Search NASA⌕ Search

SEARCH · Search NASA

Results for “HPC performance modeling”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 667 records · Page 37

Multi-physics melt pool modeling and process optimization for laser direct energy deposition of Nb-based refractory C103: Defect formation, geometric precision, and process mapping

Recent developments in additive manufacturing (AM) technology have reignited interest in the fabrication of the Nb-based refractory C103 alloy offering solutions to the challenges posed by traditional manufacturing methods. However, the limited numerical and experimental studies on laser direct energy deposition (DED) of C103 have hindered the understanding of the relationships between process parameters and build quality. This has made it challenging to consistently produce parts with the desired quality and microstructure suitable for critical applications. In this study, we focus on optimizing the laser DED process for C103 by employing a hybrid approach that combines experimental techniques and computational fluid dynamics (CFD). This approach facilitates the development of process maps for defect detection and geometric precision. To achieve this, multi-layer C103 samples were fabricated using laser DED under various process parameters, enabling the creation of a process map for defect detection. Additionally, a multi-physics, multiphase simulation framework was developed within a high-performance computing (HPC) environment to establish process maps for geometric precision. Using these process maps, printability windows were identified for achieving both the desired geometric accuracy and defect-free prints. It was observed that prints with a power-to-velocity (P/V) ratio close to unity resulted in defect-free outcomes. This study provides a foundation for reducing design lead time and rejected parts, ultimately optimizing the laser DED process for C103.

Defect formation and geometric precision↗

Cyber-Infrastructure for Data-Intensive Geospatial Computing

With the recent advent of heterogeneous High-performance Computing (HPC) to handle EO “Big Data” workloads, there is a need for a unified Cyber-infrastructure (CI) platform that can bridge the best of many HPC worlds. In this chapter, we discuss such a CI platform being developed at Geographic Information Science and Technology (GIST) group using novel and innovative techniques, and emerging technologies that are scalable to large-scale supercomputers. The CI platform utilizes a wide variety of computing such as GPGPU, distributed, real-time and cluster computing, which are being brought together architecturally to enable data-driven analysis, scientific understanding of earth system models, and research collaboration. This development addresses the need for close integration of EO and other geospatial information in the face of growing volumes of the data, and facilitates spatio-temporal analysis of disparate and dynamic data streams. Horizontal scalability and linear throughput are supported in the heart of the platform itself. It is being used to support very broad application areas, ranging from high-resolution settlement mapping, national bioenergy infrastructure to urban information and mobility systems. The platform provides spatio-temporal decision support capabilities in planning, policy and operational missions for US federal agencies. Also, the platform is designed to be functionally and technologically sustainable for continued support of the US energy and environment mission for the coming decades.

Karthik, Rajasekar↗

Collaborating CPU and GPU for large-scale high-order CFD simulations with complex grids on the TianHe-1A supercomputer

Programming and optimizing complex, real-world CFD codes on current many-core accelerated HPC systems is very challenging, especially when collaborating CPUs and accelerators to fully tap the potential of heterogeneous systems. In this paper, with a tri-level hybrid and heterogeneous programming model using MPI + OpenMP + CUDA, we port and optimize our high-order multi-block structured CFD software HOSTA on the GPU-accelerated TianHe-1A supercomputer. HOSTA adopts two self-developed high-order compact definite difference schemes WCNS and HDCS that can simulate flows with complex geometries. We present a dual-level parallelization scheme for efficient multi-block computation on GPUs and perform particular kernel optimizations for high-order CFD schemes. The GPU-only approach achieves a speedup of about 1.3 when comparing one Tesla M2050 GPU with two Xeon X5670 CPUs. To achieve a greater speedup, we collaborate CPU and GPU for HOSTA instead of using a naive GPU-only approach. We present a novel scheme to balance the loads between the store-poor GPU and the store-rich CPU. Taking CPU and GPU load balance into account, we improve the maximum simulation problem size per TianHe-1A node for HOSTA by 2.3×, meanwhile the collaborative approach can improve the performance by around 45% compared to the GPU-only approach. Further, to scale HOSTA on TianHe-1A, we propose a gather/scatter optimization to minimize PCI-e data transfer times for ghost and singularity data of 3D grid blocks, and overlap the collaborative computation and communication as far as possible using some advanced CUDA and MPI features. Scalability tests show that HOSTA can achieve a parallel efficiency of above 60% on 1024 TianHe-1A nodes. With our method, we have successfully simulated an EET high-lift airfoil configuration containing 800M cells and China's large civil airplane configuration containing 150M cells. To our best knowledge, those are the largest-scale CPU–GPU collaborative simulations that solve realistic CFD problems with both complex configurations and high-order schemes.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Improving Block-level Efficiency with scsi-mq

