SEARCH · Engineering Papers

Results for “Memory and runtime performance”

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.

SourceResearch connectionFrom yearThrough year

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

At least 55 records · Page 3

All sources NASA NTRS DOE OSTI

DOE OSTI2020

Performance Analysis of Traditional and Data-Parallel Primitive Implementations of Visualization and Analysis Kernels

Measurements of absolute runtime are useful as a summary of performance when studying parallel visualization and analysis methods on computational platforms of increasing concurrency and complexity. We can obtain even more insights by measuring and examining more detailed measures from hardware performance counters, such as the number of instructions executed by an algorithm implemented in a particular way, the amount of data moved to/from memory, memory hierarchy utilization levels via cache hit/miss ratios, and so forth. This work focuses on performance analysis on modern multi-core platforms of three different visualization and analysis kernels that are implemented in different ways: one is "traditional", using combinations of C++ and VTK, and the other uses a data-parallel approach using VTK-m. Our performance study consists of measurement and reporting of several different hardware performance counters on two different multi-core CPU platforms. The results reveal interesting performance differences between these two different approaches for implementing these kernels, results that would not be apparent using runtime as the only metric.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Devastator Parallel Discrete Event Simulation Runtime (Devastator) v1.0

The Devastator runtime is a modern C++ implementation of optimistic parallel discrete event simulation methods. Devastator allows simulation application code to productively specify their component and event functionality with C++14 constructs. It utilizes GASNet-EX for distributed memory communication and includes parallel performance optimizations such as light-weight thread message queues and asynchronous GVT. Furthermore, it supports efficient event broadcasts and pause-rewind-resume functionality to support periodic load balancing and outer loop optimization algorithms.

Chan, Cy↗

DOE OSTI2024

Adapting CLUTCH methodology to multigroup TSUNAMI-3D for eigenvalue sensitivity calculations

The sensitivity of the eigenvalue to uncertainties in nuclear data and its evaluation are important for nuclear criticality safety. TSUNAMI-3D sequences within the SCALE code system offer several options to the user community for calculating eigenvalue sensitivity coefficients with multigroup (MG) and continuous energy (CE) 3D transport capabilities. TSUNAMI-3D sequences implement the adjoint-based perturbation theory with MG KENO code, the Contributon Linked eigenvalue sensitivity/Uncertainty estimation via Track length importance CHaracterization (CLUTCH) method with CE KENO code, and the Iterated Fission Probability (IFP) method with CE KENO and Shift codes. Each method has benefits and limitations depending on the problem that is run. The work presented here aims to adapt the CLUTCH method, which enables the Contributon method's mesh-free, memory-efficient approach for calculating adjoint-weighted tallies for sensitivity calculations, to the MG TSUNAMI-3D sequence. This application would eliminate the explicit adjoint KENO calculation, as well as the memory-consuming mesh flux moment tallies required by the conventional MG TSUNAMI-3D. Smaller memory footprints in the CLUTCH methodology and relatively shorter runtimes in MG KENO transport can make MG TSUNAMI-3D a viable method for some complex problems. Moreover, this adaptation allows MG sensitivity calculations with Shift, ORNL's next-generation high-performance Monte Carlo transport code, which currently does not offer any sensitivity capabilities with MG particle transport simulations. Initial implementation of the new MG TSUNAMI-3D sequence and its preliminary results with a selected critical benchmark experiment in the Verified, Archived Library of Inputs and Data (VALID) are presented in this study.

KENO↗

DOE OSTI2024

ECP libraries and tools: An overview

The Exascale Computing Project (ECP) Software Technology and Co-Design teams addressed the growing complexities in high-performance computing (HPC) by developing scalable software libraries and tools that leverage exascale system capabilities. As we enter the exascale era, the need for reusable, optimized software solutions that can handle the unique challenges posed by these systems becomes increasingly important. The primary challenges the ECP teams faced were to create software libraries and tools that are performant on exascale architectures and portable and usable across diverse hardware platforms. Efforts addressed issues related to concurrent execution, memory management, and the integration of heterogeneous computing resources, such as GPUs from multiple vendors. The ECP’s strategy involved a structured development process encompassing the creation, optimization, and deployment of software in collaboration with industry, academia, and national laboratories. The project was organized into several technical areas: co-design of domain-specific suites with target applications, programming models and runtimes, development tools, mathematical libraries, data and visualization tools, and software ecosystem and delivery mechanisms. ECP has successfully developed a large portfolio of software libraries and tools that demonstrate significant improvements in performance and scalability on exascale systems. These products have been integrated into the Department of Energy’s computing facilities, supporting various scientific applications and ensuring robust performance across different hardware setups. ECP advancements in software development for exascale computing highlight the importance of a collaborative and adaptive approach to handling next-generation HPC systems complexities. The lessons learned emphasize the need for continuous engagement with end-users and vendors, and the importance of maintaining a balance between innovation and practical implementation. Future efforts will focus on ensuring scalability, keeping pace with rapid hardware advancements, and further enhancing the interoperability and usability of the software ecosystem. In conclusion, subsequent articles in this special issue provide in-depth discussions and case studies into specific library and tool efforts.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Direction-optimizing Label Propagation and its Application for Community Detection

