Skip to navigation Skip to content
Careers | Phone Book | A - Z Index
Performance and Algorithms Research

SciDAC-3

Researchers from FTG are engaged in a number of activities in the Scientific Discovery through Advanced Computing (SciDAC) initiative. The SciDAC program was initiated in 2001 in order to develop the Scientific Computing Software and Hardware Infrastructure needed to advance scientific discovery using supercomputers. As supercomputers continuously evolve, direct engagement of computer scientists and applied mathematicians with the scientists of targeted application domains becomes ever more necessary for taking full advantage of these new systems. In this regard, SciDAC is a partnership involving all of the Department of Energy (DOE) Office of Science (SC) programs - Advanced Scientific Computing Research (ASCR), Basic Energy Sciences (BES), Biological and Environmental Research (BER), Fusion Energy Sciences (FES), High-Energy Physics (HEP) and Nuclear Physics (NP) - to dramatically accelerate progress in scientific computing that delivers breakthrough scientific results through partnerships comprised of applied mathematicians, computer scientists, and scientists from other disciplines.

Researchers from FTG leverage their skills in auto-tuning, optimization, and modeling to improve the performance of the respective applications. Common techniques and solutions are centralized and encapsulated through the Institute for Sustained Performance, Energy, and Resilience (SUPER). The table below summarizes the FTG personnel involved with each of the SciDAC partnerships.  

SUPER - Institute for Sustained Performance, Energy, and Resilience
Atmospheric and Ocean Climate Modeling Greenland and Antarctic Ice Sheet Modeling Quantum Chromodynamics Chemistry
(LibTensor)
Chemistry (NWChem)
Roofline Toolkit
Leonid Oliker
Abhinav Sarje
Samuel Williams

 

Samuel Williams

Samuel Williams
Hongzhang Shan

 

Samuel Williams
Khaled Ibrahim
Leonid Oliker
 Hongzhang Shan

Samuel Williams

 

Publications

2018

Hongzhang Shan, Samuel Williams, Calvin W. Johnson, "Improving MPI Reduction Performance for Manycore Architectures with OpenMP and Data Compression", Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS), November 2018,

2017

Samuel Williams, Introduction to the Roofline Model, Roofline Training, November 2017,

Philip C. Roth, Hongzhang Shan, David Riegner, Nikolas Antolin, Sarat Sreepathi, Leonid Oliker, Samuel Williams, Shirley Moore, Wolfgang Windl, "Performance Analysis and Optimization of the RAMPAGE Metal Alloy Potential Generation Software", SIGPLAN International Workshop on Software Engineering for Parallel Systems (SEPS), October 2017,

Bryce Adelstein Lelbach, Hans Johansen, Samuel Williams, "Simultaneously Solving Swarms of Small Sparse Systems on SIMD Silicon", Parallel and Distributed Scientific and Engineering Computing (PDSEC), June 2017,

Hongzhang Shan, Samuel Williams, Calvin Johnson, Kenneth McElvain, "A Locality-based Threading Algorithm for the Configuration-Interaction Method", Parallel and Distributed Scientific and Engineering Computing (PDSEC), June 2017,

Thorsten Kurth, William Arndt, Taylor Barnes, Brandon Cook, Jack Deslippe, Doug Doerfler, Brian Friesen, Yun (Helen) He, Tuomas Koskela, Mathieu Lobet, Tareq Malas, Leonid Oliker, Andrey Ovsyannikov, Samuel Williams, Woo-Sun Yang, and Zhengji Zhao, "Analyzing Performance of Selected NESAP Applications on the Cori HPC System", Intel Xeon Phi Users Group (IXPUG), June 2017,

Brandon Cook, Thorsten Kurth, Brian Austin, Samuel Williams, Jack Deslippe, "Performance Variability on Xeon Phi", Intel Xeon Phi Users Group (IXPUG), June 2017,

Khaled Z. Ibrahim, Evgeny Epifanovsky, Samuel Williams, Anna I. Krylov, "Cross-scale efficient tensor contractions for coupled cluster computations through multiple programming model backends", Journal of Parallel and Distributed Computing (JPDC), February 2017, doi: 10.1016/j.jpdc.2017.02.010

2016

Taylor Barnes, Brandon Cook, Jack Deslippe, Douglas Doerfler, Brian Friesen, Yun (Helen) He, Thorsten Kurth, Tuomas Koskela, Mathieu Lobet, Tareq Malas, Leonid Oliker, Andrey Ovsyannikov, Abhinav Sarje, Jean-Luc Vay, Henri Vincenti, Samuel Williams, Pierre Carrier, Nathan Wichmann, Marcus Wagner, Paul Kent, Christopher Kerr, John Dennis, "Evaluating and Optimizing the NERSC Workload on Knights Landing", Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS), November 2016,

Hasan Metin Aktulga, Md. Afibuzzaman, Samuel Williams, Aydın Buluc, Meiyue Shao, Chao Yang, Esmond G. Ng, Pieter Maris, James P. Vary, "A High Performance Block Eigensolver for Nuclear Configuration Interaction Calculations", IEEE Transactions on Parallel and Distributed Systems (TPDS), November 2016, doi: 10.1109/TPDS.2016.2630699

Zhaoyi Meng, Alice Koniges, Yun (Helen) He, Samuel Williams, Thorsten Kurth, Brandon Cook, Jack Deslippe, and Andrea L. Bertozzi, "OpenMP Parallelization and Optimization of Graph-Based Machine Learning Algorithms", 12th International Workshop on OpenMP (iWOMP), October 2016, doi: 10.1007/978-3-319-45550-1_2

Pieter Ghysels, Xiaoye S. Li, François-Henry Rouet, Samuel Williams, Artem Napov, "An Efficient Multicore Implementation of a Novel HSS-Structured Multifrontal Solver Using Randomized Sampling", SIAM J. Sci. Comput. 38-5, pp. S358-S384, October 2016, doi: 10.1137/15M1010117

Khaled Z. Ibrahim, Evgeny Epifanovsky, Samuel Williams, Anna I. Krylov, "Cross-scale Efficient Tensor Contractions for Coupled Cluster Computations Through Multiple Programming Model Backends (tech report version)", LBNL. - Report Number: LBNL-1005853, July 1, 2016, LBNL 1005853, doi: 10.2172/1274416

Douglas Doerfer, Jack Deslippe, Samuel Williams, Leonid Oliker, Brandon Cook, Thorsten Kurth, Mathieu Lobet, Tareq Malas, Jean-Luc Vay, and Henri Vincenti, "Applying the Roofline Performance Model to the Intel Xeon Phi Knights Landing Processor", Intel Xeon Phi User Group Workshop (IXPUG), June 2016,

Abhinav Sarje, Douglas W. Jacobsen, Samuel W. Williams, Todd Ringler, Leonid Oliker, "Exploiting Thread Parallelism for Ocean Modeling on Cray XC Supercomputers", Cray User Group (CUG), London, UK, May 2016,

Abhinav Sarje, Exploiting Thread Parallelism for Ocean Modeling on Cray XC Supercomputers, Cray Users Group (CUG), May 12, 2016,

J. R. Jones, F.-H. Rouet, K. V. Lawler, E. Vecharynski, K. Z. Ibrahim, S. Williams, B. Abeln, C. Yang, C. W. McCurdy, D. J. Haxton, X. S. Li, T. N. Rescigno, "An efficient basis set representation for calculating electrons in molecules", Journal of Molecular Physics, 2016, doi: 10.1080/00268976.2016.1176262

The method of McCurdy, Baertschy, and Rescigno, J. Phys. B, 37, R137 (2004) is generalized to obtain a straightforward, surprisingly accurate, and scalable numerical representation for calculating the electronic wave functions of molecules. It uses a basis set of product sinc functions arrayed on a Cartesian grid, and yields 1 kcal/mol precision for valence transition energies with a grid resolution of approximately 0.1 bohr. The Coulomb matrix elements are replaced with matrix elements obtained from the kinetic energy operator. A resolution-of-the-identity approximation renders the primitive one- and two-electron matrix elements diagonal; in other words, the Coulomb operator is local with respect to the grid indices. The calculation of contracted two-electron matrix elements among orbitals requires only O(N log(N)) multiplication operations, not O(N^4), where N is the number of basis functions; N = n^3 on cubic grids. The representation not only is numerically expedient, but also produces energies and properties superior to those calculated variationally. Absolute energies, absorption cross sections, transition energies, and ionization potentials are reported for one- (He^+, H_2^+ ), two- (H_2, He), ten- (CH_4) and 56-electron (C_8H_8) systems.

The method of McCurdy, Baertschy, and Rescigno, J. Phys. B, 37, R137 (2004) is generalized to obtain a straightforward, surprisingly accurate, and scalable numerical representation for calculating the electronic wave functions of molecules. It uses a basis set of product sinc functions arrayed on a Cartesian grid, and yields 1 kcal/mol precision for valence transition energies with a grid resolution of approximately 0.1 bohr. The Coulomb matrix elements are replaced with matrix elements obtained from the kinetic energy operator. A resolution-of-the-identity approximation renders the primitive one- and two-electron matrix elements diagonal; in other words, the Coulomb operator is local with respect to the grid indices. The calculation of contracted two-electron matrix elements among orbitals requires only O(N log(N)) multiplication operations, not O(N^4), where N is the number of basis functions; N = n^3 on cubic grids. The representation not only is numerically expedient, but also produces energies and properties superior to those calculated variationally. Absolute energies, absorption cross sections, transition energies, and ionization potentials are reported for one- (He^+, H_2^+ ), two- (H_2, He), ten- (CH_4) and 56-electron (C_8H_8) systems.The method of McCurdy, Baertschy, and Rescigno, J. Phys. B, 37, R137 (2004) is generalized to obtain a straightforward, surprisingly accurate, and scalable numerical representation for calculating the electronic wave functions of molecules. It uses a basis set of product sinc functions arrayed on a Cartesian grid, and yields 1 kcal/mol precision for valence transition energies with a grid resolution of approximately 0.1 bohr. The Coulomb matrix elements are replaced with matrix elements obtained from the kinetic energy operator. A resolution-of-the-identity approximation renders the primitive one- and two-electron matrix elements diagonal; in other words, the Coulomb operator is local with respect to the grid indices. The calculation of contracted two-electron matrix elements among orbitals requires only O(N log(N)) multiplication operations, not O(N^4), where N is the number of basis functions; N = n^3 on cubic grids. The representation not only is numerically expedient, but also produces energies and properties superior to those calculated variationally. Absolute energies, absorption cross sections, transition energies, and ionization potentials are reported for one- (He^+, H_2^+ ), two- (H_2, He), ten- (CH_4) and 56-electron (C_8H_8) systems.

 

The method of McCurdy, Baertschy, and Rescigno, J. Phys. B, 37, R137 (2004) is generalized to obtain a straightforward, surprisingly accurate, and scalable numerical representation for calculating the electronic wave functions of molecules. It uses a basis set of product sinc functions arrayed on a Cartesian grid, and yields 1 kcal/mol precision for valence transition energies with a grid resolution of approximately 0.1 bohr. The Coulomb matrix elements are replaced with matrix elements obtained from the kinetic energy operator. A resolution-of-the-identity approximation renders the primitive one- and two-electron matrix elements diagonal; in other words, the Coulomb operator is local with respect to the grid indices. The calculation of contracted two-electron matrix elements among orbitals requires only O(N log(N)) multiplication operations, not O(N^4), where N is the number of basis functions; N = n^3 on cubic grids. The representation not only is numerically expedient, but also produces energies and properties superior to those calculated variationally. Absolute energies, absorption cross sections, transition energies, and ionization potentials are reported for one- (He^+, H_2^+ ), two- (H_2, He), ten- (CH_4) and 56-electron (C_8H_8) systems.The method of McCurdy, Baertschy, and Rescigno, J. Phys. B, 37, R137 (2004) is generalized to obtain a straightforward, surprisingly accurate, and scalable numerical representation for calculating the electronic wave functions of molecules. It uses a basis set of product sinc functions arrayed on a Cartesian grid, and yields 1 kcal/mol precision for valence transition energies with a grid resolution of approximately 0.1 bohr. The Coulomb matrix elements are replaced with matrix elements obtained from the kinetic energy operator. A resolution-of-the-identity approximation renders the primitive one- and two-electron matrix elements diagonal; in other words, the Coulomb operator is local with respect to the grid indices. The calculation of contracted two-electron matrix elements among orbitals requires only O(N log(N)) multiplication operations, not O(N^4), where N is the number of basis functions; N = n^3 on cubic grids. The representation not only is numerically expedient, but also produces energies and properties superior to those calculated variationally. Absolute energies, absorption cross sections, transition energies, and ionization potentials are reported for one- (He^+, H_2^+ ), two- (H_2, He), ten- (CH_4) and 56-electron (C_8H_8) systems.

2015

Hongzhang Shan, Kenneth McElvain, Calvin Johnson, Samuel Williams, W. Erich Ormand, "Parallel Implementation and Performance Optimization of the Configuration-Interaction Method", Supercomputing (SC), November 2015, doi: 10.1145/2807591.2807618

Alex Druinsky, Pieter Ghysels, Xiaoye S. Li, Osni Marques, Samuel Williams, Andrew Barker, Delyan Kalchev, Panayot Vassilevski, "Comparative Performance Analysis of Coarse Solvers for Algebraic Multigrid on Multicore and Manycore Architectures", International Conference on Parallel Processing and Applied Mathematics (PPAM), September 6, 2015, doi: 10.1007/978-3-319-32149-3_12

Abhinav Sarje, Parallel Performance Optimizations on Unstructured Mesh-Based Simulations, International Conference on Computational Science, June 2015,

