PRACE Summer School 2013 - Frameworks for Scientific Computing on Supercomputers

Jun 17, 2013, 8:00 AM → Jun 21, 2013, 3:00 PM CET

Lower area of Vitkovice, Ostrava, Czech Republic,

Introduction

This five-day event will offer tutorials on several distinguished software packages, frameworks and libraries widely used for solving large-scale engineering and scientific problems on high performance computing systems. The target audience consists of scientists and developers who wish to use high-level tools but at the same time require fine-grain control over the solution process. The presented open source packages were selected to cover the whole "life cycle" of such projects: preprocessing, solution and postprocessing. However, the focus is on the solution phase.

About the programme

The school will start with a tutorial on the libMesh library for the numerical simulation of partial differential equations using arbitrary unstructured discretizations on serial and parallel platforms, providing support for the adaptive mesh refinement. For the solution of the arising linear systems, the second and third day will cover the PETSc or Trilinos frameworks which are collections of open source software packages used as building blocks for the development of parallel and scalable applications based on the solution of partial differential equations. The fourth day will provide a more in-depth coverage of linear algebra algorithms for emerging computer architectures (multicore, GPU, heterogeneous, hybrid), as they are being developed in the Plasma and Magma projects. The last day, devoted to postprocessing, will deliver a tutorial on ParaView, a highly respected data analysis and visualization tool designed to run on massively parallel platforms.

The tutorials will be presented by outstanding specialists, mostly leaders or members of the tool/library developers. To get the most out of the school, the lectures will be accompanied by hands-on exercises ("labs") to be carried out on a supercomputer.

Prerequisites and Registration

Applicants are expected to be active developers/programmers of scientific applications for large-scale mathematical modelling and simulations. Thus, the prerequisites include working skills in (parallel) programming and some knowledge of basic discretization techniques.

School attendees will be selected based on applications submitted through the registration form, taking into account the order of registrations. Applicants will be informed about their acceptance not later than two weeks after the registration has closed. The number of attendees is limited to about forty. All attendees are encouraged to bring a poster on their work related to the topics of the school.

Remarks

The school is offered free of charge to students, researchers and academics residing in PRACE member states and eligible countries. Lunches, coffee breaks and social events are included. It is the responsibility of the attendees to arrange and cover travel and accommodation. The school's official language is English.

Hotel Mercure Accommodation request form HotelMercure-ReservationForm.pdf

List of Participants Participants-final.pdf

Programme Programme-poster.pdf

Slides http://www.training.prace-ri.eu/events/?tx_typopresscalendar_pi1[uid]=99&no_cache=1

Travel Information TravelInfo00.pdf

Monday, June 17

8:00 AM Registration

Welcome and PRACE introduction

libMesh: 1

1 libMesh

This tutorial will cover the fundamental design principles and usage of the libMesh finite element library. Key data structures and algorithms will be presented, along with a series of example programs of increasing complexity. Domain discretization, finite element approximation, and adaptive solution strategies will be discussed for a range of physical applications. Implementation examples will demonstrate the hybrid multilevel parallel capabilities of the software library. Interfaces to high-quality external software packages, such as PETSc and Trilinos, are a key component of the libMesh functionality and will also be hilighted. Benjamin Kirk is an engineer and NASA's Lyndon B. Johnson Space Center in Houston, USA. His doctoral work was on adaptive methods for flow and transport problems on parallel computers. His research is in numerical methods for chemically reacting flows, and in the design and implementation of adaptive numerical methods on modern computer architectures.

10:30 AM Coffee break

libMesh: 2

12:30 PM Lunch

libMesh: 3

3:30 PM Coffee break

libMesh: 4

5:30 PM Welcome reception and U6 - The Small World of Technology

Tuesday, June 18

PETSc: 1

2 PETSc

This tutorial will present scalable linear and nonlinear solver methodology within the context of the Portable Extensible Toolkit for Scientific computing (PETSc). It will cover the underlying principles of Krylov, domain decomposition, and multigrid methods, as well as field-split methods for coupled multiphysics problems. Throughout the tutorial, library interfaces will be presented, along with opportunities for solver composition, convergence diagnostics, and performance considerations for modern hardware. Jed Brown is an Assistant Computational Scientist at Argonne National Lab, USA. His doctoral work was at ETH Zurich in computational methods for ice sheet simulation. His research is in robust hardware-adapted, algorithmically optimal, multiphysics-capable composable algorithms and numerical software.

