Data Type for Low-bit Deep Neural Network Quantizationcited by 51
  6. 2025Astra: A Multi-Agent System for GPU Kernel Performance Optimizationcited by 51
  7. 2022SparseTIR: Composable Abstractions for Sparse Compilation in Deep Learningcited by 47
  8. 2025TileLang: A Composable Tiled Programming Model for AI Systemscited by 43
  9. 2021CompilerGym: Robust, Performant Compiler Optimization Environments for AI Researchcited by 41
  10. 2024Mirage: A Multi-Level Superoptimizer for Tensor Programscited by 41
  11. 2022TensorIR: An Abstraction for Automatic Tensorized Program Optimizationcited by 38
  12. 2025MultiKernelBench: A Multi-Platform Benchmark for Kernel Generationcited by 38
38 more
  1. 2022Tensor Program Optimization with Probabilistic Programscited by 30
  2. 2022NNSmith: Generating Diverse and Valid Test Cases for Deep Learning Compilerscited by 30
  3. 2022Hidet: Task-Mapping Programming Paradigm for Deep Learning Tensor Programscited by 28
  4. 2022Flashlight: Enabling Innovation in Tools for Machine Learningcited by 25
  5. 2023PIT: Optimization of Dynamic Sparse Deep Learning Models via Permutation Invariant Transformationcited by 25
  6. 2025STARK: Strategic Team of Agents for Refining Kernelscited by 25
  7. 2026FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systemscited by 25
  8. 2026K-Search: LLM Kernel Generation via Co-Evolving Intrinsic World Modelcited by 24
  9. 2025AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimizationcited by 23
  10. 2025Characterizing Mobile SoC for Accelerating Heterogeneous LLM Inferencecited by 22
  11. 2021Bolt: Bridging the Gap between Auto-tuners and Hardware-native Performancecited by 18
  12. 2024LUT Tensor Core: A Software-Hardware Co-Design for LUT-Based Low-Bit LLM Inferencecited by 18
  13. 2021The CoRa Tensor Compiler: Compilation for Ragged Tensors with Minimal Paddingcited by 17
  14. 20211xN Pattern for Pruning Convolutional Neural Networkscited by 16
  15. 2022Lite Pose: Efficient Architecture Design for 2D Human Pose Estimationcited by 16
  16. 2023TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphscited by 16
  17. 2024Allo: A Programming Model for Composable Accelerator Designcited by 16
  18. 2025Adaptive Self-improvement LLM Agentic System for ML Library Developmentcited by 16
  19. 2025MPK: A Compiler and Runtime for Mega-Kernelizing Tensor Programscited by 15
  20. 2022TLP: A Deep Learning-based Cost Model for Tensor Program Tuningcited by 14
  21. 2025Autocomp: A Powerful and Portable Code Optimizer for Tensor Acceleratorscited by 13
  22. 2026KernelFoundry: Hardware-aware evolutionary GPU kernel optimizationcited by 13
  23. 2022Benchmarking of DL Libraries and Models on Mobile Devicescited by 12
  24. 2023Relax: Composable Abstractions for End-to-End Dynamic Machine Learningcited by 12
  25. 2026AscendKernelGen: A Systematic Study of LLM-Based Kernel Generation for Neural Processing Unitscited by 12
  26. 2023Slapo: A Schedule Language for Progressive Optimization of Large Deep Learning Model Trainingcited by 11
  27. 2024Improving Parallel Program Performance with LLM Optimizers via Agent-System Interfacescited by 11
  28. 2025SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimizationcited by 11
  29. 2025LMDeploy Accelerates Mixed-Precision LLM Inference with TurboMindcited by 10
  30. 2021MetaTune: Meta-Learning Based Cost Model for Fast and Efficient Auto-tuning Frameworkscited by 9
  31. 2021Tensor Processing Primitives: A Programming Abstraction for Efficiency and Portability in Deep Learning & HPC Workloadscited by 9
  32. 2023Operator Fusion in XLA: Analysis and Evaluationcited by 9
  33. 2023DyCL: Dynamic Neural Network Compilation Via Program Rewriting and Graph Optimizationcited by 9
  34. 2021Neural Architecture Search as Program Transformation Explorationcited by 8
  35. 2024LLM-Aided Compilation for Tensor Acceleratorscited by 8
  36. 2024LayerDAG: A Layerwise Autoregressive Diffusion Model for Directed Acyclic Graph Generationcited by 8
  37. 2022Decompiling x86 Deep Neural Network Executablescited by 7
  38. 2023oneDNN Graph Compiler: A Hybrid Approach for High-Performance Deep Learning Compilationcited by 7

50 most cited of 201 All 201 (JSON)