WarpX Code Shines at the Exascale Level

February 1, 2023

ECP-funded project takes plasma accelerator modeling to new heights.

Feb. 1, 2023 — Long valued for their role in scientific discovery and in a variety of medical and industrial applications, particle accelerators have been used in many areas of fundamental research and credited with enabling Nobel Prize–winning research in physics and chemistry. But these high-end instruments also occupy a lot of space and carry hefty price tags. Even smaller accelerators, such as those used in medical centers for proton therapy, need large spaces to accommodate their hardware, power supplies, and radiation shielding.

Fortunately, over the last several years physicists, engineers, and computational scientists have been working to create more affordable and accessible particle accelerators by shrinking both the size and the cost while increasing the capability. One of the most exciting developments in these efforts is the plasma accelerator, which uses lasers or particle beams rather than radio-frequency waves to generate the accelerating field, allowing these devices to support accelerating-electric fields many orders of magnitude greater than conventional accelerators with a much smaller footprint—even able to fit on a tabletop.

The reduced size of plasma accelerators, however, presents challenges in controlling the intricate ultrafast processes at play, which are often on the picosecond and micrometer scale. Thus, realizing their compact designs requires novel mathematical and software capabilities to enable high-performance, high-fidelity modeling that can capture the full complexity of acceleration processes over a large range of space and timescales—simulations that are often computationally intensive.

To address this issue, the WarpX project—a Lawrence Berkeley National Laboratory (Berkeley Lab)–led effort that also drew the attention and support of DOE’s Exascale Computing Project (ECP)—has spent the last 6 years creating a novel, highly parallel and highly optimized single-source simulation code for modeling plasma-based particle colliders on cutting-edge exascale supercomputers, with broad importance for other accelerators and related problems.

WarpX enables computational explorations of key physics questions in the transport and acceleration of particle beams in long chains of plasma channels, which could yield significant savings in the design and characterization of plasma-based colliders before they are built. Using exascale modeling to validate these devices could also lead to broader applications, such as sterilizing food or toxic waste, implanting ions in semiconductors, treating cancer, advancing fusion research, and developing new drugs.

“A lot of research still needs to be done and, like most things with plasma, you need big simulation tools because it’s complicated due to the large number of space and time scales. This is where WarpX comes into play,” said Jean-Luc Vay, a senior scientist at Berkeley Lab who heads the Lab’s Accelerator Modeling Program in the Applied Physics and Accelerator Technologies division, leads the WarpX development team at Berkeley Lab, and is co-PI (along with Berkeley Lab’s Ann Almgren) for the ECP WarpX development team. Outside of Berkeley Lab, the ECP WarpX development team includes collaborators from Lawrence Livermore National Laboratory and SLAC National Accelerator Laboratory.

Next-Generation PIC Solution

“The idea of WarpX is to provide a particle-in-cell (PIC) solution [a technique used to solve a class of plasma physics problems from first principles] that can tackle problems that are much bigger and faster than we could do before, allowing us to explore solutions for plasma acceleration at scales not previously possible,” Vay said. “We don’t need to do just simulations—we also need to study tolerances, which is highly computationally demanding. So we are pushing the state of the art with high performance computing.” In addition, since the PIC method is important for modeling many other beams and plasma physics problems, “the code is being designed to be generic and will advance solutions for a wide range of problems in accelerators, fusion, and more, far beyond the main focus on plasma-based colliders of the ECP WarpX project,” Vay added.

The original version of the code—dubbed simply “Warp” and developed by Alex Friedman and collaborators (including Vay and other WarpX developers) for DOE’s Fusion Energy Science program—used Fortran subroutines for fast number crunching that were wrapped in a Python top layer for simulations control and steering. While that version of the code incorporated many original, cutting-edge algorithms, like adaptive mesh refinement (AMR), it was not easily portable to run efficiently on both CPUs and GPUs; this hybrid architecture is a hallmark of emerging exascale supercomputers. As a result, “it became quickly clear that a full rewrite of the code using portable C++ primitives was the way to go for developing a single source code that runs efficiently on both platforms,” Vay said—a decision that led to one of the biggest challenges the WarpX team has faced so far.

