Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “CRAY COMPUTERS”

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 739 records · Page 41

Comparative Performance Analysis of Coarse Solvers for Algebraic Multigrid on Multicore and Manycore Architectures

In this paper, we study the performance of a two-level algebraic-multigrid algorithm, with a focus on the impact of the coarse-grid solver on performance. We consider two algorithms for solving the coarse-space systems: the preconditioned conjugate gradient method and a new robust HSS-embedded low-rank sparse-factorization algorithm. Our test data comes from the SPE Comparative Solution Project for oil-reservoir simulations. We contrast the performance of our code on one 12-core socket of a Cray XC30 machine with performance on a 60-core Intel Xeon Phi coprocessor. To obtain top performance, we optimized the code to take full advantage of fine-grained parallelism and made it thread-friendly for high thread count. We also developed a bounds-and-bottlenecks performance model of the solver which we used to guide us through the optimization effort, and also carried out performance tuning in the solver’s large parameter space. Finally, as a result, significant speedups were obtained on both machines.

97 MATHEMATICS AND COMPUTING↗

Performance Assessment of OpenMP Compilers Targeting NVIDIA V100 GPUs

Heterogeneous systems are becoming increasingly prevalent. In order to exploit the rich compute resources of such systems, robust programming models are needed for application developers to seamlessly migrate legacy code from today’s systems to tomorrow’s. Over the past decade and more, directives have been established as one of the promising paths to tackle programmatic challenges on emerging systems. This work focuses on applying and demonstrating OpenMP offloading directives on five proxy applications. We observe that the performance varies widely from one compiler to the other; a crucial aspect of our work is reporting best practices to application developers who use OpenMP offloading compilers. While some issues can be worked around by the developer, there are other issues that must be reported to the compiler vendors. By restructuring OpenMP offloading directives, we gain an 18x speedup for the su3 proxy application on NERSC’s Cori system when using the Clang compiler, and a 15.7x speedup by switching max reductions to add reductions in the laplace mini-app when using the Cray-llvm compiler on Cori.

Davis, Josh↗

Multithreaded Algorithms for Maximum Matching in Bipartite Graphs

Abstract—We design, implement, and evaluate algorithms for computing a matching of maximum cardinality in a bipartite graph on multi-core and massively multithreaded computers. As computers with larger number of slower cores dominate the commodity processor market, the design of multithreaded algorithms to solve large matching problems becomes a necessity. Recent work on serial algorithms based on searching for augmenting paths for this problem have shown that their performance is sensitive to the order in which the vertices are processed for matching. In a multithreaded environment, imposing a serial order in which vertices are considered for matching would lead to loss of concurrency and performance. But this raises the question: Would parallel matching algorithms on multithreaded machines improve performance over a serial algorithm? We answer this question in the affirmative. We report efficient multithreaded implementations of two key algorithms (Hopcroft- Karp based on breadth-first-search, and Pothen-Fan based on depth-first-search) and their variants, combined with the Karp- Sipser initialization algorithm. We report extensive results and insights using three shared-memory platforms (a 48-core AMD Opteron, a 32-core Intel Nehalem, and a 128-processor Cray XMT) on a representative set of real-world and synthetic graphs. To the best of our knowledge, this is the first extensive study of augmentation-based parallel algorithms for bipartite cardinality matching.

