21,231 Hits in 5.3 sec

Performance prediction of finite-difference solvers for different computer architectures

Mathias Louboutin, Michael Lange, Felix J. Herrmann, Navjot Kukreja, Gerard Gorman
2017 Computers & Geosciences  
) implementation, and the optimization of the implementation for different computer architectures.  ...  A first principles analysis of operational intensity for key time-stepping finite-difference algorithms is presented.  ...  This gives us the following expression for OI as a function of k k , ( ) Operational intensity for finite-differences We derive a general model for the operational intensity of waveequation PDEs solvers  ... 
doi:10.1016/j.cageo.2017.04.014 fatcat:6h2d6r5dl5gp7bditc3vkxveeu

A performance spectrum for parallel computational frameworks that solve PDEs [article]

J. Chang, K. B. Nakshatrala, M. G. Knepley, L. Johnsson
2017 arXiv   pre-print
The aim of this paper is to present various strategies for better understanding the performance of any parallel computational frameworks for solving PDEs.  ...  As proof of concept, we examine commonly used finite element simulation packages and software and apply the performance spectrum to quickly analyze the performance and scalability across various hardware  ...  Compute time on the Maxwell and Opuntia systems is provided by the Center for Advanced Computing & Data Systems (CACDS) at the University of Houston, and compute time on the Edison and Cori systems is  ... 
arXiv:1705.03625v2 fatcat:zqqymw6aijcnbo22bsv4fljeae

GPU acceleration of Data Assembly in Finite Element Methods and its energy implications

Li Tang, X. Sharon Hu, Danny Z. Chen, Michael Niemier, Richard F. Barrett, Simon D. Hammond, Genie Hsieh
2013 2013 IEEE 24th International Conference on Application-Specific Systems, Architectures and Processors  
in FEM • Performance/energy experimental results • Summary 7/12/2013 7 Using MiniFE as representative FEM • MiniFE is a proxy for real FEM applications -Can predict the performance trend  ...  CG solver 54.7% 29.6% DA in miniFE • STC: stiffness matrix compute --compute-intensive • SMA: stiffness matrix assembly --memory-intensive 7/12/2013 9 • Some core improvements of Kepler over  ... 
doi:10.1109/asap.2013.6567597 dblp:conf/asap/TangHCNBHH13 fatcat:bdyst32zz5dwbc2btw7kqrkkzi

Java-Based Coupling for Parallel Predictive-Adaptive Domain Decomposition

Cécile Germain‐Renaud, Vincent Néri
1999 Scientific Programming  
This paper describes an experiment where a data‐parallel (HPF) client interfaces with a sequential computation server through Java.  ...  We show that seamless integration of data‐parallelism is possible, but requires most of the tools from the Java palette: Java Native Interface (JNI), Remote Method Invocation (RMI), callbacks and threads  ...  Acknowledgments The experiments were performed on the SP-2 from the Computing Resource Center (CRI) of Paris-11 and the National Computing Center (CNUSC).  ... 
doi:10.1155/1999/812589 fatcat:g65s3glakzbj5pvkudh5e4r55m

Agent-based computing, adaptive algorithms and bio computing

Krzysztof Centarowicz, Maciej Paszyński, David Pardo, Tibor Bosse, Han La Poutré
2010 Procedia Computer Science  
It gives the ability to integrate results of different domains of computer science and constitutes the powerful tool for solving various problems.  ...  The modern agent-oriented paradigm allows understand the adaptive (e.g. finite element / finite difference) algorithms as a collection of interacting agents making local decision about refinements.  ...  efficient parallel solver for 1D and 2D finite difference method (FDM) simulations.  ... 
doi:10.1016/j.procs.2010.04.218 fatcat:dcyr7buirnh5xkeifhfs4qn3sa

Energy Applications Challenges (SIAM CSE21) [article]

Thomas Evans
map to SIMT architectures.Multiple programming models and approaches are being used to achieve performance portability across a range of GPU-based architectures.  ...  at the SIAM CSE21 conference, MS162/192: Exascale Computing Project Performance Portability Analysis.  ...  models (Kokkos) Commonalities in the Energy Portfolio• All of the applications are time-dependent PDE solversPerformance of linear solver in strong-scaling limit (time-step) bound is a key computational  ... 
doi:10.6084/m9.figshare.14125667.v2 fatcat:gvh4wndrrbh7vcgk7ily24fmdi

Energy Applications Challenges (SIAM CSE21) [article]

Thomas Evans
map to SIMT architectures.Multiple programming models and approaches are being used to achieve performance portability across a range of GPU-based architectures.  ...  at the SIAM CSE21 conference, MS162/192: Exascale Computing Project Performance Portability Analysis.  ...  models (Kokkos) Commonalities in the Energy Portfolio• All of the applications are time-dependent PDE solversPerformance of linear solver in strong-scaling limit (time-step) bound is a key computational  ... 
doi:10.6084/m9.figshare.14125667.v3 fatcat:jiazayqlsjgbpaxe7hiolhukrq

Learned discretizations for passive scalar advection in a 2-D turbulent flow [article]

Jiawei Zhuang, Dmitrii Kochkov, Yohai Bar-Sinai, Michael P. Brenner, Stephan Hoyer
2020 arXiv   pre-print
The computational cost of fluid simulations increases rapidly with grid resolution.  ...  This has given a hard limit on the ability of simulations to accurately resolve small scale features of complex flows.  ...  This ensures that the over-all magnitude of the concentration does not affect the prediction of finite-difference coefficients, and thus our solver satisfies the "semi-linear" requirement for advection  ... 
arXiv:2004.05477v2 fatcat:isjqywlizrf4fl2434t2mcednq