Label Propagation is a machine learning algorithm typically used for classification. It has also been found to be an effective method for detecting communities in networks. It has two attractive features as a community detection method: it has nearly linear runtime and it requires no \textit{a priori} community information. We propose a new Direction Optimizing Label Propagation Algorithm (DOLPA) that relies on the use of {\em frontiers} and alternates between label {\em push} and label {\em pull} operations to enhance the performance of LPA. Specifically, DOLPA has parameters for tuning the processing order of vertices in a graph. This reduces the number of edges visited and improves the quality of solution. We apply DOLPA to community detection and present the design and implementation of the algorithm as well as its shared-memory parallelization using OpenMP. Empirically, we evaluate our algorithm using synthetic graphs as well as real-world networks. Compared with the state-of-art \textit{Parallel Label Propagation} algorithm, we achieve at least two times the F-Score while reducing the runtime by 50\% for synthetic graphs with overlapping communities. We also compare DOLPA against the state-of-art parallel implementation of the Louvain method using the same graphs and show that DOLPA achieves about three times the F-Score at 10\% the runtime. On real-world graphs, we get a speedup of up to $10\times$ using 64 threads.

Liu, X↗

DOE OSTI2021

$\mathrm{PPT}$-Multicore: performance prediction of Open$\mathrm{MP}$ applications using reuse profiles and analytical modeling

In this report we present PPT-Multicore, an analytical model embedded in the Performance Prediction Toolkit (PPT) to predict parallel applications’ performance running on a multicore processor. PPT-Multicore builds upon our previous work towards a multicore cache model. We extract LLVM basic block labeled memory trace using an architecture-independent LLVM-based instrumentation tool only once in an application’s lifetime. The model uses the memory trace and other parameters from an instrumented sequentially executed binary. We use probabilistic and computationally efficient reuse profiles to predict the cache hit rates and runtimes of OpenMP programs’ parallel sections. We model Intel’s Broadwell, Haswell, and AMD’s Zen2 architectures and validate our framework using different applications from PolyBench and PARSEC benchmark suites. The results show that PPT-Multicore can predict cache hit rates with an overall average error rate of 1.23% while predicting the runtime with an error rate of 9.08%.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

DIMPLES: Distributed Influence Maximization for Pandemic pLanning on Exascale Systems

We study exascale parallel algorithms for the selection of intervention or monitoring strategies in massive realistic socio-technical networks through scalable Influence Maximization (InfMax) algorithms. We employ novel techniques to enable efficient scaling on up to 8k nodes of OLCF Frontier, with 65k AMD GPUs and 458k AMD CPU cores. Current state-of-the-art InfMax tools are limited to networks with only a few million actors (vertices) and a few hundred million interactions (edges). By overcoming these limitations, we show that our approach is capable of processing a realistic social contact network of the United States with 285 million nodes and about 8 billion edges. This two orders-of-magnitude improvement over the previous state-of-the-art is obtained by leveraging algorithmic advancements for the InfMax problem and designing several problem-specific approaches to overlap communication with computation, improve GPU efficiency, and lower the application’s memory requirements. We evaluate strong scaling for computing 10k most influential seeds using up to 8k nodes of an exascale system, and weak scaling from 128 to 8k system nodes for seed sets ranging from 625 to 40k seeds. We achieve the fastest-known runtime of 25 minutes while performing 48 million diffusion simulations totaling 2.31 petabytes to identify 40k influential seeds using 8k nodes, and take 5.75 minutes to identify 10k seeds while using 4k nodes.

