Scientific Computing

The Material Point Method (MPM) has been very successful in providing solutions to many challenging problems involving large deformations. The nonlinear nature of MPM makes it necessary to use a full nonlinear stability analysis to determine a stable timestep. The stability analysis of Spigler and Vianello is adapted to MPM and used to derive a stable timestep bound for a model problem. This bound is contrasted against a traditional CFL bound.

Optimizations in the petascale era require modifications of existing codes to take advantage of new architectures with large core counts and SIMD vector units. This paper examines high-level and low-level optimization strategies for numerical weather prediction (NWP) codes. These strategies employ thread-local structures of arrays (SOA) and an OpenMP directive such as OMP SIMD. These optimization approaches are applied to the Weather Research Forecasting single-moment 6-class microphysics schemes (WSM6) in the US Navy NEPTUNE system. The results of this study indicate that the high-level approach with SOA and low-level OMP SIMD improves thread and vector parallelism by increasing data and temporal locality. The modified version of WSM6 runs 70x faster than the original serial code. This improvement is about 23.3x faster than the performance achieved by Ouermi et al., and 14.9x faster than the performance achieved by Michalakes et al.

Large-scale parallel applications with complex global data dependencies beyond those of reductions pose significant scalability challenges in an asynchronous runtime system. Internodal challenges include identifying the all-to-all communication of data dependencies among the nodes. Intranodal challenges include gathering together these data dependencies into usable data objects while avoiding data duplication. This paper addresses these challenges within the context of a large-scale, industrial coal boiler simulation using the Uintah asynchronous many-task runtime system on GPU architectures. We show significant reduction in time spent analyzing data dependencies through refinements in our dependency search algorithm. Multiple task graphs are used to eliminate subsequent analysis when task graphs change in predictable and repeatable ways. Using a combined data store and task scheduler redesign reduces data dependency duplication ensuring that problems fit within host and GPU memory. These modifications did not require any changes to application code or sweeping changes to the Uintah runtime system. We report results running on the DOE Titan system on 119K CPU cores and 7.5K GPUs simultaneously. Our solutions can be generalized to other task dependency problems with global dependencies among thousands of nodes which must be processed efficiently at large scale.

Controlling the spread of infectious diseases in large populations is an important societal challenge. Mathematically, the problem is best captured as a certain class of reaction-diffusion processes (referred to as contagion processes) over appropriate synthesized interaction networks. Agent-based models have been successfully used in the recent past to study such contagion processes. We describe EpiSimdemics, a highly scalable, parallel code written in Charm++ that uses agent-based modeling to simulate disease spreads over large, realistic, co-evolving interaction networks. We present a new parallel implementation of EpiSimdemics that achieves unprecedented strong and weak scaling on different architectures — Blue Waters, Cori and Mira. EpiSimdemics achieves five times greater speedup than the second fastest parallel code in this field. This unprecedented scaling is an important step to support the long term vision of real-time epidemic science. Finally, we demonstrate the capabilities of EpiSimdemics by simulating the spread of influenza over a realistic synthetic social contact network spanning the continental United States (∼280 million nodes and 5.8 billion social contacts).

We present a new method for progressive volume rendering by accumulating object-space samples over successively rendered frames. Existing methods for progressive refinement either use image space methods or average pixels over frames, which can blur features or integrate incorrectly with respect to depth. Our approach stores samples along each ray, accumulates new samples each frame into a buffer, and progressively interleaves and integrates these samples. Though this process requires additional memory, it ensures interactivity and is well suited for CPU architectures with large memory and cache. This approach also extends well to distributed rendering in cluster environments. We implement this technique in Intel's open source OSPRay CPU ray tracing framework and demonstrate that it is particularly useful for rendering volumetric data with costly sampling functions.

More than a decade ago, Chris Johnson proposed the "Theory of Visualization" as one of the top research problems in visualization. Since then, there have been several theory-focused events, including three workshops and three panels at IEEE Visualization (VIS) Conferences. Together, these events have produced a set of convincing arguments.