Azad, Md Ariful↗

NESSUS/NASTRAN Interface

The NESSUS probabilistic analysis computer program has been developed with a built-in finite element analysis program NESSUS/FEM. However, the NESSUS/FEM program is specialized for engine structures and may not contain sufficient features for other applications. In addition, users often become well acquainted with a particular finite element code and want to use that code for probabilistic structural analysis. For these reasons, this work was undertaken to develop an interface between NESSUS and NASTRAN such that NASTRAN can be used for the finite element analysis and NESSUS can be used for the probabilistic analysis. In addition, NESSUS was restructured such that other finite element codes could be more easily coupled with NESSUS. NESSUS has been enhanced such that NESSUS will modify the NASTRAN input deck for a given set of random variables, run NASTRAN and read the NASTRAN result. The coordination between the two codes is handled automatically. The work described here was implemented within NESSUS 6.2 which was delivered to NASA in September 1995. The code runs on Unix machines: Cray, HP, Sun, SGI and IBM. The new capabilities have been implemented such that a user familiar with NESSUS using NESSUS/FEM and NASTRAN can immediately use NESSUS with NASTRAN. In other words, the interface with NASTRAN has been implemented in an analogous manner to the interface with NESSUS/FEM. Only finite element specific input has been changed. This manual is written as an addendum to the existing NESSUS 6.2 manuals. We assume users have access to NESSUS manuals and are familiar with the operation of NESSUS including probabilistic finite element analysis. Update pages to the NESSUS PFEM manual are contained in Appendix E. The finite element features of the code and the probalistic analysis capabilities are summarized.

Millwater, Harry↗

Rendering and Compositing Infrastructure Improvements to VisIt for Insitu Rendering

Compared to posthoc rendering, insitu rendering often generates larger numbers of images, as a result rendering performance and scalability are critical in the insitu setting. In this work we present improvements to VisIt's rendering and compositing infrastructure that deliver increased performance and scalability in both posthoc and insitu settings. We added the capability for alpha blend compositing and use it with ordered compositing when datasets have disjoint block domain decomposition to optimize the rendering of transparent geometry. We also made improvements that increase overall efficiency by reducing communication and data movement and have addressed a number of performance issues. We structured our code to take advantage of SIMD parallelization and use threads to overlap communication and compositing. We tested our improvements on a 20 core workstation using 8 cores to render geometry generated from a $256^3$ cosmology dataset and on a Cray XC31 using 512 cores to render geometry generated from a $2000^2 \times 800$ plasma dataset. Our results show that ordered compositing provides a speed up of up to $4 \times$ over the current sort first strategy. The other improvements resulted in modest speed up with one notable exception where we achieve up to $40 \times$ speed up of rendering and compositing of opaque geometry when both opaque and transparent geometry are rendered together. We also investigated the use of depth peeling, but found that the implementation provided by VTK is substantially slower,both with and without GPU acceleration, than a local camera order sort.

97 MATHEMATICS AND COMPUTING↗

HARP: A Dynamic Inertial Spectral Partitioner

Partitioning unstructured graphs is central to the parallel solution of computational science and engineering problems. Spectral partitioners, such recursive spectral bisection (RSB), have proven effecfive in generating high-quality partitions of realistically-sized meshes. The major problem which hindered their wide-spread use was their long execution times. This paper presents a new inertial spectral partitioner, called HARP. The main objective of the proposed approach is to quickly partition the meshes at runtime in a manner that works efficiently for real applications in the context of distributed-memory machines. The underlying principle of HARP is to find the eigenvectors of the unpartitioned vertices and then project them onto the eigerivectors of the original mesh. Results for various meshes ranging in size from 1000 to 100,000 vertices indicate that HARP can indeed partition meshes rapidly at runtime. Experimental results show that our largest mesh can be partitioned sequentially in only a few seconds on an SP2 which is several times faster than other spectral partitioners while maintaining the solution quality of the proven RSB method. A parallel WI version of HARP has also been implemented on IBM SP2 and Cray T3E. Parallel HARP, running on 64 processors SP2 and T3E, can partition a mesh containing more than 100,000 vertices into 64 subgrids in about half a second. These results indicate that graph partitioning can now be truly embedded in dynamically-changing real-world applications.

Simon, Horst D.↗

GASNet-EX Performance Improvements Due to Specialization for the Cray Aries Network

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 on future exascale machines. This paper reports on the improvements in performance observed on Cray XC-series systems due to enhancements made to the GASNet-EX software. These enhancements, known as "specializations", primarily consist of replacing network-independent implementations of several recently added features with implementations tailored to the Cray Aries network. Performance gains from specialization include (1) Negotiated-Payload Active Messages improve bandwidth of a ping-pong test by up to 14%, (2) Immediate Operations reduce running time of a synthetic benchmark by up to 93%, (3) non-bulk RMA Put bandwidth is increased by up to 32%, (4) Remote Atomic performance is 70% faster than the reference on a point-to-point test and allows a hot-spot test to scale robustly, and (5) non-contiguous RMA interfaces see up to 8.6x speedups for an intra-node benchmark and 26% for inter-node. These improvements are all available in GASNet-EX version 2018.3.0 and later.

