numerical excellence
play

Numerical Excellence Franois Cassier fcassier@nag.com Introduction - PowerPoint PPT Presentation

Numerical Excellence Franois Cassier fcassier@nag.com Introduction to Multicore Demystified lecture RWTH Aachen 18 th November 2014 Experts in numerical algorithms and HPC services Agenda Few words about NAG NAG Software


  1. Numerical Excellence François Cassier fcassier@nag.com Introduction to « Multicore Demystified » lecture RWTH Aachen 18 th November 2014 Experts in numerical algorithms and HPC services

  2. Agenda  Few words about NAG  NAG Software Products:  NAG Numerical Libraries & Toolbox  NAG Fortran Compiler  NAG License at RWTH Aachen  Services Numerical Excellence

  3. NAG Background  Founded in 1970  Not-for-profit organisation  Surpluses fund on-going R&D  Mathematical and Statistical Expertise  Libraries of numerical components  Consulting  HPC Services  Computational Science and Engineering (CSE) support  Procurement advice, market watch, benchmarking Numerical Excellence

  4. Partnerships with leading academics  University of Oxford  RWTH Aachen  Universita di Roma I  University of Birmingham  T .U. Eindhoven  Illinois Institute of  Shanghai University Technology  Stanford University  Fudan University  University of Strathclyde  K U Leuven  National University of  Kings College London Singapore  University of Manchester  University of Tennessee  New York University  Tsinghua University  Osaka City University  University of Warwick Numerical Excellence

  5. NAG SOFTWARE NAG FORTRAN COMPILER 5

  6. NAG Fortran Compiler  World’s first Fortran 90 compiler  Developed and enhanced to include Fortran 95 + close to 100% features of Fortran 2003 + partial support for the new Fortran 2008 features  Regularly updated, fully supported.  EXCELLENT (the world’s best) checking compiler Numerical Excellence

  7. NAG Fortran Builder (Windows version)  Integrated Development Environment for NAG compiler on Windows PC  Perfectly integrated with NAG Library  NAG example program templates  Extra facilities: tools e.g. Fortran Polisher, Fortran converter, LAPACK examples  Integrated debugger Numerical Excellence

  8. New Features – NAG Compiler 6.0  Support for OpenMP 3.1 in addition to OpenMP 3.0  Maximum rank of an array is increased  Internal procedures can be passed as actual arguments or assigned to a procedure pointer  G0 and G0.d edit descriptors, and * (infinite) repeat factor  CONTIGUOUS attribute and IS_CONTIGUOUS intrinsic function  ERF, EFRC, ERFC_SCALED, GAMMA, LOG_GAMMA intrinsic functions  Implied-shape named array constants  ERROR_STOP statement  Kind type parameters for derived types  New interface generator (nagfor =interfaces)  Additional options within the polisher tool Numerical Excellence

  9. NAG SOFTWARE NAG NUMERICAL LIBRARIES & TOOLBOX 9

  10. NAG development philosophy  First priority: accuracy & performance  Algorithms chosen for  usefulness  robustness  accuracy  stability  speed Numerical Excellence

  11. Why Use NAG Maths Libraries and Toolboxes?  +1700 math/stat software components  High quality software – accuracy, reliability and robustness  Extensively tested, supported and maintained code  Reduce development time  Components for:  Most programming languages  Most environments for modelling and data analysis  Designed to get high performance Numerical Excellence

  12. NAG Library and Toolbox Contents   Root Finding Dense Linear Algebra   Summation of Series Sparse Linear Algebra   Quadrature Correlation & Regression Analysis  Ordinary Differential  Equations Multivariate Methods   Partial Differential Equations Analysis of Variance   Numerical Differentiation Random Number Generators   Integral Equations Univariate Estimation   Mesh Generation Nonparametric Statistics   Interpolation Smoothing in Statistics   Curve and Surface Fitting Contingency Table Analysis   Optimization Survival Analysis   Approximations of Special Time Series Analysis Functions  Operations Research Numerical Excellence

