ESSEX-II – Equipping Sparse Solvers for Exascale

ESSEX project documentation

Inhalt

Project activities

Workshops, Symposia, Seminars

2018

2017

  • Computing Bulks of Inner Eigenpairs of Large Sparse Matrices: From Applications and Algorithms to Performance and Software Engineering. Two-part minisymposium at PASC 2017, the Platform for Advanced Scientific Computing Conference, Lugano, Switzerland, June 26-28, 2017.

2016

ESSEX-II Coding Week

Participants of the ESSEX-II Coding Week in September 2016

2015

2014

  • Highly Scalable Preconditioners for Sparse Eigensolvers with Focus on Finding Interior Eigenpairs. Minisymposium at PMAA14, the 8th International Workshop on Parallel Matrix Algorithms and Applications. July 2-4, 2014, Università della Svizzera italiana, Lugano, Switzerland
  • SPPEXA PhD Seminar. Weekly seminar during summer term 2014, jointly organized by PIs of the SPPEXA projects ESSEX, ExaStencils, and TerraNeo.

2013

Talks

2018

  • Jonas Thies: Holistic Performance Engineering for Sparse Iterative Solvers. PASC MS05, Basel, Switzerland, July 2-4, 2018.
  • Georg Hager: Performance Engineering – Why and How? PASC MS05, Basel, Switzerland, July 2-4, 2018.
  • Christie Louis Alappat: RACE: Recursive Algebraic Coloring Engine. PASC MS23, Basel, Switzerland, July 2-4, 2018.
  • Gerhard Wellein: Chebyshev Filter Diagonalization on Modern Manycore Processors and GPGPUs. ISC High Performance 2018, Frankfurt, Germany, June 24-28, 2018.
  • Georg Hager: “If it doesn’t work, we learn something.” Instructive case studies from performance engineering. Minisymposium MS29 at SIAM PP18, the 2018 Conference on Parallel Processing, March 8, 2018, Tokyo, Japan.
  • Gerhard Wellein: Performance Engineering for Sparse Linear Algebra Kernels: Navigating Between Models and Expectations. Minisymposium MS85 at SIAM PP18, the 2018 Conference on Parallel Processing, March 9, 2018, Tokyo, Japan.

2017

  • Gerhard Wellein: Performance Engineering for Scalable Sparse Eigensolvers in the DFG Project ESSEX: From basic building blocks to full scale applications. JST/CREST International Symposium on Post Petascale System Software, December 12th, 2017, Tokyo, Japan.
  • Gerhard Wellein: Performance Engineering: Welcome to the world of FLOPs, Bytes and Cycles!. 34th ASE Seminar, December 13th, 2017, The University of Tokyo, Tokyo, Japan.
  • Gerhard Wellein: Performance Engineering for HPC: Models generating insights. EoCoE Face-to-Face Meeting Autumn 2017, Toulouse, France, November 29, 2017.
  • Gerhard Wellein: Performance Engineering for HPC: Models Generating Insights. Austrian HPC Meeting 2017 (AHPC17), March 1-3, 2017, Grundlsee (Austria)
  • Gerhard Wellein: Performance Engineering for HPC: Models Generating Insights. Seminar, Faculty of Informatics, USI Lugano, March 29, 2017, Lugano (Switerzland)
  • Gerhard Wellein: Performance Engineering for HPC: Models Generating Insights. Invited Talk, General Assembly of SFB/TRR-55, June 2, 2017, Regensburg (Germany)
  • Andreas Alvermann: Interior eigenvalue and eigenvalue density computations in quantum physics applications. PASC Minisymposium “Computing Bulks of Inner Eigenpairs of Large Sparse Matrices: From Applications and Algorithms to Performance and Software Engineering” PASC’17, June 26-28, 2017, Lugano, Switzerland.
  • Tetsuya Sakurai: Development of a Scalable Parallel Eigensolver for Large-Scale Simulations and Data Analysis. SPPEXA Workshop “Parallel Programming Models – Productivity and Applications (2nd Edition)”, October 18, 2017, Université de Versailles St-Quentin-en-Yvelines, Versailles, France.
  • Kengo Nakajima: Preconditioned Iterative Solvers in ppOpen-HPC/pK-Open-HPC for ESSEX-II. SPPEXA Workshop “Parallel Programming Models – Productivity and Applications (2nd Edition)”, October 18, 2017, Université de Versailles St-Quentin-en-Yvelines, Versailles, France.
  • Georg Hager: Overview of the ESSEX-II project. SPPEXA Workshop “Parallel Programming Models – Productivity and Applications (2nd Edition)”, October 18, 2017, Université de Versailles St-Quentin-en-Yvelines, Versailles, France.
  • Faisal Shahzad: CRAFT: A library for application-level Checkpoint/Restart Automatic Fault Tolerance. SPPEXA Workshop 2017, April 6, 2017, Tokyo, Japan. Faisal_Shahzad.pdf
  • Martin Galgon: Meet the BEAST. RIKEN LSPANC Meeting, Kobe, Japan, September 29, 2017.
  • Sarah Huber: Reducing linear system size with moment based methods in the FEAST framework. RIKEN LSPANC Meeting, Kobe, Japan, September 29, 2017.
  • Georg Hager: Making sense of temporally blocked stencil performance via analytic modeling. Invited talk at the 7th AICS International Symposium, Integrated Research Center of Kobe University, Kobe, Japan, February 23-24, 2017. AICS17_Hager.pdf

