Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “CRAY”

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 199 records · Page 11

Large-Scale Numerical Simulations of Human Motion

This paper examines the feasibility of using massively-parallel and vector-processing supercomputers to solve large-scale optimal control problems for human movement. Specifically, we compare the computational expense of determining the optimal controls for the single support phase of walking using a conventional serial machine (a Silicon Graphics Personal Iris 4D25 workstation), a MIMD parallel machine (an Intel iPSC/860 comprising 128 processors), and a parallel-vector-processing machine (a Cray Y-MP 8/864). With the human body modeled as a 14 degree-of-freedom linkage actuated by 46 musculotendinous units, computation of the optimal controls for walking could take up to 3 months of CPU time on the Iris. Both the Cray Y-MP and the Intel iPSC/860 are able to reduce this time to practical levels. The optimal control solution for walking can be found with about 77 hours of CPU time on the Cray, and with about 88 hours of CPU time on the Intel. Although the overall speeds of the Cray and the Intel were found to be similar, the unique capabilities of each machine are best suited to different parts of the optimal control algorithm used. The Intel performed best in the calculation of the derivatives of the performance criterion and the constraints. In contrast, the Cray performed best during parameter optimization of the controls. These results suggest that the ideal computer architecture for solving very large-scale optimal control problems is a hybrid system in which a vector-processing machine is integrated into the communication network of a MIMD parallel machine.

Anderson, Frank C.↗

Application of high-performance computing to numerical simulation of human movement

We have examined the feasibility of using massively-parallel and vector-processing supercomputers to solve large-scale optimization problems for human movement. Specifically, we compared the computational expense of determining the optimal controls for the single support phase of gait using a conventional serial machine (SGI Iris 4D25), a MIMD parallel machine (Intel iPSC/860), and a parallel-vector-processing machine (Cray Y-MP 8/864). With the human body modeled as a 14 degree-of-freedom linkage actuated by 46 musculotendinous units, computation of the optimal controls for gait could take up to 3 months of CPU time on the Iris. Both the Cray and the Intel are able to reduce this time to practical levels. The optimal solution for gait can be found with about 77 hours of CPU on the Cray and with about 88 hours of CPU on the Intel. Although the overall speeds of the Cray and the Intel were found to be similar, the unique capabilities of each machine are better suited to different portions of the computational algorithm used. The Intel was best suited to computing the derivatives of the performance criterion and the constraints whereas the Cray was best suited to parameter optimization of the controls. These results suggest that the ideal computer architecture for solving very large-scale optimal control problems is a hybrid system in which a vector-processing machine is integrated into the communication network of a MIMD parallel machine.

NASA Discipline Musculoskeletal↗

OpenSHMEM Implementation of HPCG Benchmark, In: OpenSHMEM 2016: OpenSHMEM and Related Technologies. Enhancing OpenSHMEM for Hybrid Environments

We describe the effort to implement the HPCG benchmark using OpenSHMEM and MPI one-sided communication. Unlike the High Performance LINPACK (HPL) benchmark that places emphasis on large dense matrix computations, the HPCG benchmark is dominated by sparse operations such as sparse matrix-vector product, sparse matrix triangular solve, and long vector operations. The MPI one-sided implementation is developed using the one-sided OpenSHMEM implementation. Preliminary results comparing the original MPI, OpenSHMEM, and MPI one-sided implementations on an SGI cluster, Cray XK7 and Cray XC30 are presented. The results suggest the MPI, OpenSHMEM, and MPI one-sided implementations all obtain similar overall performance but the MPI one-sided implementation seems to slightly increase the run time for multigrid preconditioning in HPCG on the Cray XK7 and Cray XC30.