  13. NAG Library : new Mark 24 (1 of 2) Mark 24 has new routines in many areas including...   Hypergeometric function Interpolation Robust accurate real confluent 5D interpolation and higher dimensions   hypergeometric function  Optimization  Nearest correlation matrix Local optimization: Non-negative least squares  Elementwise weighted nearest correlation  Global optimization: Multi-start versions of  matrix general nonlinear programming and least  squares routines Wavelet Transforms & FFTs  RNG’s Three dimensional discrete single level and  multi-level wavelet transforms. Brownian bridge and random fields   Fast Fourier Transforms (FFTs) for two-  Statistics dimensional and three dimensional real data. Gaussian mixture model   Matrix Operations Best subsets of given size (branch and bound )  Matrix exponentials (Schur – Parlett)  Vectoized probabilities and probability density  Combines two sums of squares matrices to  functions of distributions. allow large datasets to be summarised Inhomogeneous time series analysis, moving  Matrix functions including calculation of  averages condition numbers Numerical Excellence

  14. NAG Library : new Mark 24 (2 of 2) Mark 24 has new routines in many areas including... (page 2)  Data fitting Fit of 2D scattered data by two-stage  approximation (suitable for large datasets)  Quadrature 1D adaptive for badly-behaved integrals   Sparse eigenproblem Driver for real general matrix, driver for banded  complex eigenproblem  Sparse linear systems block diagonal pre conditioners and solvers   ODE solvers Thread safe initial value ODE solvers  Numerical Excellence

  15. NAG Library & Toolbox Products  NAG Fortran Library  NAG Library for SMP & Multi Cores  NAG Parallel Library  NAG C Library  NAG Toolbox for MATLAB  NAG Library for Xeon Phi  NAG Library for .NET  NAG Components for GPU/CUDA  + Additional interfaces (e.g. Python, Java, R) Numerical Excellence

  16. O/S & Compilers Supporting Wide Range of Support leading compilers O/S (Fortran, C & C++, .NET)  Windows  NAG  Linux  Intel, Pathscale, PGI  Mac  GCC, GFortran  Solaris  Absoft, Lahey, Visual C++  AIX  ...  HP-UX  32-bit, 64-bit  ... Numerical Excellence

  17. NAG Libraries Ease of Integration  Excel  C++ (various)  SciLab  OpenCL  Octave  Visual Basic  Mathematica  Java  Maple  Delphi  LabVIEW  Python  R and S-Plus  F#  SAS  …  …  and more  and more Numerical Excellence

  18. NAG SOFTWARE AT RWTH AACHEN 18

  19. NRW State Wide License Includes  NAG Fortran Compiler & NAG Fortran Builder  NAG Fortran Library  NAG Library for SMP & Multi Cores  NAG Parallel Library  NAG C Library  NAG Toolbox for MATLAB  NAG Library for Xeon Phi  NAG Library for .NET  NAG Components for GPU/CUDA  + Additional interfaces (e.g. Python, Java, R) Numerical Excellence

  20. How do I get the software?  Help yourself from:  http://www.nag.co.uk/downloads/index.asp  For all supported combinations of o/s and compiler  On Workstations, clusters & home systems  License keys can be requested at:  servicedesk@rz.rwth-aachen.de (Andrea Lindner Lorenz) Numerical Excellence

  21. How do I get some help?  Direct access to technical support:  support@nag.co.uk Numerical Excellence

  22. NUMERICAL & HPC SERVICES 22

  23. Consulting Services from NAG  Numerical & Statistical Library support Customised routines for NAG Libraries  Integration of NAG libraries into customer environments  Customisation of customer code   HPC support Supporting the business   HPC technology planning, collaborative research  Impartial advice, benchmarking, procurement consulting Supporting the developer/user   Parallel software engineering expertise & services  Porting, scalability, optimisation, new algorithms, restructuring, …  Product and HPC Training  Collaborative R&D for new technology (manycore , GPU, …) Experts in Numerical Engineering Numerical Excellence

  24. NAG Key Contacts www.nag.co.uk Technical Support and Help support@nag.co.uk francois.cassier@nag.com (commercial contact) viktor.mosenkis@nag.co.uk (local technical contact) NAGNews http://www.nag.co.uk/NAGNews/Index.asp Numerical Excellence

Recommend


More recommend