LCRC Presentations

LCRC Annual Reports

2007 PDF: (11M)

2006 PDF: (2.3M)

2005 PDF: (14M)

2004 PDF: (2.3M)

PDF (5.4M): Slides

Feedback form

PETSc, the Portable, Extensible Toolkit for Scientific computation, provides sets of tools for the parallel (as well as serial), numerical solution of PDEs that require solving large-scale, sparse nonlinear systems of equations. It includes nonlinear and linear equation solvers that employ a variety of Newton techniques and Krylov subspace methods. PETSc provides several parallel sparse matrix formats, including compressed row, block compressed row, and block diagonal storage. It is designed to facilitate extensibility. Thus, users can incorporate customized solvers and data structures when using the package. It also provides an interface to several external software packages including ADIC/ADIFOR, BlockSolve95, Hypre, ParMeTis, SUNDIALS, SuperLU, and SPAI. PETSc is fully usable from Fortran, C and C++, and is portable to UNIX and Windows. See http://www.mcs.anl.gov/petsc/ for more details.

PDF (1.8M): Slides

The tutorial discussed MPI topics such as advanced point-to-point operations and tuning, performance tools, collective operations and parallel I/O.

Part 1: Introduction to Concepts and Methods:
PDF (466k): Black & White version
PDF (1.3M): Full Color version

Part 2: Advanced Concepts.
PDF (672k): Black & White version
PDF (1.6M): Full Color version

The tutorial covers common terminology, basic parallel hardware models, and parallel programming models. We discussed why one might consider a parallel code solution, and explained some possible starting points for moving to parallel code. We also discussed libraries that can help in parallelizing code such as PETsc and parallel I/O libraries HDF5 and pNetCDF.

1. A basic introduction to profiling and performance tuning for scientific applications. The focus will be parallel applications, but, certain material is applicable to either serial or parallel codes. We will provide an overview of effective use of the more common open source profiling applications including: GProf, Valgrind, FPMPI and Jumpshot. In addition, we will cover broadly applicable performance improvement techniques, first for single CPU performance and then for MPI applications.

2. A two hour hands-on tutorial limited attendance. Attendees were required to have: the source code to their functioning application; familiarity with the build process; a small simulation they can run with the application; a Jazz account. LCRC application engineers Katherine Riley and Mike Dvorak provided hands-on assistance in applying the profiling tools, understanding output from the tools and, if time, guiding the application of certain optimizations. The intenstion of the hands-on session was to help the attendees to understand how they can apply the tools to their application and how to start applying the optimizations to their codes.

Jazz Introduction [ PDF: 56K/ HTML ]

MPI Introduction [ PPT: 707K ]

1. Introduction to the LCRC Jazz cluster, covering accounts and projects, accessing the cluster, compiling MPI code, and submitting jobs. Presented by the LCRC Staff, Mike Dvorak, Katherine Riley and Susan Coghlan.

2. Introduction to MPI, covering what MPI and MPICH are, how to use them and developing simple MPICH programs. Presented by Dr. Rusty Lusk. Slides for this section will be posted as soon as they become available.