Machine learning-based prefetch optimization for data center applications

Liao, Shih-Wei; Hung, Tzu-Han; Nguyen, Donald; Chou, Chih-Chung; Tu, Chia-Heng; Zhou, Hucheng

doi:10.1145/1654059.1654116

Cited by 61 publications

(43 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In our previous work [17,26] we used static or dynamic (performance counters) code features with SUIF, Intel and PathScale compilers to predict a set of multiple optimizations that improve execution time for new programs based on similarities between previously optimized programs. Liao et al [82] used machine learning to performance counters and decision trees to choose hardware prefetcher configurations. Several researchers [24,58,62] attempted to characterize program input in order to predict best code variant at run-time using several machine learning methods, including automatically generated decision trees and statistical modeling.…”

Section: Related Workmentioning

confidence: 99%

Milepost GCC: Machine Learning Enabled Self-tuning Compiler

Fursin

Kashnikov

Memon

et al. 2011

Int J Parallel Prog

218

251

View full text Add to dashboard Cite

Tuning compiler optimizations for rapidly evolving hardware makes porting and extending an optimizing compiler for each new platform extremely challenging. Iterative optimization is a popular approach to adapting programs to a new architecture automatically using feedback-directed compilation. However, the large number of evaluations required for each program has prevented iterative compilation from widespread take-up in production compilers. Machine learning has been proposed to tune optimizations across programs systematically but is currently limited to a few transformations, long training phases and critically lacks publicly released, stable tools.Our approach is to develop a modular, extensible, self-tuning optimization infrastructure to automatically learn the best optimizations across multiple programs and architectures based on the correlation between program features, run-time behavior and optimizations. In this paper we describe Milepost GCC, the first publicly-available open-source machine learning-based compiler. It consists of an Interactive Compilation Interface (ICI) and plugins to extract program features and exchange optimization data with the cTuning.org open public repository. It automatically adapts the internal optimization heuristic at function-level granularity to improve execution time, code size and compilation time of a new program on a given architecture. Part of the Milepost technology together with low-level ICI-inspired plugin framework is now included in the mainline GCC.We developed machine learning plugins based on probabilistic and transductive approaches to predict good combinations of optimizations. Our preliminary experimental results show that it is possible to automatically reduce the execution time of individual MiBench programs, some by more than a factor of 2, while also improving compilation 1 INRIA Saclay, France (HiPEAC member) · 2 University of Versailles Saint Quentin en Yvelines, France · 3 IBM Haifa, Israel (HiPEAC member) · 4 CAPS Entreprise, France (HiPEAC member) · 5 ARC International, UK · 6 University of Edinburgh, UK (HiPEAC member) · 2 time and code size. On average we are able to reduce the execution time of the MiBench benchmark suite by 11% for the ARC reconfigurable processor. We also present a realistic multi-objective optimization scenario for Berkeley DB library using Milepost GCC and improve execution time by approximately 17%, while reducing compilation time and code size by 12% and 7% respectively on Intel Xeon processor.

show abstract

Section: Related Workmentioning

confidence: 99%

Milepost GCC: Machine Learning Enabled Self-tuning Compiler

Fursin

Kashnikov

Memon

et al. 2011

Int J Parallel Prog

218

251

View full text Add to dashboard Cite

show abstract

“…Critical algorithms, such as garbage collection [7], may gain more than two times speedup. Application performance metrics due to prefetcher configuration may reach up to 70% improvement, depending on the extensiveness of the memory usage by the workload [2,26].…”

Section: Methodsmentioning

confidence: 99%

Enabling dynamic data centers with a smart bare-metal server platform

et al. 2010

View full text Add to dashboard Cite

Ever increasing data center complexity poses a significant burden on IT administrators. This burden can become unbearable without the help of self-managed systems that monitor themselves and automatically modify their state in order to carry out business processes according to high level objectives set by service level agreements (SLA) and policies. Among the key IT management tasks that must be automated and enhanced to realize the idea of an autonomic and highly dynamic data center, are discovery, configuration, and provisioning of new servers. In this direction, this paper describes pre-boot capabilities endowing the bare metal server with the ability to be discovered, queried, configured, and provisioned at time zero using industry standards like Common Information Model (CIM), CIM-XML, and Service Location Protocol (SLP). The capabilities are implemented as a payload of an Intel® Extensible Firmware Interface (EFI)-compliant BIOS, the Intel® Rapid Boot Toolkit (IRBT), allowing a resource manager to discover a new server during pre-boot, possibly in a baremetal state, and then perform an asset inventory, configure

show abstract

“…The most notable merit of M5P is that it can provide an interpretable model to help designers focus on the most critical performance bottlenecks. Actually, M5P has already been employed for performance modeling and analysis of computer systems [Ould-Ahmed-Vall et al 2007b;Liao et al 2009;Guo et al 2011]. In short, the reason why we use ANN, SVM, and M5P as the base models to build ELSE is that these models have been widely used in performance analysis and modeling of computer systems, and their abilities in predicting the performance and power consumption have also been well demonstrated.…”

Section: Base Regression Modelsmentioning

confidence: 99%

Robust Design Space Modeling

Guo

Chen

Zhou

et al. 2015

ACM Trans. Des. Autom. Electron. Syst.

View full text Add to dashboard Cite

Architectural design spaces of microprocessors are often exponentially large with respect to the pending processor parameters. To avoid simulating all configurations in the design space, machine learning and statistical techniques have been utilized to build regression models for characterizing the relationship between architectural configurations and responses (e.g., performance or power consumption). However, this article shows that the accuracy variability of many learning techniques over different design spaces and benchmarks can be significant enough to mislead the decision-making. This clearly indicates a high risk of applying techniques that work well on previous modeling tasks (each involving a design space, benchmark, and design objective) to a new task, due to which the powerful tools might be impractical.Inspired by ensemble learning in the machine learning domain, we propose a robust framework called ELSE to reduce the accuracy variability of design space modeling. Rather than employing a single learning technique as in previous investigations, ELSE employs distinct learning techniques to build multiple base regression models for each modeling task. This is not a trivial combination of different techniques (e.g., always trusting the regression model with the smallest error). Instead, ELSE carefully maintains the diversity of base regression models and constructs a metamodel from the base models that can provide accurate predictions even when the base models are far from accurate. Consequently, we are able to reduce the number of cases in which the final prediction errors are unacceptably large. Experimental results validate the robustness of ELSE: compared with the widely used artificial neural network over 52 distinct modeling tasks, ELSE reduces the accuracy variability by about 62%. Moreover, ELSE reduces the average prediction error by 27% and 85% for the investigated MIPS and POWER design spaces, respectively.

show abstract

Machine learning-based prefetch optimization for data center applications

Cited by 61 publications

References 27 publications

Milepost GCC: Machine Learning Enabled Self-tuning Compiler

Milepost GCC: Machine Learning Enabled Self-tuning Compiler

Enabling dynamic data centers with a smart bare-metal server platform

Robust Design Space Modeling

Contact Info

Product

Resources

About