Modeling and Estimation for the Power Consumption of Matrix Computation on Multi-core Platform

Ren, Da Qi; Suda, Reiji

doi:10.1109/cso.2009.451

Cited by 4 publications

(2 citation statements)

References 4 publications

(6 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, Tiwari, Malik, and Wolfe (1994) proposed an instruction-level power model that divides the power into two parts. Ren and Suda (2009) presented an algorithm-level power model on multi-core platform, which analysed the relationship between energy consumption and algorithm complexity which mainly include computation and communication operations. Email: tan1ming@126.com and resource constraints.…”

Section: Introductionmentioning

confidence: 99%

Energy analysis for an executable program on a single computer based on BP neural network

Tan

Zeng

2013

Journal of Experimental & Theoretical Artificial Intelligen

View full text Add to dashboard Cite

Power management of application programs is a hot research in the area of green computing and high-productivity computing. Because of the complexity of application programs, heterogeneity of processors and uncertainty of the running environment, it is difficult to propose an accurate method to predict energy consumption for application program directly. So, we present a power analysis paradigm for application program based on artificial neural network. First, we build a power analysis model based on back propagation neural network (BPNN). The three factors of software, hardware and environment are taken as the inputs of BPNN, and energy consumption and finish time as the outputs of BPNN. Next, we choose lot of classic application programs from different fields as training samples. After learning and training, an expected BPNN is obtained which can be used to predict energy consumption for other new programs. Repeated experiments show that this power analysis paradigm is rational and feasible.

show abstract

Section: Introductionmentioning

confidence: 99%

Energy analysis for an executable program on a single computer based on BP neural network

Tan

Zeng

2013

Journal of Experimental & Theoretical Artificial Intelligen

View full text Add to dashboard Cite

show abstract

“…Our research aims to develop software methods for power optimized high performance computing on CPU-GPU heterogeneous platforms [2], and to minimize the overall energy dissipation by automatically adjusting the utilization of power related components in the system with automatic tuning mechanisms. Our approach to this end is to firstly build precise models of the power consumption and of the execution time, by which we are able to predict the energy dissipation and the time of a specific computation executed on the target CPU-GPU platform; next based on that information, to tune power related parameters such as clock frequency of CPU, GPU and memory, load distribution schemes, communication protocol etc.…”

Section: Introductionmentioning

confidence: 99%

Accurate Measurements and Precise Modeling of Power Dissipation of CUDA Kernels toward Power Optimized High Performance CPU-GPU Computing

Suda

Ren

2009

2009 International Conference on Parallel and Distributed Computing, Applications and Technologies

Self Cite

View full text Add to dashboard Cite

Abstract-Power dissipation is one of the most imminent limitation factors influencing the development of High Performance Computing (HPC). Toward power-efficient HPC on CPU-GPU hybrid platform, we are investigating software methodologies to achieve optimized power utilization by algorithm design and programming technique. In this paper we discuss power measurements of GPU, propose a method of automatic extraction of power data of CUDA kernels from long measurement sequence, and execute an exactitude and effective power analysis on CUDA kernels. By using the proposed method above, we measured a sample kernel that performs single precision floating point additions on GeForce 8800 GTS. Our results suggest that the power consumption by a nonworking thread in underoccupied half-warp is 71% of the power consumed by a working thread.

show abstract

Energy Efficient Frequency Scaling on GPUs in Heterogeneous HPC Systems

Kraljic

Kerger

Schulz

2022

Architecture of Computing Systems

View full text Add to dashboard Cite

Modeling and Estimation for the Power Consumption of Matrix Computation on Multi-core Platform

Cited by 4 publications

References 4 publications

Energy analysis for an executable program on a single computer based on BP neural network

Energy analysis for an executable program on a single computer based on BP neural network

Accurate Measurements and Precise Modeling of Power Dissipation of CUDA Kernels toward Power Optimized High Performance CPU-GPU Computing

Energy Efficient Frequency Scaling on GPUs in Heterogeneous HPC Systems

Contact Info

Product

Resources

About