Combustion Simulation in the Exascale Era

By Nicole Hemsoth

November 13, 2014

One of the many excellent sessions at SC14 will address how the twin technologies of HPC and big data are coalescing to enable major scientific breakthroughs in the field of turbulent combustion. As part of the SC14 Technical Program, Jacqueline H. Chen, a Distinguished Member of Technical Staff at the Combustion Research Facility at Sandia National Laboratories, will be speaking on this topic as it relates to her main area of research: turbulent combustion models for engineering CFD simulations. As the founding Director of the Center for Exascale Simulation of Combustion in Turbulence (ExaCT), Chen is one of the foremost researchers developing massively parallel petascale direct numerical simulations (DNS) of turbulent combustion. Ultimately these simulations will be used to design efficient and clean engines that can run on alternative fuels. HPCwire recently interviewed Chen as a prelude to her upcoming presentation.

HPCwire: Your field requires a robust combination of tools to address big data as well as high performance computing needs. Can you give us a sense of how big data and HPC overlap and intersect in this field?

Chen: There are many synergies between big data coming from high fidelity combustion simulations and future exascale combustion simulations used to understand and predict the physics governing fuel efficiency and emissions of the cars, trucks, airplanes and stationary gas turbines that we rely on. Data-intensive combustion science relies on the analysis and management of large volumes of data, obtained from both computation and time-resolved experiments and a collaboration environment for sharing the data and software among an international modeling community. For both computation and analysis exascale technologies are needed to carry out this work.

Integrating the data analysis, uncertainty quantification and visualization with exascale combustion simulations represents an in situ end-to-end workflow that affects both data analytics and computation. In situ analysis means operating on data while it is still resident in memory on the exascale machine as opposed to the traditional method of post-analysis off-line. The sheer volume of data (petabytes) that needs to be analyzed prohibits moving the data off to persistent disk which is I/O bound and can’t keep pace with processing and memory. While combustion simulations are typically floating-point intensive, many of the analytics, for example, topological methods used to identify and track time-varying flow and combustion features, are rich in integer and branching operations. This dichotomy of algorithms stresses the placement of compute versus analytics tasks on different processors (co-processors and GPUs) and the data on different locations and types of memory. Co-design of algorithms for compute and analytics in the end-to-end workflow along with exascale machine characteristics is essential to be able to tightly integrate combustion simulations and analytics. For example, both compute and analytics face the challenge of optimizing data movement across multiple levels of cache and other levels of the memory hierarchy and optimization of communication both within and across nodes using low-latency networks.

HPCwire: What are the current limitations for large-scale combustion simulation, even with access to petascale-class supercomputers? Is it hardware capability, software code scalability, or all of these and more?

Chen: The current limitations for large-scale combustion simulation run the gamut from the underlying numerical algorithms for both compute and analytics to programming environments and runtimes that express fine-grained data and task parallelism and hide machine variabilities, to the actual configuration of the hardware. Using a method known as application codesign, the Exascale Combustion in Turbulence codesign Center, ExaCT, (see www.exactcodesign.org) and its multi-disciplinary team of computer scientists, applied mathematicians, and combustion computational scientists work closely together and with HPC vendors to optimize the system components from the hardware to all aspects in the software stack needed to enable effective combustion simulation on exascale machines.

HPCwire: What will the future generation of exascale-class combustion simulations provide that isn’t possible now? In other words, are you able to ask new questions?

Exascale computing offers the promise of enabling combustion simulations in parameter regimes relevant to next-generation combustors burning alternative fuels that are needed to provide the underlying science required to design fuel efficient, clean burning vehicles, planes, and power plants for electricity generation. High fidelity combustion simulations that are performed at relevant pressure, turbulence levels, and incorporating bio-fuel chemistry will be possible in laboratory configurations. The simulations will overlap experiments such that comparisons of simulation and experiment can be made, and the computations will provide complementary information, e.g., detailed species and velocity information, to what the experiments can measure. Future simulations will also incorporate in situ uncertainty quantification such that propagation of errors in fundamental chemical parameters, e.g., rate constants, and their effect on combustion heat release and emissions is understood. This information may help chemical engineers identify which rate constants from key elementary reactions need to be measured with greater precision.

