905 Hits in 5.8 sec

Generating GPU Compiler Heuristics using Reinforcement Learning [article]

Ian Colbert, Jake Daly, Norm Rubin
2021 arXiv   pre-print
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.  ...  These optimizations are often controlled by heuristics hand-designed by compiler experts using time- and resource-intensive processes.  ...  Acknowledgements We would like to thank Mike Bedy, Robert Gottlieb, Chris Reeve, Andrew Dupont, Karen Dintino, Peter Scannell and the rest of the AMD GPU compiler team for insightful discussions and infrastructure  ... 
arXiv:2111.12055v1 fatcat:2v3ekxzzfncr5cz6yuunslqspi

Coloring Big Graphs with AlphaGoZero [article]

Jiayi Huang, Mostofa Patwary, Gregory Diamos
2019 arXiv   pre-print
As a result, we are able to learn new state of the art heuristics for graph coloring.  ...  We show that recent innovations in deep reinforcement learning can effectively color very large graphs – a well-known NP-hard problem with clear commercial applications.  ...  High performance training systems use deep reinforcement learning to improve heuristics offline, which are then deployed online in production tools. • We introduce a framework for learning fast heuristics  ... 
arXiv:1902.10162v3 fatcat:vr625w5fgbf4loln2za3nb4hqm

Using meta-heuristics and machine learning for software optimization of parallel computing systems: a systematic literature review

Suejb Memeti, Sabri Pllana, Alécio Binotto, Joanna Kołodziej, Ivona Brandic
2018 Computing  
We review approaches that use machine learning or meta-heuristics for software optimization at compile-time and run-time. Additionally, we discuss challenges and future research directions.  ...  The results of this study may help to better understand the state-of-the-art techniques that use machine learning and meta-heuristics to deal with the complexity of software optimization for parallel computing  ...  Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution  ... 
doi:10.1007/s00607-018-0614-9 fatcat:da2rfxqlcjen5frzfxreimtngm

Towards Automatic Learning of Heuristics for Mechanical Transformations of Procedural Code

Guillermo Vigueras, Manuel Carro, Salvador Tamarit, Julio Mariño
2017 Electronic Proceedings in Theoretical Computer Science  
In this paper we propose a machine learning-based approach to learn heuristics to define program transformation strategies.  ...  Our approach proposes a novel combination of reinforcement learning and classification methods to efficiently tackle the problems inherent to this type of systems.  ...  We also plan to explore other features of reinforcement learning like the use of different learning rates for different states or transformation sequences in order to learn and converge faster towards  ... 
doi:10.4204/eptcs.237.4 fatcat:plfm56g4mral3ekcbey76mevk4

Topological Quantum Compiling with Reinforcement Learning [article]

Yuan-Hang Zhang, Pei-Lin Zheng, Yi Zhang, Dong-Ling Deng
2020 arXiv   pre-print
In this paper, we introduce an efficient algorithm based on deep reinforcement learning that compiles an arbitrary single-qubit gate into a sequence of elementary gates from a finite universal set.  ...  It generates near-optimal gate sequences with given accuracy and is generally applicable to various scenarios, independent of the hardware-feasible universal set and free from using ancillary qubits.  ...  The source code and a pre-trained DNN for this work can be found at quantum compiling.  ... 
arXiv:2004.04743v1 fatcat:rwqhc5zbzjfhff2pdvxrfppdw4

Machine Learning in Compiler Optimisation [article]

Zheng Wang, Michael O'Boyle
2018 arXiv   pre-print
In the last decade, machine learning based compilation has moved from an an obscure research niche to a mainstream activity.  ...  This paper provides both an accessible introduction to the fast moving area of machine learning based compilation and a detailed bibliography of its main achievements.  ...  show that machine learning techniques can be used to automatically construct effective thread-coarsening heuristics across GPU architectures [17] .  ... 
arXiv:1805.03441v1 fatcat:bhd7mpl6lzaedbuy7iln4hntki

A New Kind of Data Centric Performance Portability Challenge Item [article]

Tim Germann
Some aspects of the very heterogeneous architectures that have recently begun to emerge, such as tensor cores, specifically target use cases such as deep learning.  ...  Data science applications, including machine learning, optimization, graph analytics, and other large-scale data-driven computations, present a unique set of challenges to performance portability.  ...  Control • ML-controlled experiments • Efficient exploration of complex space • Reinforcement LearningUse RL agent to control light source experiments • Temperature control for Block  ... 
doi:10.6084/m9.figshare.14125964.v3 fatcat:ppdp4wlqarhpdkdux3dpnwksle

Towards Automatic Learning of Heuristics for Mechanical Transformations of Procedural Code [article]

Guillermo Vigueras and Manuel Carro and Salvador Tamarit and Julio Mariño
2016 arXiv   pre-print
In order to bridge the gap between heterogeneous systems and programmers, in this paper we propose a machine learning-based approach to learn heuristics for defining transformation strategies of a program  ...  Our approach proposes a novel combination of reinforcement learning and classification methods to efficiently tackle the problems inherent to this type of systems.  ...  Once platform-specific code is generated, it is compiled using a platform-specific compiler.  ... 
arXiv:1603.03022v2 fatcat:x6lnxe6b2nctdhl3xibnjwrknq