97 MATHEMATICS AND COMPUTING↗

Ames Optimized TCA Configuration

Configuration design at Ames was carried out with the SYN87-SB (single block) Euler code using a 193 x 49 x 65 C-H grid. The Euler solver is coupled to the constrained (NPSOL) and the unconstrained (QNMDIF) optimization packages. Since the single block grid is able to model only wing-body configurations, the nacelle/diverter effects were included in the optimization process by SYN87's option to superimpose the nacelle/diverter interference pressures on the wing. These interference pressures were calculated using the AIRPLANE code. AIRPLANE is an Euler solver that uses a unstructured tetrahedral mesh and is capable of computations about arbitrary complete configurations. In addition, the buoyancy effects of the nacelle/diverters were also included in the design process by imposing the pressure field obtained during the design process onto the triangulated surfaces of the nacelle/diverter mesh generated by AIRPLANE. The interference pressures and nacelle buoyancy effects are added to the final forces after each flow field calculation. Full details of the (recently enhanced) ghost nacelle capability are given in a related talk. The pseudo nacelle corrections were greatly improved during this design cycle. During the Ref H and Cycle 1 design activities, the nacelles were only translated and pitched. In the cycle 2 design effort the nacelles can translate vertically, and pitch to accommodate the changes in the lower surface geometry. The diverter heights (between their leading and trailing edges) were modified during design as the shape of the lower wing changed, with the drag of the diverter changing accordingly. Both adjoint and finite difference gradients were used during optimization. The adjoint-based gradients were found to give good direction in the design space for configurations near the starting point, but as the design approached a minimum, the finite difference gradients were found to be more accurate. Use of finite difference gradients was limited by the CPU time limit available on the Cray machines. A typical optimization run using finite difference gradients can use only 30 to 40 design variables and one optimization iteration within the 8 hour queue limit for the chosen grid size and convergence level. The efficiency afforded by the adjoint method allowed for 50-120 design variables and 5-10 optimization iterations in the 8 hour queue. Geometric perturbations to the wing and fuselage were made using the Hicks/Henne (HH) shape functions. The HH functions were distributed uniformly along the chords of the wing defining sections and lofted linearly. During single-surface design, constraints on thickness and volume at selected wing stations were imposed. Both fuselage camber and cross-sectional area distributions were permitted to change during design. The major disadvantage to the use of these functions is the inherent surface waviness produced by repeated use of such functions. Many smoothing operations were required following optimization runs to produce a configuration with reasonable smoothness. Wagner functions were also used on the wing sections but were never used on the fuselage. The Wagner functions are a family of increasingly oscillatory functions that have also been used extensively in airfoil design. The leading and trailing edge regions of the wing were designed by use of polynomial and monomial functions respectively. Twist was attempted but was abandoned because of little performance improvement available from changing the baseline twist.

Cliff, Susan E.↗

GPU Lifetimes on Titan Supercomputer: Survival Analysis and Reliability

George Ostrouchov, Don Maxwell, Rizwan Ashraf, Mallikarjun Shankar, and James Rogers. 2020. GPU Lifetimes on Titan Supercomputer: Survival Analysis and Reliability. In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (SC '20). Association for Computing Machinery, New York, NY, USA. Data and code for SC20 paper about Titan GPU reliability analysis: https://github.com/olcf/TitanGPULife. Includes R code to generate graphics for paper and additional analyses. See code/README for instructions. Includes original Titan GPU reliability data on over 100,000 collective hours of operation: data/titan.gpu.history.txt - history data, data/titan.service.txt - service nodes for exclusion. Includes output data files produced by code/TitanGPUmodel.Rmd: data/gc_full.csv - cleaned up data (see paper and R code); data/gc_summary_loc.csv - one record per GPU (variables: SN, time, nlife, nloc, last, col, row, cage, slot, node, max_loc_events, time_max_loc, dbe, dbe_loc, otb, otb_loc, out, batch, days, years, dead, dead_otb, dead_dbe) (see paper and R code). Includes .Rmd analysis document as TitanGPUmode.html. Includes Python code to process data/gc_full.csv into graphics from time-between-failure analyses: See code/tbf-analyses/README for instructions.