Parallel code portability in the petascale era requires modifying existing codes to support new architectures with large core counts and SIMD vector units. OpenMP is a well established and increasingly supported vehicle for portable parallelization. As architectures mature and compiler OpenMP implementations evolve, best practices for code modernization change as well. In this paper, we examine the impact of newer OpenMP features (in particular OMP SIMD) on the Intel Xeon Phi Knights Landing (KNL) architecture, applied in optimizing loops in the single moment 6-class microphysics module (WSM6) in the US Navy's NEPTUNE code. We find that with functioning OMP SIMD constructs, low thread invocation overhead on KNL and reduced penalty for unaligned access compared to previous architectures, one can leverage OpenMP 4 to achieve reasonable scalability with relatively minor reorganization of a production physics code.

The University of Utah's Carbon Capture Multidisciplinary Simulation Center (CCMSC) is using the Uintah Computational Framework to predict performance of a 1000 MWe ultra-supercritical clean coal boiler. The center aims to utilize the Intel Xeon Phi-based DOE systems, Theta and Aurora, through the Aurora Early Science Program by using the Kokkos C++ library to enable node-level performance portability. This paper describes infrastructure advancements and portability improvements made possible by our integration of Kokkos within Uintah. Scalability results are presented that compare serial and data parallel task execution models for a challenging radiative heat transfer calculation, central to the center's predictive boiler simulations. These results demonstrate both good strong-scaling characteristics to 256 Knights Landing (KNL) processors on the NSF Stampede system, and show the KNL-based calculation to compete with prior GPU-based results for the same calculation.

The lithiation and delithiation of a silicon battery anode is modeled using the material point method (MPM). The main challenges in modeling this process using the MPM is to simulate stress dependent diffusion coupled with concentration dependent stress within a material that undergoes large deformations. MPM is chosen as the numerical method of choice because of its ability to handle large deformations. A method for modeling diffusion within MPM is described. A stress dependent model for diffusivity and three different constitutive models that fully couple the equations for stress with the equations for diffusion are considered. Verifications tests for the accuracy of the numerical implementations of the models and validation tests with experimental results show the accuracy of the approach. The application of the fully coupled stress diffusion model implemented in MPM is applied to modeling the lithiation and delithiation of silicon nanopillars.

The increasing gap between available compute power and I/O capabilities is resulting in simulation pipelines running on leadership computing facilities being reformulated. In particular, in-situ processing is complementing conventional post-process analysis; however, it can be performed by using the same compute resources as the simulation or using secondary dedicated resources.

In this paper, we focus on three different in-situ analysis strategies, which use the same compute resources as the ongoing simulation but different data movement strategies. We evaluate the costs incurred by these strategies in terms of run time, scalability and power/energy consumption. Furthermore, we extrapolate power behavior to peta-scale and investigate different design choices through projections. Experimental evaluation at full machine scale on Titan supports that using fewer cores per node for in-situ analysis is the optimum choice in terms of scalability. Hence, further research effort should be devoted towards developing in-situ analysis techniques following this strategy in future high-end systems.

The current diversity in nodal parallel computer architectures is seen in machines based upon multicore CPUs, GPUs and the Intel Xeon Phi's. A class of approaches for enabling scalability of complex applications on such architectures is based upon Asynchronous Many Task software architectures such as that in the Uintah framework used for the parallel solution of solid and fluid mechanics problems. Uintah has both an applications layer with its own programming model and a separate runtime system. While Uintah scales well today, it is necessary to address nodal performance portability in order for it to continue to do. Incrementally modifying Uintah to use the Kokkos performance portability library through prototyping experiments results in improved kernel performance by more than a factor of two.

We present a system for interactive in situ visualization of large particle simulations, suitable for general CPU-based HPC architectures. As simulations grow in scale, in situ methods are needed to alleviate IO bottlenecks and visualize data at full spatio-temporal resolution. We use a lightweight loosely-coupled layer serving distributed data from the simulation to a data-parallel renderer running in separate processes. Leveraging the OSPRay ray tracing framework for visualization and balanced P-k-d trees, we can render simulation data in real-time, as they arrive, with negligible memory overhead. This flexible solution allows users to perform exploratory in situ visualization on the same computational resources as the simulation code, on dedicated visualization clusters or remote workstations, via a standalone rendering client that can be connected or disconnected as needed. We evaluate this system on simulations with up to 227M particles in the LAMMPS and Uintah computational frameworks, and show that our approach provides many of the advantages of tightly-coupled systems, with the flexibility to render on a wide variety of remote and co-processing resources.

Traditional scientific visualization software approaches do not fare well in massively threaded environments. To address the needs of the high-performance computing community, the VTK-m framework fills the gaps in functionality by bringing together the most recent research.