Machine Learning in Compiler Optimization

Zheng Wang, Michael O'Boyle
2018 Proceedings of the IEEE  
| In the last decade, machine-learning-based compilation has moved from an obscure research niche to a mainstream activity.  ...  In this paper, we describe the relationship between machine learning and compiler optimization and introduce the main concepts of features, models, training, and deployment.  ...  show that machine learning techniques can be used to automatically construct effective threadcoarsening heuristics across GPU architectures [17] .  ... 
doi:10.1109/jproc.2018.2817118 fatcat:vuebhfw7efcdpm5yyzaumia3vi

CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research [article]

Chris Cummins, Bram Wasti, Jiadong Guo, Brandon Cui, Jason Ansel, Sahir Gomez, Somya Jain, Jia Liu, Olivier Teytaud, Benoit Steiner, Yuandong Tian, Hugh Leather
2021 arXiv   pre-print
CompilerGym enables anyone to experiment on production compiler optimization problems through an easy-to-use package, regardless of their experience with compilers.  ...  We build upon the popular OpenAI Gym interface enabling researchers to interact with compilers using Python and a familiar API.  ...  This may be through handcrafted heuristics, search, supervised machine learning, or reinforcement learning. hampers progress in the field.  ... 
arXiv:2109.08267v2 fatcat:s2a3qrrk7zczflszoeta3ztl7q

RLWS: A Reinforcement Learning based GPU Warp Scheduler [article]

Jayvant Anantpur, Nagendra Gulur Dwarakanath, Shivaram Kalyanakrishnan, Shalabh Bhatnagar, R. Govindarajan
2017 arXiv   pre-print
GPU workloads are becoming very diverse in nature and hence one heuristic may not work for all cases.  ...  We propose a Reinforcement Learning based Warp Scheduler (RLWS) which learns to schedule warps based on the current state of the core and the long-term benefits of scheduling actions, adapting not only  ...  [28] used reinforcement learning to schedule basic blocks.  ... 
arXiv:1712.04303v1 fatcat:aqt46qj3gzaf5n3ivckfq7fcnu

Transferable Graph Optimizers for ML Compilers [article]

Yanqi Zhou, Sudip Roy, Amirali Abdolrashidi, Daniel Wong, Peter Ma, Qiumin Xu, Hanxiao Liu, Phitchaya Mangpo Phothilimthana, Shen Wang, Anna Goldie, Azalia Mirhoseini, James Laudon
2021 arXiv   pre-print
Most compilers for machine learning (ML) frameworks need to solve many correlated optimization problems to generate efficient machine code.  ...  Current ML compilers rely on heuristics based algorithms to solve these optimization problems one at a time.  ...  RL for Graph Optimization Reinforcement learning has been used for device placement [21, 9, 19] and has demonstrated run time reduction over human-crafted placements and conventional heuristics.  ... 
arXiv:2010.12438v2 fatcat:ju26bxgmajbgfa4wtwvgrc6k2a

NPAS: A Compiler-aware Framework of Unified Network Pruning and Architecture Search for Beyond Real-Time Mobile Acceleration [article]

Zhengang Li, Geng Yuan, Wei Niu, Pu Zhao, Yanyu Li, Yuxuan Cai, Xuan Shen, Zheng Zhan, Zhenglun Kong, Qing Jin, Zhiyu Chen, Sijia Liu (+4 others)
2021 arXiv   pre-print
To deal with large search space, we propose a meta-modeling procedure based on reinforcement learning with fast evaluation and Bayesian optimization, ensuring the total number of training epochs comparable  ...  In this work, we first propose (i) a general category of fine-grained structured pruning applicable to various DNN layers, and (ii) a comprehensive, compiler automatic code generation framework supporting  ...  The starting learning rate is set to 0.001, and the cosine learning rate scheduler is used if not specified in our paper.  ... 
arXiv:2012.00596v3 fatcat:zinvvpwb5fb2zmqsyifdrfhwga

A Survey of Machine Learning for Computer Architecture and Systems [article]

Nan Wu, Yuan Xie
2021 arXiv   pre-print
It has been a long time that computer architecture and systems are optimized to enable efficient execution of machine learning (ML) algorithms or models.  ...  scope of (micro-)architecture design (memory, branch prediction, NoC), coordination between architecture/system and workload (resource allocation and management, data center management, and security), compiler  ...  DIFFERENT ML TECHNIQUES There are three general frameworks in ML: supervised learning, unsupervised learning and reinforcement learning.  ... 
arXiv:2102.07952v1 fatcat:vzj776a6abesljetqobakoc3dq

MLGO: a Machine Learning Guided Compiler Optimizations Framework [article]

Mircea Trofin
2021 arXiv   pre-print
However, the adoption of ML in general-purpose, industry strength compilers has yet to happen.  ...  As a case study, we present the details and results of replacing the heuristics-based inlining-for-size optimization in LLVM with machine learned models.  ...  [10] automatically extract features from source code, and use supervised learning to learn heuristics for predicting optimal mapping for heterogeneous parallelism and GPU thread coarsening factors.  ... 
arXiv:2101.04808v1 fatcat:jl7owbq5xvf5xmo3qhpksrqtrq
« Previous Showing results 1 — 15 out of 905 results