HPCwire: In your presentation at SC14, it is expected that you will touch on issues around integration and workflow designs for more complex simulations. Data volumes will increase, computational, memory, network and other elements will improve. But what new issues will this create—and how might those in your field begin thinking about those issues now?

Chen: There are a whole host of issues related to integration and workflow designs for combustion simulation. In ExaCT, we have been considering how extreme data volumes as well as the cost of data movement require rethinking the simulation workflows and necessitating in-situ formulations where the coupled simulation codes as well as the data analytics and uncertainty quantification happen close to where the data resides, on the extreme scale system itself. As a result, it is critical that the co-design processes go beyond individual computational methods and their skeletons, to include the interactions and data exchanges between these skeletons as well as with analytics and UQ components. The goal of a so-called meta-skeleton framework is to enable co-design of end-to-end data-intensive simulations workflows that include coupled simulations as well as the uncertainty quantification and analytics components (which may have very different characteristics from the simulation codes). It enables computer scientists to reason about the rich design spaces available for the placement and scheduling of computation and data in space and time, across local, remote and accelerator cores, and at various levels of the deep memory hierarchy, all while considering performance and energy/power constraints and the associated tradeoffs. The meta-skeleton integrates skeletons of the interacting components and architecture-independent representation of their behaviors (e.g., machine independent characteristics, such as memory access patterns and communication patterns) with system level empirical as well as analytic models and simulators. It then allows users to explore combinations of algorithmic, runtime placement and scheduling, data movement patterns, and system architecture design choices (including hypothetical system designs) and evaluate relative performance and energy/power behaviors. Ongoing efforts will provide high-level description of the skeletons to capture their behaviors and their interactions to evaluate these co-design choices and design models. Ultimately, the choices could be used to develop efficient mappers used in dynamic runtime systems to map data to memory and tasks to processors on an extreme scale system.

HPCwire: What are some of the challenges for codes in your field as we look toward the exascale era?

Chen: One of the major challenges for turbulent combustion codes is more effective use of memory. Typically combustion codes are not compute bound, but rather, memory bandwidth limited. As the ratio of memory to compute shrinks moving towards exascale, and the size of combustion chemistry increases, i.e. the number of species and elementary reactions in bio-fuels and conventional petroleum fuels, it is harder to fit the large working set associated with chemistry into available memory resources on the machine. Clever approaches towards maximizing compute resources which are abundant across the machine while keeping data in memory without fissioning loops are needed. Effective ways to extract both data and task parallelism in an automatic way using just-in-time compilation and dynamic runtimes are needed to get the throughput required in combustion simulations.

HPCwire: What will be the greatest challenges you expect for combustion simulations once exascale-capable machines become the norm (so, presumably in the 2020s)?

Chen: Once exascale machine architectures are deployed in the 2020s the main challenge will be how to program these machines, that is, how to adapt legacy codes and write new application codes. Programming abstractions, runtimes and mappings for efficient operation across distributed, multi-node machines with inherent latencies are needed. Programming models that support hierarchical decomposition of data and tasks to match the machine structure and a deferred execution model to hide latencies inherent to large distributed memory machines are required. Domain specific languages are needed that enable application scientists to write readable code at a high level of abstraction and to automatically emit low level optimized kernels that are suited to specific hardware (e.g., CUDA code for GPUs) at a level of abstraction above the concrete hardware. In combustion, a combustion chemistry domain specific language compiler called SINGE was recently written by computer scientists in ExaCT that makes effective use of warp specialization to fit extremely large working sets into on-chip memory on GPUs.

HPCwire: What other topics will emerge during your talk at SC14 that might be of interest and under the radar in the description with all of this in mind?

Chen: I think this covers most of it.

Read more about Chen’s upcoming session, Big Data and Combustion Simulation, here.

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!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

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

March 18, 2024

Nvidia's latest and fastest GPU, codenamed 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…

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…

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…

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…

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

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…

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…

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…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire