mlsanityData to 7 Oct 2026

Learning to Optimize Tensor Programs

arXiv 1805.08166 ↗First posted 21 May 2018cs.LG stat.MLcited by 138

Not on the map. Why?

Abstract (arXiv)

We introduce a learning-based framework to optimize tensor programs for deep learning workloads. Efficient implementations of tensor operators, such as matrix multiplication and high dimensional convolution, are key enablers of effective deep learning systems. However, existing systems rely on manually optimized libraries such as cuDNN where only a narrow range of server class GPUs are well-supported. The reliance on hardware-specific operator libraries limits the applicability of high-level graph optimizations and incurs significant engineering costs when deploying to new hardware targets. We use learning to remove this engineering burden. We learn domain-specific statistical cost models to guide the search of tensor operator implementations over billions of possible program variants. We further accelerate the search by effective model transfer across workloads. Experimental results show that our framework delivers performance competitive with state-of-the-art hand-tuned libraries for low-power CPU, mobile GPU, and server-class GPU.

Key facts. Cited by 138 papers in the map (data through 7 Oct 2026). First posted on arXiv 21 May 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 28; 2022: 26; 2023: 23; 2024: 19; 2025: 24; 2026: 18.

138 papers.

Sort:

By area:

  1. 2024Efficient LLM Scheduling by Learning to Rankcited by 83
  2. 2021DNNFusion: Accelerating Deep Neural Networks Execution with Advanced Operator Fusioncited by 56
  3. 2022SparseTIR: Composable Abstractions for Sparse Compilation in Deep Learningcited by 47
  4. 2024T-MAC: CPU Renaissance via Table Lookup for Low-Bit LLM Deployment on Edgecited by 47
  5. 2025CudaForge: An Agent Framework with Hardware Feedback for CUDA Kernel Optimizationcited by 42
  6. 2021MLGO: a Machine Learning Guided Compiler Optimizations Frameworkcited by 40
  7. 2024Tiny Machine Learning: Progress and Futurescited by 40
  8. 2024Mirage: A Multi-Level Superoptimizer for Tensor Programscited by 39
  9. 2021FLAT: An Optimized Dataflow for Mitigating Attention Bottleneckscited by 38
  10. 2022TensorIR: An Abstraction for Automatic Tensorized Program Optimizationcited by 38
  11. 2021A Deep Learning Based Cost Model for Automatic Code Optimizationcited by 33
  12. 2022Tensor Program Optimization with Probabilistic Programscited by 30
38 more
  1. 2021A Runtime-Based Computational Performance Predictor for Deep Neural Network Trainingcited by 29
  2. 2022NNSmith: Generating Diverse and Valid Test Cases for Deep Learning Compilerscited by 29
  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. 2021NAAS: Neural Accelerator Architecture Searchcited by 24
  6. 2021Bolt: Bridging the Gap between Auto-tuners and Hardware-native Performancecited by 18
  7. 2021The CoRa Tensor Compiler: Compilation for Ragged Tensors with Minimal Paddingcited by 17
  8. 2023TpuGraphs: A Performance Prediction Dataset on Large Tensor Computational Graphscited by 16
  9. 2024Allo: A Programming Model for Composable Accelerator Designcited by 16
  10. 2023Supersonic: Learning to Generate Source Code Optimizations in C/C++cited by 15
  11. 2022TLP: A Deep Learning-based Cost Model for Tensor Program Tuningcited by 14
  12. 2025MPK: A Compiler and Runtime for Mega-Kernelizing Tensor Programscited by 14
  13. 2021SECDA: Efficient Hardware/Software Co-Design of FPGA-based DNN Accelerators for Edge Inferencecited by 12
  14. 2023Incremental Randomized Smoothing Certificationcited by 12
  15. 2023Relax: Composable Abstractions for End-to-End Dynamic Machine Learningcited by 12
  16. 2023Slapo: A Schedule Language for Progressive Optimization of Large Deep Learning Model Trainingcited by 11
  17. 2025SwizzlePerf: Hardware-Aware LLMs for GPU Kernel Performance Optimizationcited by 11
  18. 2026KernelFoundry: Hardware-aware evolutionary GPU kernel optimizationcited by 11
  19. 2021UNIT: Unifying Tensorized Instruction Compilationcited by 10
  20. 2021HASCO: Towards Agile HArdware and Software CO-design for Tensor Computationcited by 10
  21. 2021MetaTune: Meta-Learning Based Cost Model for Fast and Efficient Auto-tuning Frameworkscited by 9
  22. 2023Learning Large Graph Property Prediction via Graph Segment Trainingcited by 9
  23. 2024LayerDAG: A Layerwise Autoregressive Diffusion Model for Directed Acyclic Graph Generationcited by 8
  24. 2022Decompiling x86 Deep Neural Network Executablescited by 7
  25. 2023oneDNN Graph Compiler: A Hybrid Approach for High-Performance Deep Learning Compilationcited by 7
  26. 2024LOOPer: A Learned Automatic Code Optimizer For Polyhedral Compilerscited by 7
  27. 2024Optimal Kernel Orchestration for Tensor Programs with Korchcited by 7
  28. 2021BoGraph: Structured Bayesian Optimization From Logs for Expensive Systems with Many Parameterscited by 6
  29. 2021Bring Your Own Codegen to Deep Learning Compilercited by 5
  30. 2022Transfer-Tuning: Reusing Auto-Schedules for Efficient Tensor Program Code Generationcited by 5
  31. 2025QiMeng-TensorOp: Automatically Generating High-Performance Tensor Operators with Hardware Primitivescited by 5
  32. 2025SuperCoder: Assembly Program Superoptimization with Large Language Modelscited by 5
  33. 2021Customized Monte Carlo Tree Search for LLVM/Polly's Composable Loop Optimization Transformationscited by 4
  34. 2021AIRCHITECT: Learning Custom Architecture Design and Mapping Spacecited by 4
  35. 2021Learning to Superoptimize Real-world Programscited by 4
  36. 2021A Highly Configurable Hardware/Software Stack for DNN Inference Accelerationcited by 4
  37. 2022Towards Hardware-Specific Automatic Compression of Neural Networkscited by 4
  38. 2023Tango: rethinking quantization for graph neural network training on GPUscited by 4

50 most cited of 138 All 138 (JSON)