So the team turned to AMReX, an ECP software library that provides a robust, efficient, and scalable implementation of AMR capabilities and versatile portable C++ primitives, which was developed by a team led by John Bell and Ann Almgren in the Applied Mathematics and Computational Research Division (AMCRD) at Berkeley Lab.

WarpX: longitudinal electric field in a laser-plasma accelerator rendered with the ECP software libraries Ascent & VTK-m as the simulation was running. Credit: Axel Huebl, Berkeley Lab.

“AMReX is a numerical library that helps us implement physical block-structured mesh refinement algorithms on multiple computer architectures,” said Axel Huebl, a computational physicist and research software engineer in Berkeley Lab’s Accelerator Technology and Applied Physics Division and lead developer of WarpX. “Mesh refinement enables us to focus the computational power on the most interesting parts of a simulation, while staying effective for the larger, macroscopic evolution of, for instance, the plasma physics we model.”

Toward this end, as the WarpX team simulated longer chains of plasma accelerators at higher grid resolutions, the efficiency of the simulations was limited by the numerical limitations of the existing algorithms. So new algorithms were devised that removed some specific limitations, resulting in speedups of an order of magnitude or more in some cases. And after just one year of development, the AMReX team had the enhanced WarpX code up and running on GPUs, Vay noted.

“In addition to the question of what languages and programming tools we should use, we also had to change the way a lot of the algorithms are implemented,” said Andrew Myers, a computer systems engineer in AMCRD and a member of the WarpX team who has been instrumental in the AMReX implementation. “For example, we had to redesign the way a lot of the particle algorithms in AMReX work.”

“The AMReX team did an amazing job, doing a lot of tests to determine what the solutions could be and then developing this layer that provides a single source code that allows us to compile for CPUs or NVIDIA, AMD, or Intel GPUs,” Vay said.

Four Supercomputers, a Gordon Bell Prize, and More

With support from ECP and team members’ respective institutions, these development efforts are already paying off. In simulations run in 2022, the WarpX project demonstrated a 500x improvement in performance on the exascale supercomputer Frontier over the preceding version of Warp and was the first ECP application to reach its project goal by running at scale on Frontier. It was also the first code to perform 3D simulations of laser-matter interactions on Frontier, Fugaku, and Summit—something that has so far been out of reach for standard codes.

“With WarpX we can take the same code base and compile it for all of those different machines without having to make any changes to the code, and that was a big focus point of the effort and the project,” Myers said.

“That is something that ECP really puts an emphasis on,” Huebl added. “If you are working for such a long time on a project, how do you make this development sustainable? Our codes outlive the machines most of the time, so we intentionally plan it in a way that we can continue in five years when the next machine is coming along, and we don’t have to start from scratch and can optimize the code without having to rewrite everything.”

Another key milestone for the WarpX project was the successful implementation and deployment of the code for first-of-kind mesh-refined massively parallel 3D PIC simulations of kinetic plasma optimized on 4 of the 10 fastest supercomputers in the world (Frontier, Fugaku, Summit, and Perlmutter). This accomplishment earned the WarpX development team and collaborators from France and Japan the Association for Computing Machinery’s prestigious Gordon Bell Prize in 2022.

And the WarpX project is far from over. Several activities are already under way that will further leverage the latest iteration of the code for a variety of applications:

  • WarpX is being applied to the exploration of outstanding questions for plasma-based collider designs on tens of consecutive plasma stages, toward the modeling of multi-TeV high-energy physics colliders based on tens to thousands of plasma-based accelerator stages. Novel algorithms are being studied to improve the accuracy and speed of the code for these studies, including a more versatile and robust implementation of AMR.
  • A growing number of users in research labs, academia, and industry are applying WarpX to topics such as laser-ion acceleration, structure-based wakefield acceleration, laser-plasma interaction, plasma instabilities, plasma mirrors, fusion devices, magnetic fusion sheaths, magnetic reconnection, pulsars physics, thermionic converters, electron clouds in accelerators. Several of them are also contributing to the code testing and development.
  • The success of WarpX has prompted spinoff projects such as ImpactX, a rewrite of the popular conventional accelerator suite IMPACT (now ImpactX), and HiPACE++, a rewrite of the quasistatic code HiPACE for plasma accelerators. Both have been rewritten for CPUs and GPUs using the AMReX library and sharing data structures and modules with WarpX. Another spinoff is Artemis, which is built on top of WarpX with additional functionalities for the modeling of micromagnetics and electrodynamic waves in next-generation microelectronics.