Minutoli, Marco [Pacific Northwest National Labora↗

DOE OSTI2025

SpecSims: A Scalable Speculative Tree-based Simulation Cloning Framework for Finite Memory Machines

Simulation cloning is a technique in which cloned simulations whose state spaces differ partially from their parent simulation due to intervening events are spawned at runtime and concurrently advanced. It is a powerful method to carry out what-if analysis by speculatively exploring and evaluating the impact of various permutations of intervening cascade of events. Due to the exponential growth in the number of possible clones even for a small number of distinct intervening events, the practical efficacy of the approach is often severely limited by the maximum available memory of the computing host. In this paper, we introduce a novel speculative simulation cloning framework that executes a simulation cloning campaign capable of efficiently exploring an exponentially large space of clone simulations created by permutation of intervening events under a finite memory constraint. We provide a theoretical analysis of the runtime characteristics of our proposed approach and highlight its novel advantages such as memory-aware and as-long-as-needed execution. Furthermore, in support of our analytical findings and to demonstrate its practical feasibility, we implement a prototype of the cloning framework on a shared memory system and report its performance characteristics in the context of a heat diffusion simulation, and a power grid simulation subject to cascading disruptions from geomagnetic disturbances.

Simulation framework↗

DOE OSTI2020

Assessment of the Griffin Reactor Multiphysics Application Using the Empire Micro Reactor Design Concept

In late 2019, INL and ANL agreed to jointly develop the reactor physics code named Griffin based on the integration of the two code suites, MAMMOTH/Rattlesnake (INL) and MC2 - 3/PROTEUS (ANL). Griffin is being developed based on the MOOSE framework and MOOSE quality assurance procedures. This decision was made to be able to allow DOE-NE to efficiently invest funding to this area and to provide effective and timely support for existing and potential users; the latter includes industry and government organizations who are developing various types of advanced reactors in the near and long term. Since MAMMOTH/Rattlesnake has been developed based on the MOOSE framework, the INL/ANL Griffin development team agreed to build Griffin beginning with a merger of MAMMOTH and Rattlesnake into a single code and moving forward by implementing capabilities from the PROTEUS suite into Griffin. Moving forward, both ANL and INL efforts are equally invested in the Griffin project, with management support, to provide an advanced reactor multiphysics tool to assist in reactor design, optimization, and safety analysis. Much work remains in moving Griffin forward to migrate PROTEUS capabilities and to optimize performance to meet user needs. The main objective of this work is to assess the current status of Griffin capabilities in terms of performance and accuracy, to determine priorities for PROTEUS migration, and to identify capabilities and features to improve for supporting the code integration effort. For this assessment, the Empire micro reactor problem that was developed in the ARPA-E MEITNER program was selected as an advance reactor concept of interest to the technical community. The Empire reactor problem was expanded from its original incomplete specification to be a small heat-pipe-cooled micro reactor core with ~113 cm radius and 70 cm in height, composed of 18 fuel assemblies, 12 control drums, and beryllium radial and axial reflectors. In the current model, using 5 cm axial reflectors specified in the original Empire assembly model, more than 10% of neutrons leak axially and through the empty center safety hole, as well as through heat pipe channels in fuel assembly elements that extend through the top reflector region. Several calculation models of the core were defined for systematic assessment, including 2-D and 3-D fuel assemblies and whole cores with cylindrical boundaries. Cross sections were generated using Serpent 2, and meshes were produced using the Argonne mesh tool or the INL neutronics meshing tools combined with CUBIT. Cross sections and meshes were converted to the ISOXML and Exodus formats, respectively, so that Griffin and PROTEUS could use consistent data for solving the reactor problems. With the prepared cross sections and meshes, PROTEUS was run first to ensure that all input data were correctly generated and input options in terms of angle, mesh, and energy group were accurately determined. Comparisons against Serpent 2 solutions were made in terms of eigenvalue and pin power. The same calculations and comparisons were then conducted using Griffin. For the fuel assembly and whole core problems, the PROTEUS eigenvalues agreed well with reference Serpent 2 solutions within 100 and 30 pcm, respectively, and pin power differences relative to Serpent 2 were overall less than 2.2% and RMS 0.8% for the whole core models. This indicated that all input data were properly prepared. Using the same data, Griffin was run selecting the SAAF-CFEM SN solver with Legendre-Gaussian quadrature and NDA and DSA for acceleration. It was found that the SAAF-CFEM solver of Griffin required finer meshes to achieve eigenvalue and pin power solutions in good agreement with Serpent 2, consequently requiring more memory requirement and longer computation time. On the other hand, the SPH-Diffusion 2-D core calculations performed using Griffin were able to recover the exact eigenvalue from the reference Serpent 2 solutions, resulting in a pin-power distribution with an RMS of 0.6% and maximum absolute difference of less than 1.4%. The runtimes for SPH-Diffusion for the 2-D core were less than 3 minutes on 40 cores. During this evolution of this evaluation, many updates were made in Griffin by the Griffin development team of INL (focusing on software updates) and ANL (reviewing and supporting software updates) to complete this assessment. Observations from the code assessment are presented in the conclusion section of this report, followed by a discussion of recommendations for future work.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2025