2016

  • Jonas Thies: Performance Engineering for Sparse Eigensolers on Heterogenous Clusters. SIAM conference on Parallel Processing for Scientific Computing, April 12-15, 2016, Paris, France. http://elib.dlr.de/108964/1/presPP16.pdf
  • Achim Basermann: Highly scalable sparse eigensolvers for large quantum physics problems on heterogeneous computing systems. Platform for Advanced Scientific Computing Conference 2016 (PASC16), June 8-10, 2016, Lausanne, Switzerland. http://elib.dlr.de/110049/1/ESSEX_PASC16_8-10_6_16.pdf
  • Bruno Lang: Efficient subspace iteration with Chebyshev-type filtering. The 9th International Workshop on Parallel Matrix Algorithms and Applications, July 6-8, 2016, Bordeaux, France. http://elib.dlr.de/110053/1/talk-Lang.pdf

2015

2014

  • Gerhard Wellein: Sparse iterative solvers, asynchronicity and fault tolerance.
    Talk at the SPPEXA Workshop on Application Interfaces for an Exascale OS, Technical University of Dresden, Germany, December 8-12, 2014. ESSEX_FFMK_2014_FS.pdf
  • Jonas Thies: A 3D-Parallel Interior Eigenvalue Solver. Talk at SPPEXA workshop on Numerical Methods for High-Performance Computers, Dec. 1-3, 2014, Interdisciplinary Center for Scientific Computing, Heidelberg, Germany. presSPPEXA14_JT.pdf
  • Gerhard Wellein: Application level asynchronous check-pointing / restart: first experiences with GPI. Dagstuhl Seminar 14402: Resilience in Exascale Computing, Schloss Dagstuhl, Germany, September 28 – October 1, 2014. Dagstuhl_FT_2014.pdf
  • Gerhard Wellein: ESSEX – Equipping Sparse Solvers for Exascale. Talk at the SPPEXA-Workshop, Euro-Par 2014 Conference, Porto, Portugal, August 25, 2014. ESSEX-SPPEXA-EuroPar2014.pdf
  • Lukas Krämer: Control of the search space size and solving linear systems in the FEAST algorithm. Talk at PMAA14, the 8th International Workshop on Parallel Matrix Algorithms and Applications. July 2-4, 2014, Università della Svizzera italiana, Lugano, Switzerland. pmaa14_kraemer.pdf
  • Lukas Krämer: Improving robustness of the FEAST algorithm and solving eigenvalue problems from graphene nanoribbons. Talk at GAMM Jahrestagung 2014, March 10-14, 2014, Erlangen. gamm_kraemer.pdf
  • Moritz Kreutzer: Building Blocks for Sparse Linear Algebra on Heterogeneous Systems. Talk at PHSP14, the Workshop on Programming of Heterogeneous Systems in Physics. July 14-15, 2014, Friedrich-Schiller-Universität, Jena, Germany. 2014_07_Jena_Heterogeneous_Sparse_Linear_Algebra.pdf
  • Jan Treibig: Building Blocks for Sparse Linear Algebra on Heterogeneous Hardware. Talk at the Workshop on Exploiting Different Levels of Parallelism for Exascale Computing at ICS2014, the ACM International Conference on Supercomputing, June 10-14, 2014, Bavarian Academy of Sciences, Munich. ICS_WS2_Exascale.pdf
  • Jonas Thies: A Hybrid Parallel Iterative Solver for Indefinite Systems in Interior Eigenvalue Computations. Talk at PMAA14, the 8th International Workshop on Parallel Matrix Algorithms and Applications. July 2-4, 2014, Università della Svizzera italiana, Lugano, Switzerland. Thies-PMAA14.pdf
  • Faisal Shahzad: A feasibility study of checkpoint/restart as a fault tolerance technique. Seminar talk at the Chair of Operating Systems, Technical University of Dresden, May 16, 2014. 140516-Desden-seminar-talk_FT.pdf
  • Gerhard Wellein: Performance Engineering for Stencil Updates on Modern Processors.
    Talk at the Minisymposium “Optimizing Stencil-based Algorithms,” SIAM Conference on Parallel Processing for Scientific Computing (PP14) , Portland (OR), USA, Feb 18-21, 2014.

