A Machine Learning Approach for Performance Prediction and Scheduling on Heterogeneous CPUs

Nemirovsky, Daniel; Arkose, Tugberk; Marković, Nikola; Nemirovsky, Mario; Ünsal, Osman; Cristal, Adrián

doi:10.1109/sbac-pad.2017.23

Cited by 36 publications

(29 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…An implementation approach is to modify the scheduler code within the OS to collect the attributes and also run the predictions using the trained machine learning algorithms. This is the approach taken in [18,19] and has been shown to produce around 30% performance improvements over state-of-the-art schedulers. Other examples of machine learning predictors being exploited by mechanisms to improve system performance are in the area of branch prediction [9] and cache line reusability [10,25].…”

Section: Exploiting the Predictorsmentioning

confidence: 99%

“…Recently, there has been pioneering studies conducted on applying machine/deep learning to CPU scheduling. In the works [18,19] artificial neural network performance predictors are used by the scheduler to improve the system throughput over a Linux based scheduler by over 30%. Other approaches to using machine/deep learning for scheduling has been to classify applications, as well as to identify process attributes and a program's execution history.…”

Section: Related Workmentioning

confidence: 99%

“…However, the few works that have done so in the areas of CPU scheduling [18,19], cache replacement [10,25], and branch prediction [9] have shown tremendous promise. These are but a few of the opportunities we foresee where machine learning could provide a significant advantage towards improving the efficiency of computer systems.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

General Guide to Applying Machine Learning to Computer Architecture

Nemirovsky

Arkose²,

Marković

et al. 2018

JSFI

View full text Add to dashboard Cite

The resurgence of machine learning since the late 1990s has been enabled by significant advances in computing performance and the growth of big data. The ability of these algorithms to detect complex patterns in data which are extremely difficult to achieve manually, helps to produce effective predictive models. Whilst computer architects have been accelerating the performance of machine learning algorithms with GPUs and custom hardware, there have been few implementations leveraging these algorithms to improve the computer system performance. The work that has been conducted, however, has produced considerably promising results. The purpose of this paper is to serve as a foundational base and guide to future computer architecture research seeking to make use of machine learning models for improving system efficiency. We describe a method that highlights when, why, and how to utilize machine learning models for improving system performance and provide a relevant example showcasing the effectiveness of applying machine learning in computer architecture. We describe a process of data generation every execution quantum and parameter engineering. This is followed by a survey of a set of popular machine learning models. We discuss their strengths and weaknesses and provide an evaluation of implementations for the purpose of creating a workload performance predictor for different core types in an x86 processor. The predictions can then be exploited by a scheduler for heterogeneous processors to improve the system throughput. The algorithms of focus are stochastic gradient descent based linear regression, decision trees, random forests, artificial neural networks, and k-nearest neighbors.

show abstract

Section: Exploiting the Predictorsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

General Guide to Applying Machine Learning to Computer Architecture

Nemirovsky

Arkose²,

Marković

et al. 2018

JSFI

View full text Add to dashboard Cite

show abstract

“…Huang [10] uses the computation and communication overlapped through multi-stream concurrent technology to reduce the data transmission bottleneck between CPU and GPU. [11] adopt machine learn ing to determine the optimal partit ioning. [12] proposed a systematic approach by using modeling, profiling and prediction technique to solve workload partitioning.…”

Section: Background and Related Workmentioning

confidence: 99%

Workload Partitioning Algorithm Based on Performance Curve of GPU in Heterogeneous Platforms

Yang¹,

Chen²,

Li³

et al. 2018

Proceedings of the 2018 International Conference on Advanced Control, Automation and Artificial Intelligence (ACAAI 2018)

View full text Add to dashboard Cite

With the development of GPU's general computing power, hybrid systems composed of multi-core CPU and GPU are becoming more and more popular in data parallel applications. Because the performance of GPU is related to the magnitude of the load received, effective load allocation methods are very important for improving the performance of data parallel applications. The existing static load distribution methods fail to use the characteristics effectively -GPU performance changed with the load, causing the load unbalanced. Dynamic load distribution methods easily reduce the performance of the system due to the excessive synchronization and data transmission operation. In this paper, we propose a new workload partitioning algorithm, which takes advantage of the characteristics of GPU performance varying with the workload in off-line analysis stage, and uses the successive decreasing method to determine the optimal load allocation rati o between multi-core CPU and GPU. The effectiveness of the load allocation algorithm is verified on the remote sensing data set based on the median filtering algorithm.

show abstract

“…Regression trees [26] have been used for performance and power prediction of a GPU. Artificial neural networks have successfully been used for performance prediction of a parallelized application [21], but also for workload characterization of general purpose processors [27,28], superscalar processors [29], and microarchitectures [30]. A variant of the MART method has also been used for predicting performances of distributed systems [31].…”

Section: Introductionmentioning

confidence: 99%

Avionics Graphics Hardware Performance Prediction with Machine Learning

Girard

Legault

Bois

et al. 2019

Scientific Programming

View full text Add to dashboard Cite

Within the strongly regulated avionic engineering field, conventional graphical desktop hardware and software application programming interface (API) cannot be used because they do not conform to the avionic certification standards. We observe the need for better avionic graphical hardware, but system engineers lack system design tools related to graphical hardware. The endorsement of an optimal hardware architecture by estimating the performance of a graphical software, when a stable rendering engine does not yet exist, represents a major challenge. As proven by previous hardware emulation tools, there is also a potential for development cost reduction, by enabling developers to have a first estimation of the performance of its graphical engine early in the development cycle. In this paper, we propose to replace expensive development platforms by predictive software running on a desktop computer. More precisely, we present a system design tool that helps predict the rendering performance of graphical hardware based on the OpenGL Safety Critical API. First, we create nonparametric models of the underlying hardware, with machine learning, by analyzing the instantaneous frames per second (FPS) of the rendering of a synthetic 3D scene and by drawing multiple times with various characteristics that are typically found in synthetic vision applications. The number of characteristic combinations used during this supervised training phase is a subset of all possible combinations, but performance predictions can be arbitrarily extrapolated. To validate our models, we render an industrial scene with characteristic combinations not used during the training phase and we compare the predictions to those real values. We find a median prediction error of less than 4 FPS.

show abstract

A Machine Learning Approach for Performance Prediction and Scheduling on Heterogeneous CPUs

Cited by 36 publications

References 17 publications

General Guide to Applying Machine Learning to Computer Architecture

General Guide to Applying Machine Learning to Computer Architecture

Workload Partitioning Algorithm Based on Performance Curve of GPU in Heterogeneous Platforms

Avionics Graphics Hardware Performance Prediction with Machine Learning

Contact Info

Product

Resources

About