Abhinav Sarje, Sukhyun Song, Douglas Jacobsen, Kevin Huck, Jeffrey Hollingsworth, Allen Malony, Samuel Williams, and Leonid Oliker, "Parallel Performance Optimizations on Unstructured Mesh-Based Simulations", Procedia Computer Science, 1877-0509, June 2015, 51:2016-2025, doi: 10.1016/j.procs.2015.05.466

This paper addresses two key parallelization challenges the unstructured mesh-based ocean modeling code, MPAS-Ocean, which uses a mesh based on Voronoi tessellations: (1) load imbalance across processes, and (2) unstructured data access patterns, that inhibit intra- and inter-node performance. Our work analyzes the load imbalance due to naive partitioning of the mesh, and develops methods to generate mesh partitioning with better load balance and reduced communication. Furthermore, we present methods that minimize both inter- and intra- node data movement and maximize data reuse. Our techniques include predictive ordering of data elements for higher cache efficiency, as well as communication reduction approaches. We present detailed performance data when running on thousands of cores using the Cray XC30 supercomputer and show that our optimization strategies can exceed the original performance by over 2×. Additionally, many of these solutions can be broadly applied to a wide variety of unstructured grid-based computations.

Hongzhang Shan, Samuel Williams, Wibe de Jong, Leonid Oliker, "Thread-Level Parallelization and Optimization of NWChem for the Intel MIC Architecture", Programming Models and Applications for Multicores and Manycores (PMAM), February 2015,

Thorsten Kurth, Andrew Pochinsky, Abhinav Sarje, Sergey Syritsyn, Andre Walker-Loud, "High-Performance I/O: HDF5 for Lattice QCD", arXiv:1501.06992, January 2015,

Practitioners of lattice QCD/QFT have been some of the primary pioneer users of the state-of-the-art high-performance-computing systems, and contribute towards the stress tests of such new machines as soon as they become available. As with all aspects of high-performance-computing, I/O is becoming an increasingly specialized component of these systems. In order to take advantage of the latest available high-performance I/O infrastructure, to ensure reliability and backwards compatibility of data files, and to help unify the data structures used in lattice codes, we have incorporated parallel HDF5 I/O into the SciDAC supported USQCD software stack. Here we present the design and implementation of this I/O framework. Our HDF5 implementation outperforms optimized QIO at the 10-20% level and leaves room for further improvement by utilizing appropriate dataset chunking.

2014

Khaled Z. Ibrahim, Samuel W. Williams, Evgeny Epifanovsky, Anna I. Krylov, "Analysis and Tuning of Libtensor Framework on Multicore Architectures", High Performance Computing Conference (HIPC), December 2014,

Alex Druinsky, Brian Austin, Sherry Li, Osni Marques, Eric Roman, Samuel Williams, "A Roofline Performance Analysis of an Algebraic Multigrid Solver", Supercomputing (SC), November 2014,

Yu Jung Lo, Samuel Williams, Brian Van Straalen, Terry J. Ligocki, Matthew J. Cordery, Leonid Oliker, Mary W. Hall, "Roofline Model Toolkit: A Practical Tool for Architectural and Program Analysis", Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS), November 2014, doi: 10.1007/978-3-319-17248-4_7

Hongzhang Shan, Samuel Williams, Wibe de Jong, Leonid Oliker, "Thread-Level Parallelization and Optimization of NWChem for the Intel MIC Architecture", LBNL Technical Report, October 2014, LBNL 6806E,

W.A. de Jong, L. Lin, H. Shan, C. Yang and L. Oliker, "Towards modelling complex mesoscale molecular environments", International Conference on Computational and Mathematical Methods in Science and Engineering (CMMSE), 2014,

H. M. Aktulga, A. Buluc, S. Williams, C. Yang, "Optimizing Sparse Matrix-Multiple Vector Multiplication for Nuclear Configuration Interaction Calculations", International Parallel and Distributed Processing Symposium (IPDPS 2014), May 2014, doi: 10.1109/IPDPS.2014.125

2013

Hongzhang Shan, Brian Austin, Wibe de Jong, Leonid Oliker, Nick Wright, Edoardo Apra, "Performance Tuning of Fock Matrix and Two Electron Integral Calculations for NWChem on Leading HPC Platforms", Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS), November 2013, doi: 10.1007/978-3-319-10214-6_13

Abhinav Sarje, Samuel Williams, David H. Bailey, "MPQC: Performance analysis and optimization", LBNL Technical Report, February 2013, LBNL 6076E,

2011

Samuel Williams, Oliker, Carter, John Shalf, "Extracting ultra-scale Lattice Boltzmann performance via hierarchical and distributed auto-tuning", Proceedings of the International Conference on High Performance Computing, Networking, Storage and Analysis (SC), New York, NY, USA, ACM, January 2011, 55, doi: 10.1145/2063384.2063458