Dates
October 22nd - 24th, 2024
Location
Hybrid Meeting: CSRI, 1450 Innovation Parkway SE, Albuquerque, NM 87123
Contact: Phyllis Rutka
Contact: Curtis Ober
Trilinos Background
The Trilinos Project is an effort to develop algorithms and enabling technologies within an object-oriented software framework for the solution of large-scale, complex multi-physics engineering and scientific problems. A unique design feature of Trilinos is its focus on packages. Full details of Trilinos and links to package websites can be found at the Trilinos home page.
Meeting Overview
We are excited to announce that the 2024 Trilinos User-Developer Group Meeting (TUG24) will be held from October 22nd - 24th, 2024, at Sandia National Laboratories in Albuquerque, New Mexico. This hybrid event will feature both virtual and in-person attendance options.
Agenda
- Tuesday, October 22nd (AM): Keynote Speaker and Area/Package Updates
- Wednesday, October 23rd (AM): Application and User Talks
- Thursday, October 24th (AM/PM): Developer’s Day
Morning sessions (8 AM - 12 PM MDT) are scheduled to accommodate our European colleagues. Developer’s Day may extend into the afternoon depending on the amount of material to be covered.
Presentations
We invite you to present your work related to Trilinos. Please email your presentation proposals (e.g., author(s) and title) to ccober@sandia.gov and parutka@sandia.gov. Sharing updates, experiences, and successes is a valuable part of TUG, and we encourage submissions for both Application/User Talks and Developer’s Day.
Discussion Topics
We also encourage developers and stakeholders to submit topics and questions for Developer’s Day presentations and discussions. This is an opportunity to address current issues and future directions. Please send your discussion topics to ccober@sandia.gov and parutka@sandia.gov.
Registration
Although this meeting will be hybrid (i.e., attend in-person at Sandia or online via MS Teams), we are encouraging everyone to come and join us in Albuquerque, New Mexico.
There is no registration fee, but registration is required for reporting purposes. Please complete the registration form and send it to parutka@sandia.gov (see deadlines below). After registration closes, you will receive an invite for the meetings. Please do not forward the invite as registration is necessary to attend. Click here for the registration form.
Sandians need to register, as it will help with monitoring attendance, and the reporting requirements for this meeting.
Virtual Attendance
- Register by October 21st, 2024.
In-Person Attendance
- Sensitive Countries: Contact us by August 21st, 2024, due to the 60-day lead time for paperwork.
- Non-Sandians and Non-US Citizens: Declare your intention by September 23rd, 2024.
- Non-Sandians and US Citizens: Register by October 21st, 2024.
- Sandians and Non-US Citizens: Ensure Building CSRI is included on your FNRSP and register by October 21st, 2024.
- Sandians and US Citizens: Register by October 21st, 2024.
Keynote
Tensor Decomposition of Large-Scale Data with GenTen (E. Phipps)
Tensors, or high dimensional arrays of data, are ubiquitous in science, engineering, and informatics, arising in diverse areas including multiphysics simulation, remote sensing, text analytics, and cyber security. Tensor decomposition methods, akin to matrix factorization methods for two-dimensional data, facilitate powerful analysis and reduction of such data, including data compression, surrogate modeling, pattern identification, and anomaly detection. In this talk, we describe a recently developed tool for tensor decomposition of large-scale data called GenTen. GenTen leverages several packages in Trilinos to compute tensor decompositions in parallel on modern CPU and GPU computing architectures. We describe the kinds of decompositions GenTen can compute, demonstrate the kinds of analysis it facilitates, and how GenTen uses Trilinos packages in these computations. On-going challenges and future research directions will also be discussed.
Tentative Agenda
All times in Mountain Daylight Time (MDT) Denver, U.S. (UTC-06:00).