Current generation solid-state storage devices are exposing a new bottlenecks in the SCSI and block layers of the Linux kernel, where IO throughput is limited by lock contention, inefficient interrupt handling, and poor memory locality. To address these limitations, the Linux kernel block layer underwent a major rewrite with the blk-mq project to move from a single request queue to a multi-queue model. The Linux SCSI subsystem rework to make use of this new model, known as scsi-mq, has been merged into the Linux kernel and work is underway for dm-multipath support in the upcoming Linux 4.0 kernel. These pieces were necessary to make use of the multi-queue block layer in a Lustre parallel filesystem with high availability requirements. We undertook adding support of the 3.18 kernel to Lustre with scsi-mq and dm-multipath patches to evaluate the potential of these efficiency improvements. In this paper we evaluate the block-level performance of scsi-mq with backing storage hardware representative of a HPC-targerted Lustre filesystem. Our findings show that SCSI write request latency is reduced by as much as 13.6%. Additionally, when profiling the CPU usage of our prototype Lustre filesystem, we found that CPU idle time increased by a factor of 7 with Linux 3.18 and blk-mq as compared to a standard 2.6.32 Linux kernel. Our findings demonstrate increased efficiency of the multi-queue block layer even with disk-based caching storage arrays used in existing parallel filesystems.

Cluster Computing↗

Enhancing Monte Carlo proxy applications on GPUs

In Monte Carlo neutron transport simulations, a computational routine commonly known as the ”cross-section lookup” has been identified as being the most computationally expensive part of these applications. A tool which is commonly used as a proxy application for these routines, named ”XSBench”, was created to simulate popular algorithms used in these routines on CPUs. Currently, however, as GPU-based HPC resources have become more widely available, there has been significant interest and efforts invested in moving these traditionally CPU-based simulations to GPUs. Unfortunately, the algorithms commonly used in the cross-section lookup routine were originally devised and developed for CPU-based platforms, and have seen limited study on GPUs to date. Additionally, platforms such as XSBench implement approximations which may have a negligible effect on CPUs, but may be quite impactful to performance on GPUs given the more resource-limited nature of the latter. As a result, we have created VEXS, a new tool for modeling the cross-section lookup routine which removes or at least reduces the approximations made by XSBench in order to provide a more realistic prediction of algorithm performance on GPUs. In this paper, we detail our efforts to remove and reduce these approximations, show the resulting improvement in performance prediction in comparison to a reference production code, Shift, and provide some basic profiling analysis of the resulting application.

Shriver, Forrest↗

COMPOSE-HPC: A Transformational Approach to Exascale

The goal of the COMPOSE-HPC project is to 'democratize' tools for automatic transformation of program source code so that it becomes tractable for the developers of scientific applications to create and use their own transformations reliably and safely. This paper describes our approach to this challenge, the creation of the KNOT tool chain, which includes tools for the creation of annotation languages to control the transformations (PAUL), to perform the transformations (ROTE), and optimization and code generation (BRAID), which can be used individually and in combination. We also provide examples of current and future uses of the KNOT tools, which include transforming code to use different programming models and environments, providing tests that can be used to detect errors in software or its execution, as well as composition of software written in different programming languages, or with different threading patterns.

99 GENERAL AND MISCELLANEOUS↗

Dilute Combustion Control Using Spiking Neural Networks

Dilute combustion with exhaust gas recirculation (EGR) in spark-ignition engines presents a cost-effective method for achieving higher levels of engine efficiency. At high levels of EGR, however, cycle-to-cycle variability (CCV) of the combustion process is exacerbated by sporadic occurrences of misfires and partial burns. Previous studies have shown that temporal deterministic patterns emerge at such conditions and certain combustion cycles have a significant influence over future events. Due to the complexity of the combustion process and the nature of CCV, harnessing all the deterministic information for control purposes has remained challenging even with physics based 0-D, 1-D, and high-fidelity computational fluid dynamics (CFD) models. In this study, we present a data-driven approach to optimize the combustion process by controlling CCV adjusting the cycle-to-cycle fuel injection quantity. Readily available data from in-cylinder pressure was used to train a spiking neural network (SNN) which learns the optimal way to manage fuel injection in order to reduce CCV while maintaining acceptable levels of fuel consumption. SNNs are particularly well suited for powertrain control applications due to their ability to be deployed on FPGA-based neuromorphic hardware which are small, inexpensive, and have a low power demand. The high-performance computing (HPC) resources of Oak Ridge National Laboratory were used to run an evolutionary-based training approach for choosing the best SNN configuration that minimizes the size of the network while achieving the desired goal. The neuromorphic hardware with the optimized SNN deployed was connected to the rapid prototyping engine control system for real-time control implementation and tested on a single cylinder version of a GM LNF 4-cylinder engine. The results show a significant reduction of CCV with a small percentage of additional fuel used to stabilize the charge.