DIF3D-VARIANT 12.0: Updates and New Features

The DIF3D code has been a workhorse of fast reactor analysis work at Argonne National Laboratory for over 40 years. In 1995, a transport option called VARIANT was added to DIF3D to improve the flux solutions for fast reactor problems which we term DIF3D-VARIANT today. DIF3D-VARIANT performs nodal neutron transport calculations using P N or SP N theory in Cartesian and hexagonal two- and three-dimensional geometries. The limited computing capabilities of the time restricted DIF3D-VARIANT to use at most a 6 th order spatial approximation combined with a P3 flux approximation and P1 scattering kernel for a 33 group structure on most studied reactor problems. Computer capabilities have increased steadily since 1995 and today much larger space-angle-energy approximations are possible. This manuscript serves as an update to the theory section of the original DIF3D-VARIANT manual and details more than twenty years of changes made to DIF3D to make version 12 which was released on November 1 st , 2024. The primary focus of the initial work was to extend the space-angle approximations available in DIF3D-VARIANT such that the error due to transport approximations could be better understood. This work was started and completed in 2002 and marked the official version 10. Unfortunately, those higher order approximations could not be used at that time due to the memory constraints of the BPOINTER part of DIF3D (limited to 2 GB). In version 11, completed in 2012, BPOINTER was circumvented in DIF3D-VARIANT for the largest arrays by introducing a Fortran 90 module called LMA (Large Memory Array). This seamlessly replaces all of the functionality of the BPOINTER concept, but it allows 64 bit addressing for every array such that they can be larger than 2 GB. It is now common for DIF3D-VARIANT jobs to consume 50 GB of memory on modern workstations when using high order space-angle approximations and a large number of groups. Many improvements were made to version 11 from 2012 to 2022 when work to create version 12 started. For version 12, several parts of DIF3D were updated to improve performance and thread parallelism was introduced to further reduce the runtime. Numerous minor bugs were discovered in DIF3D-VARIANT as part of the process of creating the perturbation and sensitivity code PERSENT. All of these algorithmic problems were identified in the transition from version 10 to version 11 which prevented DIF3D-VARIANT from running efficiently and reliably. Firstly, the coarse mesh rebalance scheme would routinely diverge and a study detailed in this report demonstrates how it was also typically not effective. This is not a failure of the coarse mesh rebalance methodology, but a failure of its implementation in DIF3D-VARIANT for hexagonal geometries. The fission source extrapolation algorithm was also found to be unreliable on larger group structure problems, leading to divergence in some cases and a negligible improvement in performance overall. Finally, the “Omega” acceleration applied to the partial current solver routine of DIF3D-VARIANT was found to cause DIF3D-VARIANT to converge to the wrong answer. To resolve these issues, both the coarse mesh rebalance and fission source extrapolation were permanently disabled in version 11. The Tchebychev acceleration was put in as a temporary reliable alternative but it is generally inferior to coarse mesh rebalance or coarse mesh finite difference. For the Omega acceleration, the factor was restricted to guarantee that it would not cause follow-on errors in PERSENT. Due to limited funding to support maintenance and development of DIF3D in the last 10 years, no effort was spent since to resolve the outer iteration acceleration. Except for the threading work, all of the changes discussed in this manuscript refer to changes made between version 10 and version 11. Performance comparisons are done to demonstrate the improvements from version 9 to version 12. As will be demonstrated, the updated versi

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2020

A Code-Agnostic Driver Application for Coupled Neutronics and Thermal-Hydraulic Simulations

While the literature has numerous examples of Monte Carlo and computational fluid dynamics (CFD) coupling, most are hard-wired codes intended primarily for research rather than as standalone, general-purpose applications. In this work, we describe an open source application, ENRICO, that enables coupled neutronic and thermal-hydraulic simulations between multiple codes that can be chosen at runtime (as opposed to a coupling between two specific codes). The application has been designed such that the control flow logic, domain mapping, nonlinear fixed-point iteration, solution transfers, and convergence checks are all agnostic to the underlying physics solvers used. Special emphasis has also been placed on enabling efficient execution on distributed-memory computing environments. The transfer of solution fields between solvers is performed in memory rather than through filesystem I/O. Additionally, solvers can be configured to run on overlapping or disjoint sets of processes. To date, coupling with the OpenMC and Shift Monte Carlo codes, the Nek5000 CFD code, and a simplified heat diffusion and subchannel solver has been implemented in ENRICO. We present results for coupled simulations of a single light-water reactor fuel assembly based on the NuScale reactor using various combinations of the physics solvers. For this problem, the coupled simulations are shown to converge in about four Picard iterations. A comparison of the heat source and temperature distributions computed by ENRICO using OpenMC coupled with Nek5000 and Shift coupled with Nek5000 illustrates remarkable agreement between the codes.

42 ENGINEERING↗

DOE OSTI2022

GASNet-EX RMA Communication Performance on Recent Supercomputing Systems

Partitioned Global Address Space (PGAS) programming models, typified by systems such as Unified Parallel C (UPC) and Fortran coarrays, expose one-sided Remote Memory Access (RMA) communication as a key building block for High Performance Computing (HPC) applications. Architectural trends in supercomputing make such programming models increasingly attractive, and newer, more sophisticated models such as UPC++, Legion and Chapel that rely upon similar communication paradigms are gaining popularity. GASNet-EX is a portable, open-source, high-performance communication library designed to efficiently support the networking requirements of PGAS runtime systems and other alternative models in emerging exascale machines. The library is an evolution of the popular GASNet communication system, building upon 20 years of lessons learned. We present microbenchmark results which demonstrate the RMA performance of GASNet-EX is competitive with MPI implementations on four recent, high-impact, production HPC systems. These results are an update relative to previously published results on older systems. The networks measured here are representative of hardware currently used in six of the top ten fastest supercomputers in the world, and all of the exascale systems on the U.S. DOE road map.

Hargrove, Paul H↗

DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

A Co-design Framework for Online Data Analysis and Reduction

Science applications preparing for the exascale era are increasingly exploring in situ computations comprising of simulation-analysis-reduction pipelines coupled in-memory. Efficient composition and execution of such complex pipelines for a target platform is a codesign process that evaluates the impact and tradeoffs of various application- and system-specific parameters. In this article, we describe a toolset for automating performance studies of composed HPC applications that perform online data reduction and analysis. We describe Cheetah, a new framework for composing parametric studies on coupled applications, and Savanna, a runtime engine for orchestrating and executing campaigns of codesign experiments. Furthermore, this toolset facilitates understanding the impact of various factors such as process placement, synchronicity of algorithms, and storage versus compute requirements for online analysis of large data. Ultimately, we aim to create a catalog of performance results that can help scientists understand tradeoffs when designing next-generation simulations that make use of online processing techniques. We illustrate the design of Cheetah and Savanna, and present application examples that use this framework to conduct codesign studies on small clusters as well as leadership class supercomputers.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Sparse Binary Matrix-Vector Multiplication on Neuromorphic Computers

Neuromorphic computers offer the opportunity for low-power, efficient computation. Though they have been primarily applied to neural network tasks, there is also the opportunity to leverage the inherent characteristics of neuromorphic computers (low power, massive parallelism, collocated processing and memory) to perform non-neural network tasks. Here, we demonstrate how an approach for performing sparse binary matrix-vector multiplication on neuromorphic computers. We describe the approach, which relies on the connection between binary matrix-vector multiplication and breadth first search, and we introduce the algorithm for performing this calculation in a neuromorphic way. We validate the approach in simulation. Finally, we provide a discussion of the runtime of this algorithm and discuss where neuromorphic computers in the future may have a computational advantage when performing this computation.

Schuman, Catherine↗

DOE OSTI2023

Code modernization strategies for short-range non-bonded molecular dynamics simulations

Modern HPC systems are increasingly relying on greater core counts and wider vector registers. Thus, applications need to be adapted to fully utilize these hardware capabilities. One class of applications that can benefit from this increase in parallelism are molecular dynamics simulations. In this paper, we describe our efforts at modernizing the ESPResSo++ simulation package for molecular dynamics by restructuring its particle data layout for efficient memory accesses and applying vectorization techniques to benefit the calculation of short-range non-bonded forces, which results in an overall three times speedup and serves as a baseline for further optimizations. We also implement fine-grained parallelism for multi-core CPUs through HPX, a C++ runtime system which uses lightweight threads and an asynchronous many-task approach to maximize concurrency. Our goal is to evaluate the performance of an HPX-based approach compared to the bulk-synchronous MPI-based implementation. This requires the introduction of an additional layer to the domain decomposition scheme that defines the task granularity. On spatially inhomogeneous systems, which impose a corresponding load-imbalance in traditional MPI-based approaches, we demonstrate that by choosing an optimal task size, the efficient work-stealing mechanisms of HPX can overcome the overhead of communication resulting in an overall 1.4 times speedup compared to the baseline MPI version.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2026

How efficiently can AI recognize Wireless Devices?

This poster presents a hardware benchmarking methodology for a 3-layer CNN waveform classifier deployed using ONNX Runtime on an NVIDIA Jetson AGX Orin. The dataset consist of 9 signal types, -30 to +30 dB SNR with 5dB increments. Benchmarking on the Jetson AGX Orin gave an accuracy of 91.9% and GPU throughput of 107,120 predictions/sec (23× faster than CPU). The Jetson GPU reached approximately 27M samples/sec with stable performance but fell below the 40 MHz rate needed for real-time radio feeds. Sustained testing of 5 minutes confirmed stable performance with no memory leaks, establishing a reproducible benchmarking baseline for future edge-deployment optimization.

99 - GENERAL AND MISCELLANEOUS↗

DOE OSTI2025

Experiences with SYCL on AMD GPUs with Kokkos

With the recent diversification of the hardware landscape in the high-performance computing (HPC) community, performance-portability solutions are becoming more and more important. One of the most popular choices is Kokkos, which recently became a Linux Foundation project. Most of its development is supported by the US Department of Energy and the French Alternative Energies and Atomic Energy Commission. Kokkos is implemented as a C++ library with multiple backends to support CPUs as well as various GPU architectures. These backends include OpenMP, CUDA, HIP, and also SCYL. This approach enables users to leverage the preferred vendor toolchain for the respective platform (e.g. CUDA, ROCm, OneAPI). The SYCL backend is used to target Intel GPUs, in particular to support the Aurora exascale supercomputer. However, SYCL itself also offers a large degree of portability, and in fact Kokkos’ CI for SYCL has been running on NVIDIA hardware due to a lack of access to Intel GPUs. In this report, we describe our experience with using Kokkos SYCL backend on AMD GPUs targeting the Frontier supercomputer at Oak Ridge National Laboratory. The two major SYCL implementations are DPC++ and AdaptiveCpp. While the Kokkos SYCL backend has been implemented using the former, the latter was the first implementation to target AMD GPUs. We will discuss the experience with both of these SYCL implementations in terms of functionality and performance. Using Kokkos to evaluate SYCL toolchains has a number of benefits. Kokkos’ use of SYCL is fairly complex, exercising features such as graphs, relocatable device functions, atomics – including for non-arithmetic types, as well as pinned and page migratable memory allocations. Kokkos also needs to implement capabilities such as Kokkos’ hierarchical parallelism that are not a straight-forward mapping to SYCL capabilities. Furthermore, a large number of libraries and applications that represent diverse use cases are implemented in Kokkos, providing readily available test cases for a toolchain evaluation. Preliminary results show that support for AMD GPUs in DPC++ is much less mature than for NVIDIA GPUs or Intel GPUs. While the situation has improved significantly over the last year, we still encounter many runtime failures, dispatching problems, and code generation issues. With AdaptiveCpp the challenges arise even earlier in the evaluation process. Since Kokkos’ SYCL implementation is largely focused on supporting Intel GPUs, we opted to leverage SYCL extensions which are available in DPC++ but not in AdaptiveCpp. Furthermore, AdaptiveCpp appears to be less conformant with the SYCL2020 standard which Kokkos relies on. In some cases, we are able to work around the lack of feature support, in other cases we have to disable certain Kokkos capabilities to evaluate the toolchain. Our evaluation will leverage Kokkos’ unit tests to establish basic functionality and feature completeness. We then use simple benchmarks for components of a CG implementation as a measure of usability and performance of the SYCL toolchains.

97 MATHEMATICS AND COMPUTING↗