Assessing the role of mini-applications in predicting key performance characteristics of scientific and engineering applications

R.F. Barrett, P.S. Crozier, D.W. Doerfler, M.A. Heroux, P.T. Lin, H.K. Thornquist, T.G. Trucano, C.T. Vaughan
2015 Journal of Parallel and Distributed Computing  
The result was miniFE, putting the linear system into the context of an implicit finite element solver.  ...  These questions concerned the direction of some coding implementations targeting emerging and expected future architectures, including multi-core, many-core, and GPU-accelerated high performance computers  ...  Acknowledgments Support for this work was provided through the Advanced Simulation and Computing (ASC) program funded by U.S. Department of Energy's National Nuclear Security Agency.  ... 
doi:10.1016/j.jpdc.2014.09.006 fatcat:qfe6i6adofhrldfynkskjmco4q

Consistent and symmetry preserving data-driven interface reconstruction for the level-set method [article]

Aaron B. Buhendwa, Deniz A. Bezgin, Nikolaus Adams
2021 arXiv   pre-print
Recently, machine learning has been used to substitute parts of conventional computational fluid dynamics, e.g. the cell-face reconstruction in finite-volume solvers or the curvature computation in the  ...  The combined model is implemented into a CFD solver and demonstrated for two-phase flows. Furthermore, we provide details of floating point symmetric implementation and computational efficiency.  ...  The black line indicates the central finite difference. For interpretation of the model architectures see section 2.3.  ... 
arXiv:2104.11578v1 fatcat:obqgsvmtnrdzlirzvdfy3g6ybm

Solving inverse-PDE problems with physics-aware neural networks [article]

Samira Pakravan, Pouria A. Mistani, Miguel Angel Aragon-Calvo, Frederic Gibou
2020 arXiv   pre-print
This subsequently focuses the computational load to only the discovery of the hidden fields and therefore is more data efficient.  ...  We call this architecture Blended inverse-PDE networks (hereby dubbed BiPDE networks) and demonstrate its applicability for recovering the variable diffusion coefficient in Poisson problems in one and  ...  In our architecture, we use the standard 5-point stencil finite difference discretization of the Poisson equation in the solver layer, i.e.  ... 
arXiv:2001.03608v3 fatcat:hkhelkijvzawvbsfdoyj52wwfu

Performance and accuracy assessments of an incompressible fluid solver coupled with a deep convolutional neural network

Ekhi Ajuria Illarramendi, Michaël Bauerheim, Bénédicte Cuenot
2022 Data-Centric Engineering  
on the accuracy of the solution.This drawback might lead to inaccuracies, potentially unstable simulations and prevent performing fair assessments of the CNN speedup for different network architectures  ...  The resolution of the Poisson equation is usually one of the most computationally intensive steps for incompressible fluid solvers.  ...  Results show that for the same error level, networks with multiple scales allow to perform accurate predictions faster than classical solvers.  ... 
doi:10.1017/dce.2022.2 fatcat:fudeg4gp75gazoxamhu52wm4ui

Machine learning accelerated computational fluid dynamics [article]

Dmitrii Kochkov, Jamie A. Smith, Ayya Alieva, Qing Wang, Michael P. Brenner, Stephan Hoyer
2021 arXiv   pre-print
For both direct numerical simulation of turbulence and large eddy simulation, our results are as accurate as baseline solvers with 8-10x finer resolution in each spatial dimension, resulting in 40-80x  ...  Here we use end-to-end deep learning to improve approximations inside computational fluid dynamics for modeling two-dimensional turbulent flows.  ...  Classical methods for computational fluid dynamics (CFD), such as finite differences, finite volumes, finite elements and pseudo-spectral methods, are only accurate if fields vary smoothly on the mesh,  ... 
arXiv:2102.01010v1 fatcat:rp75forirfe43c5rtifo2qbonm

HONEI: A collection of libraries for numerical computations targeting multiple processor architectures

Danny van Dyk, Markus Geveler, Sven Mallach, Dirk Ribbrock, Dominik Göddeke, Carsten Gutwenger
2009 Computer Physics Communications  
We demonstrate the flexibility and performance of our approach with two test applications, a Finite Element multigrid solver for the Poisson problem and a robust and fast simulation of shallow water waves  ...  HONEI abstracts the hardware, and applications written on top of HONEI can be executed on a wide range of computer architectures such as CPUs, GPUs and the Cell processor.  ...  Acknowledgements Parts of this work were supported by the German Science Foundation (DFG), projects TU102/22-1 and TU102/22-2. We thank all participants of PG512 at TU Dortmund for initial support.  ... 
doi:10.1016/j.cpc.2009.04.018 fatcat:aodts4clsvapfhwpwizmot4rha

Supporting sets of arbitrary connections on iWarp through communication context switches

Anja Feldmann, Thomas M. Stricker, Thomas E. Warfel
1993 Proceedings of the fifth annual ACM symposium on Parallel algorithms and architectures - SPAA '93  
Looking at basic communication patterns as well as patterns generated by an iterative finite element PDE solver, we compare ConSet's performance (using the compiler's schedules) to that of message passing  ...  In this paper we introduce the ConSet communication model for distributed memory parallel computers.  ...  Shewchukfor assistancewith the FEM application, Gary Miller for general guidance, Eric Schwabe for early work on the communication compiler, and Thomas Gross, David O'Hallaron, and the rest of the CMU/  ... 
doi:10.1145/165231.165257 dblp:conf/spaa/FeldmannSW93 fatcat:yzuebwuthjfbbfdnxanozq5sie
« Previous Showing results 1 — 15 out of 21,231 results