33 ADVANCED PROPULSION SYSTEMS↗

Is Knowledge about Running Applications Helping Improve Runtime Prediction of HPC Jobs?

High-performance computing systems rely upon scheduling algorithms to achieve high utilization. These schedulers rely upon user estimates of job resource requirements, such as runtime, to determine optimal scheduling of incoming jobs. These user estimates, however, are prone to error. To mitigate this error, significant research has been directed at providing better estimates of job runtime, usually employing machine learning techniques. These techniques are dependent upon the input features selected. Among the possible features is the primary application used by the job. In a survey of more than 20 papers directed at improving runtime prediction, only four included primary application as an input feature. We focus this investigation specifically on the value of adding primary application as an input feature, and find that it does improve model performance, especially for jobs with longer runtimes, though this improvement varies based on the application used. We recommend further research to determine the cause of this variability as well as an optimal strategy for employing a mixture of models both including and not including primary application as a feature.

MATHEMATICS AND COMPUTING↗

Is Knowledge About Running Applications Helping Improve Runtime Prediction of HPC Jobs?

High-performance computing systems rely upon scheduling algorithms to achieve high utilization. These schedulers rely upon user estimates of job resource requirements, such as runtime, to determine optimal scheduling of incoming jobs. These user estimates, however, are prone to error. To mitigate this error, significant research has been directed at providing better estimates of job runtime, usually employing machine learning techniques. These techniques are dependent upon the input features selected. Among the possible features is the primary application used by the job. In a survey of more than 20 papers directed at improving runtime prediction, only four included primary application as an input feature. We focus this investigation specifically on the value of adding primary application as an input feature, and find that it does improve model performance, especially for jobs with longer runtimes, though this improvement varies based on the application used. We recommend further research to determine the cause of this variability as well as an optimal strategy for employing a mixture of models both including and not including primary application as a feature.

feature selection↗

A Framework for Integrating Quantum Simulation and High Performance Computing

Scientific applications are starting to explore the viability of quantum computing. This exploration typically begins with quantum simulations that can run on existing classical platforms, albeit without the performance advantages of real quantum resources. In the context of high-performance computing (HPC), the incorporation of simulation software can often take advantage of the powerful resources to help scale-up the simulation size. The configuration, installation and operation of these quantum simulation packages on HPC resources can often be rather daunting and increases friction for experimentation by scientific application developers. We describe a framework to help streamline access to quantum simulation software running on HPC resources. This includes an interface for circuit-based quantum computing tasks, as well as the necessary resource management infrastructure to make effective use of the underlying HPC resources. The primary contributions of this work include a classification of different usage models for quantum simulation in an HPC context, a review of the software architecture for our approach and a detailed description of the prototype implementation to experiment with these ideas using two different simulators (TNQVM & NWQ-Sim). We include initial experimental results running on the Frontier supercomputer at the Oak Ridge Leadership Computing Facility (OLCF) using a synthetic workload generated via the SupermarQ quantum benchmarking framework.

Shehata, Amir [ORNL] (ORCID:0000000224531426)↗

Multi-meson systems in lattice QCD / Many-body QCD

Nuclear physics entails the study of the properties and interactions of hadrons, such as the proton and neutron, and atomic nuclei and it is central to our understanding of our world at the smallest scales. The underlying basis for nuclear physics is provided by the Standard Model of particle physics which describes how matter interacts through the strong, electromagnetic and weak (electroweak) forces. This theory was developed in the 1970s and provides an extremely successful description of our world at the most fundamental level to which it has been probed. The Standard Model has been, and continues to be, subject to stringent tests at particle accelerators around the world, so far passing without blemish. However, at the relatively low energies that are relevant for nuclear physics, calculations involving the strong interaction, governed by the equations of Quantum Chromodynamics (QCD), are enormously challenging, and to date, the only systematic way to perform them is numerically, using a framework known as lattice QCD (LQCD). In this approach, one discretizes space-time and numerically solves the equations of QCD on a space-time lattice; for realistic calculations, this requires highly optimized algorithms and cutting-edge high performance computing (HPC) resources. Progress over the project period is discussed in detail in the following subsections

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Advanced simulation capability for environmental management, integrated tool-sets and simulator to enhance public communication - 15186

The U.S. Department of Energy (US DOE) Office of Environmental Management (EM), Office of Soil and Groundwater Remediation, is supporting multi-laboratory development of the Advanced Simulation Capability for Environmental Management (ASCEM). ASCEM is an open source and modular computing framework that incorporates new advances and tools for predicting contaminant fate and transport in natural and engineered systems. ASCEM facilitates transparent, efficient and integrated approaches to modeling and site characterization, and provides robust and standardized assessments of performance and risk for EM cleanup and closure activities. Site subsurface simulation modeling and performance assessments are complex and involve sensitivity evaluations and uncertainty quantification to better understand the risks at a site. The ASCEM tool-set was designed in close collaboration with the sites and other potential end-users (both the public and regulators) to allow better communication and understanding of these complex modeling efforts. ASCEM's integrated tool-sets provide a user interface and tools for managing work-flows, including conceptual model development, management of data and metadata for model input, sensitivity analysis, model calibration and uncertainty analysis, model execution on diverse computational platforms, and processing of model output, including visualization. The hierarchical and modular design of ASCEM HPC simulator Amanzi easily lends itself to translating this conceptual model of a site into a numerical model that can easily be spatially and temporally discretized to run on the HPC simulator. Together this integrated work-flow allows for a seamless and integrated analysis tracking all provenance information and modeling uncertainties. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

CHASM Challenge Problem: Lagrangian Hydrodynamics

Computer simulations of many science and engineering problems require modeling the equations of hydrodynamics which describe the motion of materials relative to each other induced by various forces. Many important DoD simulation problems involve complex multi-material systems that undergo large deformations. Examples include the analysis of armor defense, penetration mechanics, blast effects, structural integrity, and conventional munitions such as shaped charges and explosively formed projectiles. Indeed, the original motivation for developing codes that solve the equations of hydrodynamics, herein referred to as “hydrocodes”, was to solve problems with defense applications. The FY2010 Requirements Analysis Report issued by the DoD High Performance Computing Modernization Program (HPCMP) Office shows that a major portion of DoD HPC activities involves hydrocodes [HPCMP2010]. The report surveyed 496 projects across the Services and various Agencies, representing 4,050 HPCMP users at more than 125 locations, including government, contractors, and academia, and grouped each project into one of ten categories.

42 ENGINEERING↗

Evaluating the Efficacy of the Cloud for Cluster Computation

Computing requirements vary by industry, and it follows that NASA and other research organizations have computing demands that fall outside the mainstream. While cloud computing made rapid inroads for tasks such as powering web applications, performance issues on highly distributed tasks hindered early adoption for scientific computation. One venture to address this problem is Nebula, NASA's homegrown cloud project tasked with delivering science-quality cloud computing resources. However, another industry development is Amazon's high-performance computing (HPC) instances on Elastic Cloud Compute (EC2) that promises improved performance for cluster computation. This paper presents results from a series of benchmarks run on Amazon EC2 and discusses the efficacy of current commercial cloud technology for running scientific applications across a cluster. In particular, a 240-core cluster of cloud instances achieved 2 TFLOPS on High-Performance Linpack (HPL) at 70% of theoretical computational performance. The cluster's local network also demonstrated sub-100 ?s inter-process latency with sustained inter-node throughput in excess of 8 Gbps. Beyond HPL, a real-world Hadoop image processing task from NASA's Lunar Mapping and Modeling Project (LMMP) was run on a 29 instance cluster to process lunar and Martian surface images with sizes on the order of tens of gigapixels. These results demonstrate that while not a rival of dedicated supercomputing clusters, commercial cloud technology is now a feasible option for moderately demanding scientific workloads.

Cloud Computing↗

Decaf: Decoupled Dataflows for In Situ High-Performance Workflows

Decaf is a dataflow system for the parallel communication of coupled tasks in an HPC workflow. The dataflow can perform arbitrary data transformations ranging from simply forwarding data to complex data redistribution. Decaf does this by allowing the user to allocate resources and execute custom code in the dataflow. All communication through the dataflow is efficient parallel message passing over MPI. The runtime for calling tasks is entirely message-driven; Decaf executes a task when all messages for the task have been received. Such a messagedriven runtime allows cyclic task dependencies in the workflow graph, for example, to enact computational steering based on the result of downstream tasks. Decaf includes a simple Python API for describing the workflow graph. This allows Decaf to stand alone as a complete workflow system, but Decaf can also be used as the dataflow layer by one or more other workflow systems to form a heterogeneous task-based computing environment. In one experiment, we couple a molecular dynamics code with a visualization tool using the FlowVR and Damaris workflow systems and Decaf for the dataflow. In another experiment, we test the coupling of a cosmology code with Voronoi tessellation and density estimation codes using MPI for the simulation, the DIY programming model for the two analysis codes, and Decaf for the dataflow. Such workflows consisting of heterogeneous software infrastructures exist because components are developed separately with different programming models and runtimes, and this is the first time that such heterogeneous coupling of diverse components was demonstrated in situ on HPC systems.

97 MATHEMATICS AND COMPUTING↗

Automatic Generation of High-Performance Convolution Kernels on ARM CPUs for Deep Learning

In this work, we present FastConv, a template-based code auto-generation open source library that can automatically generate high-performance deep learning convolution kernels of arbitrary matrices/tensors shapes. FastConv is based on the Winograd algorithm, which is reportedly the highest performing algorithm for the time-consuming convolution layers of convolutional neural networks. ARM CPUs cover a wide range designs and specifications, from embedded devices to HPC-grade CPUs. The leads to the dilemma of how to consistently optimize Winograd-based convolution solvers for convolution layers of different shapes. FastConv addresses this problem by using templates to auto-generate multiple shapes of tuned kernels variants suitable for skinny tall matrices. As a performance portable library, FastConv transparently searches for the best combination of kernel shapes, cache tiles, scheduling of loop orders, packing strategies, access patterns, and online/offline computations. Auto-tuning is used to search the parameter configuration space for the best performance for a given target architecture and problem size. The experiments with layer-wise evaluation on the VGG--16 model confirms a 1.25x performance gains is got by tuning the Winograd library. Integrated comparison results shows 1.02x to 1.40x, 1.14x to 2.17x, and 1.22x and 2.48x speedup is achieved over NNPACK, Arm NN, and FeatherCNN on the Kunpeng 920 beside few cases. Furthermore, problem size performance portability experiments with various convolution shapes shows that FastConv achieves 1.2x to 1.7x speedup and 2x to 22x speedup over NNPACK and ARM NN inference engine using Winograd on Kunpeng 920 . CPU performance portability evaluation on the VGG--16 show an average speedup over NNPACK of 1.42x, 1.21x, 1.26x, 1.37x, 2.26x, and 11.02x is observed on Kunpeng 920, Snapdragon 835, 855, 888, Apple M1, and AWS Graviton2, respectively.

97 MATHEMATICS AND COMPUTING↗

A Framework for the Analysis of Compiler Optimizations

Compilers transform program source code to machine executable code. During this transformation, they perform a number of compiler optimizations to improve the performance of the generated executable code. Importantly, applying those optimizations depends on the source code structure, such as the parallel programming model used to parallelize an algorithm. Often, implementations of the same algorithm with different programming models have vastly different performance because the compiler optimized them differently. We create FAROS, a framework to structure and automate the analysis of compiler optimizations on programs. FAROS automates the building process, execution profiling, and analysis of compiler optimization of programs, through a configuration interface. It outputs compiler optimization reports to show which optimizations applied to which line of source code, leveraging compilation remarks output by the compiler. Also, FAROS supports benchmarking performance of different program versions by collecting execution time results. In this first release of FAROS, we provide a configuration file to analyze compiler optimization differences for sequential vs OpenMP compilation, including 38 programs consisting of HPC proxy/mini/large applications, and NAS and Rodina kernels for analysis.

Georgakoudis, Giorgis↗

FitCache: A Transparent Drop-In Framework for Multi-Tier Caching to Accelerate Distributed Deep Learning Workloads

Training in Deep learning (DL) remains highly compute- and data-intensive, with I/O becoming a critical bottleneck as models and datasets scale. Recent studies report that data loading can dominate training time, especially on large-scale HPC systems with shared parallel file systems (PFS). Existing caching approaches either rely on single-tier designs or require intrusive modifications to training pipelines, limiting their portability and effectiveness. In this work, we present FitCache, a transparent drop-in framework for multi-tier caching to accelerate distributed DL training by coordinating fast local memory (e.g., DRAM, Persistent Memory (PMem)) and NVMe as hierarchical caches atop PFS. Our design adapts to hardware diversity, i.e., if NVMe is missing, memory transparently acts as a caching tier, ensuring stable performance. FitCache transparently intercepts I/O requests and issues concurrent fetches across all tiers, returning data from the fastest responder without centralized metadata or static redirection paths. FitCache adapts to dynamic workloads and heterogeneous clusters while maintaining POSIX compatibility. Experiments on Frontier (2048 GPUs) and smaller research clusters show that FitCache reduces training time by up to 40% and per-batch I/O latency by up to 71.6% compared to Lustre Orion PFS, offering a drop-in solution for scalable DL training.

Hu, Guangxing [ORNL] (ORCID:0009000283203614)↗