Tuesday, October 22nd
| Opening | |
| 8:00 | Welcome (C. Ober) |
|---|---|
| 8:15 | Keynote Tensor Decomposition of Large-Scale Data with GenTen (E. Phipps) |
| Updates from Trilinos Areas | |
| 9:15 | Trilinos, HPSF, Epetra Deprecation (C. Ober, S. Browne, J. Willenbring) |
| 10:15 | Break |
| 10:30 | DevOps (a.k.a. Framework) Updates (S. Browne)
|
| 11:00 | Trilinos Core Updates (R. Pawlowski)
|
| 11:15 | Solver Updates (C. Glusa)
|
| 11:30 | Discretizations and Analysis Updates (M. Perego)
|
| 11:45 | Lunch / Adjourn |
| 1:00 | Nvidia HPC Session NVIDIA’s HPC SDK (Graham Lopez) This presentation will provide an overview of the NVIDIA HPC SDK and its role on NVIDIA HPC systems. We will look at the design and motivation behind the HPC SDK to understand how and various choices are made in the product. We’ll look at a few case studies of how support for programming models, new hardware, and new software products have shaped the past evolution of the HPC SDK, where it currently stands today, and the future directions of the HPC SDK. Finally, we invite the audience to share their thoughts, feedback, and ideas on the software. We'll explore speculative concepts and emerging trends that could shape the future of HPC application development. Nvidia Attendees (Tentative):
|
| 5:00 | Adjourn |
Wednesday, October 23rd
| Applications Session | |
| 7:50 | Welcome (C. Ober) |
|---|---|
| 8:00 | Trilinos use in 4C-multiphysics (M. Firmbach) |
| 8:30 | RAMSES impact on Sandia’s pulsed power program (D. Sirajuddin) |
| 9:00 | Build Times on Modern Heterogenous Systems: Steps towards a faster future (J. Elliot, J. Doerfert) |
| 9:30 | Multi-Region Solver and Algebraically-Transformed Block-Diagonal Preconditioner for High-Q Cavity Problems (V. Dang, R. Pfeiffer, J. Kotulski) |
| 10:00 | Break |
| 10:15 | Recent Teko Usage in Aria (M. Phillips) |
| 10:45 | On Scalable Multiphysics Block Preconditioning of an Implicit VMS Resistive MHD Formulation with Application to Magnetic Confinement Fusion J. Shadid, J. Bonilla (LANL), P. Ohm (RIKEN, Japan), E. Phillips, M. Crockatt, R. Pawlowski, R. Tuminaro, J. Hu, and X. Tang (LANL) |
| 11:15 | My Very First Large(ish) Language Model: Doing Model Inference, Fine Tuning & RAG at Sandia (C. Siefert, G. Harper) |
| 12:00 | Lunch |
| 1:00 | Trilinos Collaboration
|
| 5:00 | Adjourn |
Thursday, October 24th
| Developer Sessions | |
| 8:00 | Welcome (C. Ober) |
|---|---|
| 8:10 | Using Intrepid2 for high-order FE assembly in a multi-physics code (M. Arnst, R. Tomasetti) |
| 8:30 | AutoTester2: Modernized Trilinos CI Testing (S. Brown, A. Chauphan, J. Frye, J. LaPre) |
| 10:00 | Break |
| 10:15 | Keeping Trilinos working performantly every night, everywhere (C. Siefert, et. al.) |
| 10:45 | Belos Serial Dense Matrix Abstraction (H. Thornquist) |
| 11:15 | Hands-On with ATLAS/SHIRTY As A Trilinos Developer: Coding with LLMs at Sandia (C. Siefert, C. Glusa, J. Hu) |
| 11:45 | Lunch |
| 1:00 | ASC DevOps Research Project and Trilinos (R. Bartlett, J. Willenbring, R. Milewicz) |
| 1:30 | Software Sustainability / Stewardship CI Update (J. Willenbring) |
| 2:00 | Break (30 min.)
|
| 2:30 | Open Discussion
|
| 5:00 | Adjourn |