The material point method (MPM) has been very successful in providing solutions to many challenging problems involving large deformations. Nevertheless there are some important issues that remain to be resolved with regard to its analysis. One key challenge applies to both MPM and particle-in-cell (PIC) methods and arises from the difference between the number of particles and the number of the nodal grid points to which the particles are mapped. This difference between the number of particles and the number of grid points gives rise to a non-trivial null space of the linear operator that maps particle values onto nodal grid point values. In other words, there are non-zero particle values that when mapped to the grid point nodes result in a zero value there. Moreover, when the nodal values at the grid points are mapped back to particles, part of those particle values may be in that same null space. Given positive mapping weights from particles to nodes such null space values are oscillatory in nature. While this problem has been observed almost since the beginning of PIC methods there are still elements of it that are problematical today as well as methods that transcend it. The null space may be viewed as being connected to the ringing instability identified by Brackbill for PIC methods. It will be shown that it is possible to remove these null space values from the solution using a null space filter. This filter improves the accuracy of the MPM methods using an approach that is based upon a local singular value decomposition (SVD) calculation. This local SVD approach is compared against the global SVD approach previously considered by the authors and to a recent MPM method by Zhang and colleagues.

The current diversity in nodal parallel computer architectures is seen in machines based upon multicore CPUs, GPUs and the Intel Xeon Phi's. A class of approaches for enabling scalability of complex applications on such architectures is based upon Asynchronous Many Task software architectures such as that in the Uintah framework used for the parallel solution of solid and fluid mechanics problems. Uintah has both an applications layer with its own programming model and a separate runtime system. While Uintah scales well today, it is necessary to address nodal performance portability in order for it to continue to do. Incrementally modifying Uintah to use the Kokkos performance portability library through prototyping experiments results in improved kernel performance by more than a factor of two.

The detonation of hundreds of explosive devices from either a transportation or storage accident is an extremely dangerous event. This paper focuses on identifying ways of packing/storing arrays of explosive cylinders that will reduce the probability of a Deflagration to Detonation Transition (DDT). The Uintah Computational Framework was utilized to predict the conditions necessary for a large scale DDT to occur. The results showed that the arrangement of the explosive cylinders and the number of devices packed in a "box" greatly effects the probability of a detonation.

The 2015 SIAM Conference on Computational Science and Engineering (CSE) was held March 14-18, 2015, in Salt Lake City, Utah. The SIAM Journal on Scientific Computing (SISC) created this special section in association with the CSE15 conference. The special section focuses on two topics that are of significant current interest to CSE researchers: CSE software and big data in CSE.

Read More: http://epubs.siam.org/doi/abs/10.1137/16N974188

Over the past two decades the field of computational science and engineering (CSE) has penetrated both basic and applied research in academia, industry, and laboratories to advance discovery, optimize systems, support decision-makers, and educate the scientific and engineering workforce. Informed by centuries of theory and experiment, CSE performs computational experiments to answer questions that neither theory nor experiment alone is equipped to answer. CSE provides scientists and engineers of all persuasions with algorithmic inventions and software systems that transcend disciplines and scales. Carried on a wave of digital technology, CSE brings the power of parallelism to bear on troves of data. Mathematics-based advanced computing has become a prevalent means of discovery and innovation in essentially all areas of science, engineering, technology, and society; and the CSE community is at the core of this transformation. However, a combination of disruptive developments---including the architectural complexity of extreme-scale computing, the data revolution that engulfs the planet, and the specialization required to follow the applications to new frontiers---is redefining the scope and reach of the CSE endeavor. This report describes the rapid expansion of CSE and the challenges to sustaining its bold advances. The report also presents strategies and directions for CSE research and education for the next decade.

Sort-last parallel rendering can be improved by considering the rendering of multiple images at a time. Most parallel rendering algorithms consider the generation of only a single image. This makes sense when performing interactive rendering where the parameters of each rendering are not known until the previous rendering completes. However, in situ visualization often generates multiple images that do not need to be created sequentially. In this paper we present a simple and effective approach to improving parallel image generation throughput by amortizing the load and overhead among multiple image renders. Additionally, we validate our approach by conducting a performance study exploring the achievable speed-ups in a variety of image-based in situ use cases and rendering workloads. On average, our approach shows a 1.5 to 3.7 fold improvement in performance, and in some cases, shows a 10 fold improvement.