ProTuner: Tuning Programs with Monte Carlo Tree Search

Haj-Ali, Ameer; Genç, Hasan; Huang, Qijing; Moses, W.W.; Wawrzynek, John; Asanović, Krste; Stoica, Ion

doi:10.48550/arxiv.2005.13685

Cited by 4 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Black-box optimization has been adopted in high-performance computing libraries [16,3]. Recent advances in automatic tensor program optimization brought a rich set of techniques to accelerate search through better cost modeling [10,4,34,44] and learning-based search [2,25,1,19]. Different variations of pre-defined search spaces have also been proposed that couple with the automatic tensor program optimization frameworks [10,43,1].…”

Section: Related Workmentioning

confidence: 99%

“…Tracing of the program execution across different runs leads to a set of linearized probabilistic programs on the right. Only sampling and transformation instructions are traced, while all other constructs and control flow in the host language are ignored.Auto-scheduling[43,45,1,19] requires developers to implement workload agnostic transformation rules. MetaSchedule achieves the same programmability and functionality through specific probabilistic transformation modules that corresponds to the search space generation rules.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Tensor Program Optimization with Probabilistic Programs

Shao¹,

Zhou²,

Feng³

et al. 2022

Preprint

View full text Add to dashboard Cite

Automatic optimization for tensor programs becomes increasingly important as we deploy deep learning in various environments, and efficient optimization relies on a rich search space and effective search. Most existing efforts adopt a search space which lacks the ability to efficiently enable domain experts to grow the search space. This paper introduces MetaSchedule, a domain-specific probabilistic programming language abstraction to construct a rich search space of tensor programs. Our abstraction allows domain experts to analyze the program, and easily propose stochastic choices in a modular way to compose program transformation accordingly. We also build an end-to-end learning-driven framework to find an optimized program for a given search space. Experimental results show that MetaSchedule can cover the search space used in the state-of-the-art tensor program optimization frameworks in a modular way. Additionally, it empowers domain experts to conveniently grow the search space and modularly enhance the system, which brings 48% speedup on end-to-end deep learning workloads.

show abstract

Section: Related Workmentioning

confidence: 99%

mentioning

confidence: 99%

Tensor Program Optimization with Probabilistic Programs

Shao¹,

Zhou²,

Feng³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Halide [49] introduces the concept of compute and schedule to represent software and optimization, while TVM [9] generalizes the concept and allows users to use tensorize primitive to lower part of software to spatial accelerators manually. Automatic schedulers such as Halide Scheduler [2,30,38], FlexTensor [70], Pro-Tuner [19], ALT [63], Rammer [34], NeoFlow [69], and Ansor [68] focus on general-purpose hardware and ignore the mapping problem for spatial accelerators such as Tensor Core. Polyhedral model is widely used in compilers for constrained optimization [5,56,57].…”

Section: Related Workmentioning

confidence: 99%

Amos

Zheng

Chen

Wei

et al. 2022

Proceedings of the 49th Annual International Symposium on Computer Architecture

View full text Add to dashboard Cite

Hardware specialization is a promising trend to sustain performance growth. Spatial hardware accelerators that employ specialized and hierarchical computation and memory resources have recently shown high performance gains for tensor applications such as deep learning, scientific computing, and data mining. To harness the power of these hardware accelerators, programmers have to use specialized instructions with certain hardware constraints. However, these hardware accelerators and instructions are quite new and there is a lack of understanding of the hardware abstraction, performance optimization space, and automatic methodologies to explore the space. Existing compilers use handtuned computation implementations and optimization templates, resulting in sub-optimal performance and heavy development costs.In this paper, we propose AMOS, which is an automatic compilation framework for spatial hardware accelerators. Central to this framework is the hardware abstraction that not only clearly specifies the behavior of spatial hardware instructions, but also formally defines the mapping problem from software to hardware. Based on * Work done while the author was a student at Peking University.

show abstract

“…It adopts a reinforcement learning-based search strategy, using the cost of the tensor program as the reward value during reinforcement learning. ProTuner [18] uses Monte Carlo tree search to solve the inaccurate estimation problem in the Halide auto-scheduler. The TIRAMISU cost model [5] extracts loop information, buffer access matrix, and AST information from the tensor program and propagates the data forward recursively according to the AST structure.…”

Section: Related Workmentioning

confidence: 99%

TLP: A Deep Learning-based Cost Model for Tensor Program Tuning

Zhai¹,

Liu²,

Chu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Tensor program tuning is a non-convex objective optimization problem, to which search-based approaches have proven to be effective. At the core of the search-based approaches lies the design of the cost model. Though deep learning-based cost models perform significantly better than other methods, they still fall short and suffer from the following problems. First, their feature extraction heavily relies on expert-level domain knowledge in hardware architectures. Even so, the extracted features are often unsatisfactory and require separate considerations for CPUs and GPUs. Second, a cost model trained on one hardware platform usually performs poorly on another, a problem we call cross-hardware unavailability.In order to address these problems, we propose TLP and MTL-TLP. TLP is a deep learning-based cost model that facilitates tensor program tuning. Instead of extracting features from the tensor program itself, TLP extracts features from the schedule primitives. We treat schedule primitives as tensor languages. TLP is thus a Tensor Language Processing task. In this way, the task of predicting the tensor program latency through the cost model is transformed into a natural language processing (NLP) regression task. MTL-TLP combines Multi-Task Learning and TLP to cope with the crosshardware unavailability problem.

show abstract

ProTuner: Tuning Programs with Monte Carlo Tree Search

Cited by 4 publications

References 0 publications

Tensor Program Optimization with Probabilistic Programs

Tensor Program Optimization with Probabilistic Programs

Amos

TLP: A Deep Learning-based Cost Model for Tensor Program Tuning

Contact Info

Product

Resources

About