    • Gerhard Wellein: Performance Engineering for Stencil Updates on Modern Processors. Talk at the Workshop “ExaStencils 2014”, Dresden, Germany, March 31, 2014.
    • Gerhard Wellein: Performance Engineering for Stencil Updates on Modern Processors with the ECM Model. Talk at the “Computing Sciences Exascale Seminar Series,” Lawrence Berkeley National Laboratory, Berkeley (CA), USA, July 25, 2014.
  • Georg Hager: Sparse Matrix-Vector Multiplication with Wide SIMD Units: Performance Models and a Unified Storage Format. Talk at the Minisymposium on Sparse Computations on Accelerators at the SIAM Conference on Parallel Processing for Scientific Computing (PP14), Feb 18-21, Portland, OR. SELL-C-sigma.pdf

2013

  • Faisal Shahzad: PGAS implementation of Sparse Matrix Vector Multiplication and Lattice Boltzmann Method using GPI. Talk at the GASPI workshop, University of Erlangen, Dec 16, 2013. 131216_GASPI-workshop_Erlangen.pdf
  • Jonas Thies: A Hybrid Parallel Block Jacobi Davidson Method. Talk at the Workshop on New Algorithms for Exascale Computing at the University of Cologne, December 4-6, 2013. Slides: A Hybrid Parallel Block Jacobi-Davidson Method.pdf
  • Moritz Kreutzer: A unified sparse matrix storage format for heterogeneous systems. Poster and talk at the Early Research Showcase track at SC13 (Denver, CO), Nov 21, 2013. Talk: 2013-11_SC13.pdf
  • Holger Fehske: Quantum phase transition in a 1D transport model with boson affected hopping: Luttinger liquid versus charge-density-wave behavior. Talks at Universities Aachen und Hannover, November 2013
  • Holger Fehske: The Edwards Fermion-Boson model. Talk at University of Wuppertal, November 2013
  • Holger Fehske: Quantum transport within a background medium: Fluctuations versus correlations. Invited talk at the International conference on Electron correlation in nanostructures, Yalta, Ukraine, October 2013
  • Faisal Shahzad: PGAS Implementation of Sparse Matrix Vector Multiplication and Lattice Boltzmann Method Using GPI. Talk at the 7th International Comference on PGAS Programming Models (PGAS13), Edinburgh, Scotland, Oct 3-4, 2013. 131004_PGAS13_GPI-LBM-SPMVM.pdf
  • Faisal Shahzad: An Evaluation of Different I/O Techniques for Checkpoint/Restart. Talk at the Workshop on Large Scale Parallel Processing 2013 (LSPP13), Boston, MA, May 24, 2013. 130524_LSPP13_FS.pdf

Teaching

2017

2016

2015

  • Georg Hager: Model-guided performance engineering of numerical kernels. Invited talk at the meeting of the SFB Transregio 55 “Hadron Physics from Lattice QCD,” University of Wuppertal, Germany, July 10, 2015. Hager_Analytic_PM_BUW_15.pdf

2014

  • Georg Hager and Gerhard Wellein: Node-level performance engineering. Two-day PATC short course at LRZ Garching, December 4-5, 2014.
  • Georg Hager and Gerhard Wellein: Node-Level Performance Engineering. Two-day short course at the second “SPPEXA Doctoral Retreat”, Sarntal, South Tyrol, Sep 22-26, 2014.
  • Georg Hager: Node-Level Architecture and Performance Engineering. Guest lecture at the University of Wuppertal, Feb 3, 2014. Wuppertal_guest.pdf

2013

Qualification

2017

  • Christopher Zachow: Lösung dünnbesetzter Dreieckcssysteme auf modernen Mehrkernprozessoren. Bachelorarbeit, Universität zu Köln, 2017. (Gutachter: A. Klawonn & J. Thies)
  • Rebekka-Sarah Hennig: Vergleich von Partitionierungsmethoden für dünnbesetzte Matrizen, die den Hamilton-Operator von Spin-Ketten repräsentieren. Bachelorarbeit, Universität Bonn, 2017. (Gutachter: M. A. Schweitzer & J. Thies)

2016

  • Christie Louis Alappat: Implementation and Performance Engineering of the Kaczmarz Method for Parallel Systems. Master thesis, University of Erlangen-Nuremberg, 2016. MAThesis-LouisChrstie.pdf

2014

  • Georg Hager: Performance Engineering as a guiding principle for efficient implementations of algorithms in computational science. Habilitation thesis at the Ernst-Moritz-Arndt University of Greifswald. The habilitation procedure was successfully completed in November 2014.
  • Lukas Krämer: Integration based solvers for standard and generalized Hermitian eigenvalue problems. Dissertation, Bergische Universität Wuppertal, 2014.
  • Melven Röhrig-Zöllner: Parallel solution of large sparse eigenproblems using a Block-Jacobi-Davidson method. Master thesis, DLR Köln and RWTH Aachen University, 2014. MA_M_Roehrig-Zoellner.pdf