Changing computing paradigms towards power efficiency

Klavík, Pavel; Malossi, A. Cristiano I.; Bekas, Constantin; Curioni, A.

doi:10.1098/rsta.2013.0278

Cited by 20 publications

(18 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Scale-up strategies include building larger arrays and/or operating several of them in parallel. To address problems with a broader range of condition numbers, it will be necessary to increase the precision of the computational memory unit beyond that achieved in the present work to allow the Krylov-subspace inner solver to converge more easily 29 . Possible avenues are improving the memristive device characteristics with respect to variability and conductance noise 33 , mapping a single column of the matrix to multiple physical columns of an array encoding different bits (Supplementary Note III), and using error-correction techniques within the computational memory unit 41 .…”

Section: Performance Assessment and Current Limitationsmentioning

confidence: 98%

See 1 more Smart Citation

Mixed-precision in-memory computing

et al. 2018

Self Cite

View full text Add to dashboard Cite

As CMOS scaling reaches its technological limits, a radical departure from traditional von Neumann systems, which involve separate processing and memory units, is needed in order to significantly extend the performance of today's computers. In-memory computing is a promising approach in which nanoscale resistive memory devices, organized in a computational memory unit, are used for both processing and memory. However, to reach the numerical accuracy typically required for data analytics and scientific computing, limitations arising from device variability and non-ideal device characteristics need to be addressed. Here we introduce the concept of mixed-precision in-memory computing, which combines a von Neumann machine with a computational memory unit. In this hybrid system, the computational memory unit performs the bulk of a computational task, while the von Neumann machine implements a backward method to iteratively improve the accuracy of the solution. The system therefore benefits from both the high precision of digital computing and the energy/areal efficiency of in-memory computing. We experimentally demonstrate the efficacy of the approach by accurately solving systems of linear equations, in particular, a system of 5, 000 equations using 998, 752 phase-change memory devices.

show abstract

Section: Performance Assessment and Current Limitationsmentioning

confidence: 98%

“…The errorcorrection term is computed by solving Az = r with an inexact inner solver using the residual r = b−Ax, calculated with high precision. 29 The algorithm runs until the norm of the residual falls below a desired tolerance, tol.…”

Section: Mixed-precision In-memory Linear Equation Solvermentioning

confidence: 99%

Mixed-precision in-memory computing

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…While in many scientific applications the use of double-precision floating-point is most common, this precision is not always required. For example, iterative methods can exhibit resilience against low precision arithmetic as has been shown for the computation of inverse matrix roots [Lass et al 2018a] and for solving systems of linear equations [Angerer et al 2016;Haidar et al 2018Haidar et al , 2017Klavík et al 2014]. Mainly driven by the growing popularity of artificial neural networks [Gupta et al 2015], we can observe growing support of low-precision data types in hardware accelerators.…”

Section: Approximate Computingmentioning

confidence: 99%

Accurate Sampling with Noisy Forces from Approximate Computing

et al. 2020

View full text Add to dashboard Cite

In scientific computing, the acceleration of atomistic computer simulations by means of custom hardware is finding ever growing application. A major limitation, however, is that the high efficiency in terms of performance and low power consumption entails the massive usage of low-precision computing units. Here, based on the approximate computing paradigm, we present an algorithmic method to rigorously compensate for numerical inaccuracies due to low-accuracy arithmetic operations, yet still obtaining exact expectation values using a properly modified Langevin-type equation.

show abstract

“…This hints to the fact that the FLOPS/W metric used in the Green500 with Linpack, is actually inappropriate to represent energy efficiency of general algorithms [21,22]. Finally, Figures 10 and 12 show a direct view over the power vs. energy consumption relation; here, the importance of the net power (especially on the P755) with respect to the overall power consumption is highlighted.…”

Section: Ict -Energy Concepts For Energy Efficiency and Sustainabilitmentioning

confidence: 99%

Energy-Aware High Performance Computing

Wlotzka¹,

Heuveline²,

Dolz³

et al. 2017

ICT - Energy Concepts for Energy Efficiency and Sustainability

Self Cite

View full text Add to dashboard Cite

High performance computing centres consume substantial amounts of energy to power large-scale supercomputers and the necessary building and cooling infrastructure. Recently, considerable performance gains resulted predominantly from developments in multi-core, many-core and accelerator technology. Computing centres rapidly adopted this hardware to serve the increasing demand for computational power. However, further performance increases in large-scale computing systems are limited by the aggregate energy budget required to operate them. Power consumption has become a major cost factor for computing centres. Furthermore, energy consumption results in carbon dioxide emissions, a hazard for the environment and public health; and heat, which reduces the reliability and lifetime of hardware components. Energy efficiency is therefore crucial in high performance computing.This chapter addresses key issues of energy-aware high performance computing. We outline some numerical methods which are often used in scientific applications, and present an energy profiling and tracing technique suitable to analyse the power consumption of applications. The next section is devoted to the performance and energy characterization of the sparse matrix-vector product, a basic numerical building block. Finally, we discuss opportunities for saving energy in computations by means of two examples. First, we present energy-aware runtimes on shared memory multi-core platforms for the Conjugate Gradient method. Second, we present energy-efficient techniques for multigrid methods on distributed memory clusters.

show abstract

Changing computing paradigms towards power efficiency

Cited by 20 publications

References 14 publications

Mixed-precision in-memory computing

Mixed-precision in-memory computing

Accurate Sampling with Noisy Forces from Approximate Computing

Energy-Aware High Performance Computing

Contact Info

Product

Resources

About