42 ENGINEERING↗

Analyzing the Interplay of Failures and Workload on a Leadership-Class Supercomputer

The unprecedented computational power of cur- rent supercomputers now makes possible the exploration of complex problems in many scientific fields, from genomic analysis to computational fluid dynamics. Modern machines are powerful because they are massive: they assemble millions of cores and a huge quantity of disks, cards, routers, and other components. But it is precisely the size of these machines that glooms the future of supercomputing. A system that comprises many components has a high chance to fail, and fail often. In order to make the next generation of supercomputers usable, it is imperative to use some type of fault tolerance platform to run applications on large machines. Most fault tolerance strategies can be optimized for the peculiarities of each system and boost efficacy by keeping the system productive. In this paper, we aim to understand how failure characterization can improve resilience in several layers of the software stack: applications, runtime systems, and job schedulers. We examine the Titan supercomputer, one of the fastest systems in the world. We analyze a full year of Titan in production and distill the failure patterns of the machine. By looking into Titan s log files and using the criteria of experts, we provide a detailed description of the types of failures. In addition, we inspect the job submission files and describe how the system is used. Using those two sources, we cross correlate failures in the machine to executing jobs and provide a picture of how failures affect the user experience. We believe such characterization is fundamental in developing appropriate fault tolerance solutions for Cray systems similar to Titan.

Cray↗

Bringing HPE Slingshot 11 support to Open MPI

The Cray HPE Slingshot 11 network is used on the new exascale systems arriving at the U.S. Department of Energy (DoE) laboratories (e.g., Frontier, Aurora, Perlmutter). As such, the support of this network is an important capability to meet the needs of exascale applications. Here, this article highlights recent work to develop supporting infrastructure to enable Open MPI to efficiently support these new platforms. A key component of this effort involves development of a new Open Fabrics Interface (OFI) provider, LinkX. We discuss the design and development of enhancements that take advantage of the new Slingshot 11 network and AMD GPUs. We include performance data from tests on the Frontier supercomputer using synthetic communication benchmarks, and the vendor provided MPI as a baseline for comparison. The tests demonstrate full functionality of Open MPI on the system and initial results show favorable performance when compared to the highly tuned vendor implementation.

97 MATHEMATICS AND COMPUTING↗

Preparing NERSC users for Cori, a Cray XC40 system with Intel many integrated cores

The newest NERSC supercomputer Cori is a Cray XC40 system consisting of 2,388 Intel Xeon Haswell nodes and 9,688 Intel Xeon-Phi “Knights Landing” (KNL) nodes. Compared to the Xeon-based clusters NERSC users are familiar with, optimal performance on Cori requires consideration of KNL mode settings; process, thread, and memory affinity; fine-grain parallelization; vectorization; and use of the high-bandwidth MCDRAM memory. This paper describes our efforts preparing NERSC users for KNL through the NERSC Exascale Science Application Program, Web documentation, and user training. We discuss how we configured the Cori system for usability and productivity, addressing programming concerns, batch system configurations, and default KNL cluster and memory modes. Here, system usage data, job completion analysis, programming and running jobs issues, and a few successful user stories on KNL are presented.

97 MATHEMATICS AND COMPUTING↗

Implementation and Optimization of miniGMG - a Compact Geometric Multigrid Benchmark

Multigrid methods are widely used to accelerate the convergence of iterative solvers for linear systems used in a number of different application areas. In this report, we describe miniGMG, our compact geometric multigrid benchmark designed to proxy the multigrid solves found in AMR applications. We explore optimization techniques for geometric multigrid on existing and emerging multicore systems including the Opteron-based Cray XE6, Intel Sandy Bridge and Nehalem-based Infiniband clusters, as well as manycore-based architectures including NVIDIA's Fermi and Kepler GPUs and Intel's Knights Corner (KNC) co-processor. This report examines a variety of novel techniques including communication-aggregation, threaded wavefront-based DRAM communication-avoiding, dynamic threading decisions, SIMDization, and fusion of operators. We quantify performance through each phase of the V-cycle for both single-node and distributed-memory experiments and provide detailed analysis for each class of optimization. Results show our optimizations yield significant speedups across a variety of subdomain sizes while simultaneously demonstrating the potential of multi- and manycore processors to dramatically accelerate single-node performance. However, our analysis also indicates that improvements in networks and communication will be essential to reap the potential of manycore processors in large-scale multigrid calculations.

97 MATHEMATICS AND COMPUTING↗

Dose planning objectives in anal canal cancer IMRT: the TROG ANROTAT experience

Intensity modulated radiotherapy (IMRT) is ideal for anal canal cancer (ACC), delivering high doses to irregular tumour volumes whilst minimising dose to surrounding normal tissues. Establishing achievable dose objectives is a challenge. The purpose of this paper was to utilise data collected in the Assessment of New Radiation Oncology Treatments and Technologies (ANROTAT) project to evaluate the feasibility of ACC IMRT dose planning objectives employed in the Australian situation. Ten Australian centres were randomly allocated three data sets from 15 non-identifiable computed tomography data sets representing a range of disease stages and gender. Each data set was planned by two different centres, producing 30 plans. All tumour and organ at risk (OAR) contours, prescription and dose constraint details were provided. Dose–volume histograms (DVHs) for each plan were analysed to evaluate the feasibility of dose planning objectives provided. All dose planning objectives for the bone marrow (BM) and femoral heads were achieved. Median planned doses exceeded one or more objectives for bowel, external genitalia and bladder. This reached statistical significance for bowel V30 (P = 0.04), V45 (P < 0.001), V50 (P < 0.001), external genitalia V20 (P < 0.001) and bladder V35 (P < 0.001), V40 (P = 0.01). Gender was found to be the only significant factor in the likelihood of achieving the bowel V50 (P = 0.03) and BM V30 constraints (P = 0.04). The dose planning objectives used in the ANROTAT project provide a good starting point for ACC IMRT planning. To facilitate clinical implementation, it is important to prioritise OAR objectives and recognise factors that affect the achievability of these objectives.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

HiCOO: Hierarchical cooperation for scalable communication in Global Address Space programming models on Cray XT systems

Global Address Space (GAS) programming models enable a convenient, shared-memory style addressing model. Typically this is realized through one-sided operations that can enable asynchronous communication and data movement. With the size of petascale systems reaching 10,000s of nodes and 100,000s of cores, the underlying runtime systems face critical challenges in (1) scalably managing resources (such as memory for communication buffers), and (2) gracefully handling unpredictable communication patterns and any associated contention. For any solution that addresses these resource scalability challenges, equally important is the need to maintain the performance of GAS programming models. In this paper, we describe a Hierarchical COOperation (HiCOO) architecture for scalable communication in GAS programming models. HiCOO formulates a cooperative communication architecture: with inter-node cooperation amongst multiple nodes (a.k.a multinode) and hierarchical cooperation among multinodes that are arranged in various virtual topologies. We have implemented HiCOO for a popular GAS runtime library, Aggregate Remote Memory Copy Interface (ARMCI). By extensively evaluating different virtual topologies in HiCOO in terms of their impact to memory scalability, network contention, and application performance, we identify MFCG as the most suitable virtual topology. The resulting HiCOO architecture is able to realize scalable resource management and achieve resilience to network contention, while at the same time maintaining or enhancing the performance of scientific applications. In one case, it reduces the total execution time of an NWChem application by 52%.