“WarpX was built as a general code with one target application in mind, but it can do much more, and people are already exploring further applications,” Vay said.

Being part of the ECP effort has had a demonstrable impact on this body of research, he added. “The ECP project has been an incredible boost to our scientific code developments and productivity,” he said. “Many of us on the WarpX team agree that it has fostered, between WarpX, AMReX, and partners, the best collaboration experience that we have had to date in our scientific lives. In addition, ECP is enabling us to realize our vision of an integrated PIC ecosystem of codes that is part of a community of integrated scientific software.”

Beyond ECP, more work is now needed to further boost the speed and efficiency of the code, and to address the broad applications in accelerators, fusion, and more. “Make no mistake,” said Vay, “the work on WarpX is not over.” For example, continued work on novel algorithms is needed to improve the accuracy and speed of the code, including a more versatile and robust implementation of AMR.

The ECP WarpX Development Team

Berkeley Lab

  • Ann Almgren (co-PI)
  • Arianna Formenti
  • Marco Garten
  • Kevin Gott
  • Junmin Gu
  • Axel Huebl
  • Revathi Jambunathan
  • Hannah Klion
  • Prabhat Kumar
  • Rémi Lehe
  • Andrew Myers
  • Ryan Sandberg
  • Olga Shapoval
  • Maxence Thevenet (now DESY)
  • Jean-Luc Vay (PI)
  • Weiqun Zhang
  • Edoardo Zoni

Lawrence Livermore National Laboratory

  • David Grote

SLAC

  • Lixin Ge
  • Cho Ng

(Plus contributors from laboratories, universities and industry organizations in the United States, Europe, and Asia)


Source: Kathy Kincade, Lawrence Berkeley National Laboratory and ECP

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from its predecessors, including the red-hot H100 and A100 GPUs. Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. While Nvidia may not spring to mind when thinking of the quant Read more…

2024 Winter Classic: Meet the HPE Mentors

March 18, 2024

The latest installment of the 2024 Winter Classic Studio Update Show features our interview with the HPE mentor team who introduced our student teams to the joys (and potential sorrows) of the HPL (LINPACK) and accompany Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the field was normalized for boys in 1969 when the Apollo 11 missi Read more…

Apple Buys DarwinAI Deepening its AI Push According to Report

March 14, 2024

Apple has purchased Canadian AI startup DarwinAI according to a Bloomberg report today. Apparently the deal was done early this year but still hasn’t been publicly announced according to the report. Apple is preparing Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimization algorithms to iteratively refine their parameters until Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, code-named Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the fi Read more…

Survey of Rapid Training Methods for Neural Networks

March 14, 2024

Artificial neural networks are computing systems with interconnected layers that process and learn from data. During training, neural networks utilize optimizat Read more…

PASQAL Issues Roadmap to 10,000 Qubits in 2026 and Fault Tolerance in 2028

March 13, 2024

Paris-based PASQAL, a developer of neutral atom-based quantum computers, yesterday issued a roadmap for delivering systems with 10,000 physical qubits in 2026 a Read more…

India Is an AI Powerhouse Waiting to Happen, but Challenges Await

March 12, 2024

The Indian government is pushing full speed ahead to make the country an attractive technology base, especially in the hot fields of AI and semiconductors, but Read more…

Charles Tahan Exits National Quantum Coordination Office

March 12, 2024

(March 1, 2024) My first official day at the White House Office of Science and Technology Policy (OSTP) was June 15, 2020, during the depths of the COVID-19 loc Read more…

AI Bias In the Spotlight On International Women’s Day

March 11, 2024

What impact does AI bias have on women and girls? What can people do to increase female participation in the AI field? These are some of the questions the tech Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Analyst Panel Says Take the Quantum Computing Plunge Now…

November 27, 2023

Should you start exploring quantum computing? Yes, said a panel of analysts convened at Tabor Communications HPC and AI on Wall Street conference earlier this y Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Training of 1-Trillion Parameter Scientific AI Begins

November 13, 2023

A US national lab has started training a massive AI brain that could ultimately become the must-have computing resource for scientific researchers. Argonne N Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire