PETSc: Celebrating 20 Years

Panel 1: The Outer Loop: What needs to be done in algorithms, libraries, and applications to support computing sensitivities and optimizations of simulations as well as solving stochastic systems? Does this require refactoring/evolving the lower-level algorithm/software stack, i.e., PETSc? What level of abstraction should be used/exposed for the lower levels? For TS and SNES it is usually function evaluation plus Jacobian; should it be something else?

Moderator: Bill Gropp, University of Illinois at Urbana-Champaign

Panelists: Paul Bauman (SUNY Buffalo), Jed Brown (Argonne National Laboratory and University of Colorado, Boulder), Patrick Farrell (University of Oxford), Tobin Isaac (University of Texas at Austin)

Wednesday, June 17

7:30 – 8:15 Bus from Argonne Guest House to Building 240 Conference Center

8:00 – 8:30 Continental breakfast

8:30 – 10:00 (HZ)

8:30 SLEPc: Current Achievements and Plans for the Future

Jose Roman, Universitat Politècnica de València, Spain

9:00 The Immersed Boundary Method for Advection-Electrodiffusion

Philhwa Lee, University of Michigan

9:30 Gyrokinetic Particle-in-Cell Methods for Tokamak Edge Plasmas

Mark Adams, Lawrence Berkeley National Laboratory

10:00 – 10:30 Break

10:30 –12:30 (DK)

10:30 PETSc: a SWOT Analysis,

David Keyes, KAUST, Saudi Arabia

11:00 Fast Computation of Fully Resolved, Neuromechanically Driven Locomotion,

Namu Patel, Northwestern University

11:30 15 years of PFLOTRAN,

Richard Mills, Intel

12:00 PETSc on GPUs and MIC: Current Status and Future Directions

Karl Rupp, Vienna University of Technology, Austria

12:30 – 2:00 Lunch/cafeteria

2:00 – 3:45 (DK)

2:00 Panel 2: Leveraging the Community: Beyond Software Libraries

2:45 Simplifying Multiphysics Through Application Composition

Derek Gaston, Idaho National Laboratory

3:15 Scalable Parallel Solvers for Finite Elements and Isogeometric Discretizations in

Computational Cardiology

Luca Pavarino, University of Milan, Italy

3:45 Lightning Talks 2

4:15 – 4:30 Break (set up posters)

4:30 – 5:30 Poster Session 2 (see list of posters on page 5)

5:30 Adjourn (Dinner on your own)

5:30 – 6:15 Bus from Building 240 Conference Center to Argonne Guest House

Panel 2: Leveraging the Community: Beyond Software Libraries: In mathematics, one builds new results using previously published theorems and proofs; publications, including books, are a fairly good vehicle to communicate results so that they may be reused. In numerical simulations, publications are unable to convey the depth of information required to leverage previous results. Application codes are rarely usable (or even buildable) beyond their original developers and thus are a poor way of leveraging previous work. Software libraries and packages have achieved success at enabling the leveraging of previous work. What other successful strategies exist for community leverage and reuse (e.g., StackExchange), and how can the impact of these approaches be measured fairly to prevent gaming the system?

Moderator: Matthew Knepley (Rice University)

Panelists: Derek Gaston (Idaho National Laboratory), Gerard Gorman (Imperial College), Glenn Hammond (Sandia National Laboratories), Jack Poulson (Stanford University)

Thursday, June 18

7:30 – 8:15 Bus from Argonne Guest to Building 240 Conference Center

8:00 – 8:30 Continental breakfast

8:30 – 10:00 (JB)

8:30 PetIGA: A Framework for High Performance Isogeometric Analysis,

Lisandro Dalcin, KAUST, Saudi Arabia

9:00 C urrent and Planned AMR support in PETSc

Tobin Isaac, University of Texas at Austin

9:30 Solving the Load Flow and Helmholz Equations using PETSc

Domenico Lahaye, Delft University of Technology, Netherlands

10:00 – 10:30 Break

10:30 –12:00 (KR)