Computer Science↗

Performance Evaluation of Darshan 3.0.0 on the Cray XC30

Abstract—Darshan is a lightweight I/O characterization tool used to gather and summarize salient I/O workload statistics from HPC applications. Darshan was designed to minimize any possible perturbations of an application’s performance, leading it to be enabled by default on a number of production HPC systems. For each file accessed by a given application, Darshan records the count and types of I/O operations performed, histograms of access sizes, cumulative timers on the amount of time spent doing I/O, and other statistical data. This type of data has proved invaluable in understanding and improving the I/O performance of HPC applications. Darshan 3.0.0 is the new modularized version of the traditional Darshan library and file format, allowing users to easily add more in-depth I/O characterization data to Darshan logs. In this work we perform an empirical evaluation of Darshan 3.0.0 to ensure that it continues to meet performance expectations for broad deployment. In particular, we evaluate the imposed overhead on instrumented I/O operations, time taken to shut down and generate corresponding Darshan log files, and resultant log file sizes for different workloads. These performance results are compared to results of Darshan 2.3.0 on the Edison XC30 system at NERSC to determine whether the new version is lightweight enough to run full-time on production HPC systems. Our evaluation shows that Darshan has limited impact on application I/O performance and can fully generate a corresponding log file for most application workloads in under two seconds.

97 MATHEMATICS AND COMPUTING↗

SMC 2021 : Analyzing Resource Utilization and User Behavior on Titan Supercomputer

Resource utilization statistics of submitted jobs on a supercomputer can help us understand how users from various scientific domains use HPC platforms and better design a job scheduler. We explore to generate insight regarding workload distribution and usage pattern domains from job scheduler trace, GPU failure information, and project-specific information collected from Titan supercomputer. Furthermore, we want to know how the scheduler performance varies over time and how the users' scheduling behavior changes following a system failure. These observations have the potential to provide valuable insight, which is helpful to prepare for system failures. These practices will help us develop and apply novel machine learning algorithms in understanding system behavior, requirement, and better scheduling of HPC systems. There are two datasets, RUR and GPU. RUR: This dataset is the job scheduler traces collected from the Titan supercomputerfrom 01/01/2015 to 07/31/2019 (2015.csv - 2019.csv). These were collected usingResource Utilization Report (RUR), a Cray-developed resource-usage data collectionand reporting system. It contains the usage information of its critical resources (CPU,Memory, GPU, and I/O) of each running job on Titan during that period [2]. ProjectAreas: Every job is associated with a project ID. TheProjectAreas.csvdatasetprovides a mapping of the project ID to its domain science. GPU: There have been some hardware-related issues in the GPUs in Titan that caused some GPUs to fail, sometimes irrecoverably during some job runs. This dataset provides information regarding these failures during the execution of the submitted jobs. GPUs on Titan are uniquely identified by a serial number (SN), and they are installed in a location. A GPU can be installed in a location, then removed from that location following a failure, and then re-installed in a different location after fixing the problem. If the failure can't be recovered, the GPU might be removed entirely from Titan. There are two prominent types of failures that resulted in the removal of GPUs from Titan: Double Bit Error (DBE) and Out of the Bus (OTB). The dataset (gc_full.csv) has the following fields: 1. SN : Serial number of a GPU 2. location : The location where it is installed 3. insert : The time when it was inserted into that location 4. remove : The time when it was removed from that location 5. duration : Amount of time the GPU spent in this location 6. out : If the device was taken out entirely w/o a re-installment into a new location. 7. event : If the GPU was taken out entirely, the reason for its removal. To learn more about this dataset, please refer to the git repositoryhttps://github.com/olcf/TitanGPULifeand the related publication [1]. References [1] George Ostrouchov, Don Maxwell, Rizwan A Ashraf, Christian Engelmann, MallikarjunShankar, and James H Rogers. Gpu lifetimes on titan supercomputer: Survival analysisand reliability. InSC20: International Conference for High Performance Computing,Networking, Storage and Analysis, pages 1-14. IEEE, 2020. [2] Feiyi Wang, Sarp Oral, Satyabrata Sen, and Neena Imam. Learning from five-yearresource-utilization data of titan system. In2019 IEEE International Conference onCluster Computing (CLUSTER), pages 1-6. IEEE, 2019.

42 ENGINEERING↗