10:30 AM Coffee break

PETSc: 2

12:30 PM Lunch & Poster session

PETSc: 3

3:30 PM Coffee break

PETSc: 4

5:30 PM Bar hopping (optional, at your own cost)

Wednesday, June 19

Trilinos: 1

3 Trilinos

This tutorial will make you familiar with the high-performance numerical toolkit Trilinos. You will get insight into the core capabilities of Trilinos and some of its over more than 50 packages, covering almost every aspect of numerical computation. On the low level, we will discuss data organization and the architecture of Trilinos, on a higher level about numerical methods such as Krylov and LU solvers for linear problems, an array of nonlinear solvers, optimization packages, multigrid preconditioners, and discretization capabilties. Nico Schlömer is Assistant Scientist at TU Berlin, Germany, and member of the Optimal Control research group. He studied mathematics in Dresden, Germany, and Auckland, New Zealand, and did his doctoral work on numerical methods for nonlinear Schrödinger equations at the University of Antwerp, Belgium. His main research interests are in numerical linear algebra and nonlinear problems; particularly large, coupled systems.

10:30 AM Coffee break

Trilinos: 2

12:30 PM Lunch

Trilinos: 3

3:30 PM Coffee break

Trilinos: 4

5:30 PM Brewery excursion and dinner (by bus)

Thursday, June 20

Plasma & Magma: 1

4 Magma & Plasma

This tutorial will present different libraries made by the Innovative Computing Laboratory at University of Tennessee to address the problem of solving large dense linear problems on today's heterogeneous architectures. The course will first give reminders on dense linear algebra algorithms and the PLASMA library that targets shared memory multi-core architectures, will be presented as a way to introduce to dataflow programming models. Then, as of today the main concern is on accelerators, the focus will be on the MAGMA library and the development of GPU kernels. Finally, we will extend those to libraries to distributed heterogeneous architectures thanks to the PaRSEC runtime. Mathieu Faverge is an Associate Professor at IPB - ENSEIRB-Matmeca at Bordeaux, France and is part of the Inria HiePACS team. Before that, he received a PhD degree in Computer Science from the University of Bordeaux 1, France and has been a Post Doctoral Research Associate at the University of Tennessee Knoxville's Innovative Computing Laboratory for three years. His main research interests are numerical linear algebra algorithms for sparse and dense problems on massively parallel architectures, and especially DAG algorithms relying on dynamic schedulers. He has experience with hierarchical shared memory, heterogeneous and distributed systems, and his contributions to the scientific community include efficient linear algebra algorithms for those systems.

10:30 AM Coffee break

Plasma & Magma: 2

12:30 PM Lunch & Blast furnace excursion (in two groups)

Plasma & Magma: 3

4:00 PM Coffee break

Plasma & Magma: 4

6:00 PM Bar hopping (optional, on your own cost)

Friday, June 21

ParaView: 1

5 ParaView

This tutorial will present visualization tool Paraview. Key data structures and basic function for visualization of large data sets will be presented. Basic structure and usage of Paraview will be covered within this tutorial. Different visualization algorithms such as isosurfaces, streamlines and glyphs on different data types such as structured grids, unstructured grids and polygonal data will be presented to give overview of visualization capabilities of Paraview. Short intro into visualization of large data sets using parallel capabilities of Paraview will be also included. Tomas Karasek is a senior researcher at IT4Innovations VSB-Technical University of Ostrava and he is a part of Supercomputing for Industry (SC4I) team. He received his PhD degree in Applied mechanics from the VSB-Technical University of Ostrava. After obtaining his degree he worked for 7 years as a scientist and later on as a deputy department director at Institute of High Performance Computing in Singapore. His main research interest is in Computational Structure Dynamics, Computational Fluid Dynamics and Fluid Structure Interaction mainly for their application for solving industrial problems.

10:30 AM Coffee break

ParaView: 2

Closing of the PRACE Summer School 2013

1:00 PM Lunch