Computer Science↗

Optimizing blocking and nonblocking reduction operations for multicore systems: Hierarchical design and implementation

Many scientific simulations, using the Message Passing Interface (MPI) programming model, are sensitive to the performance and scalability of reduction collective operations such as MPI_Allreduce and MPI_Reduce. These operations are the most widely used abstractions to perform mathematical operations over all processes that are part of the simulation. In this work, we propose a hierarchical design to implement the reduction operations on multicore systems. This design aims to improve the efficiency of reductions by 1) tailoring the algorithms and customizing the implementations for various communication mechanisms in the system 2) providing the ability to configure the depth of hierarchy to match the system architecture, and 3) providing the ability to independently progress each of this hierarchy. Using this design, we implement MPI_Allreduce and MPI_Reduce operations (and its nonblocking variants MPI_Iallreduce and MPI_Ireduce) for all message sizes, and evaluate on multiple architectures including InfiniBand and Cray XT5. We leverage and enhance our existing infrastructure, Cheetah, which is a framework for implementing hierarchical collective operations to implement these reductions. The experimental results show that the Cheetah reduction operations outperform the production-grade MPI implementations such as Open MPI default, Cray MPI, and MVAPICH2, demonstrating its efficiency, flexibility and portability. On Infini-Band systems, with a microbenchmark, a 512-process Cheetah nonblocking Allreduce and Reduce achieves a speedup of 23x and 10x, respectively, compared to the default Open MPI reductions. The blocking variants of the reduction operations also show similar performance benefits. A 512-process nonblocking Cheetah Allreduce achieves a speedup of 3x, compared to the default MVAPICH2 Allreduce implementation. On a Cray XT5 system, a 6144-process Cheetah Allreduce outperforms the Cray MPI by 145%. The evaluation with an application kernel, Conjugate Gradient solver, shows that the Cheetah reductions speeds up total time to solution by 195%, demonstrating the potential benefits for scientific simulations.

Venkata, Manjunath Gorentla↗

OpenSHMEM Reference Implementation using UCCS-uGNI Transport Layer. In: PGAS '14 Proceedings of the 8th International Conference on Partitioned Global Address Space Programming Models, Article No. 18

OpenSHMEM is a library interface implementation and specification that enables the implementation of the Partitioned Global Address Space (PGAS) model. It exports modern RDMA network functionality and communication semantics to applications very efficiently. There are many closed source implementations of OpenSHMEM for modern RDMA interconnects such as InfiniBand and Cray's Gemini and Aries. Given the important role that Cray systems play in HPC, in this paper, we present an open source implementation of OpenSHMEM for Cray XE/XK/XC systems. To implement OpenSHMEM, we use the uGNI interface. uGNI is a generic interface that is designed for multiple programming models. The interface fits well the goal of UCCS. Having OpenSHMEM with UCCS-uGNI allows usage of the same implementation over multiple interconnects. This also translates into many advantages that come with common code such as resource sharing, increasing productivity because of less code maintenance, etc. Preliminary results show that OpenSHMEM-UCCS performs comparable to state-of-the-art Cray SHMEM for Put, Get, and AMO operations.

Janjusic, Tomislav↗

MapReduceXMT v. Beta 0.1

The MapReduceXMT library ports the MapReduce framework onto the Cray XMT. MapReduce is a programming paradigm and an approach to data management for unstructured problems. It has gained relevance due to its ability to map serial operations onto parallel distributed architectures, significantly improving developer/analyst productivity. The MapReduceXMT implements the key aspects of MapReduce for the Cray XMT, a massively threaded system that is inherently difficult to program. MapReduceXMT allows users to utilize the machine effectively and efficiently without extensive training in multi-threaded programming. The MapReduceXMT library ports the MapReduce framework onto the Cray XMT. MapReduce is a programming paradigm and an approach to data management for unstructured problems. It has gained relevance due to its ability to map serial operations onto parallel distributed architectures, significantly improving developer/analyst productivity. The MapReduceXMT implements the key aspects of MapReduce for the Cray XMT, a massively threaded system that is inherently difficult to program. MapReduceXMT allows users to utilize the machine effectively and efficiently without extensive training in multi-threaded programming.

Goodman, Eric↗

Approximate Weighted Matching On Emerging Manycore and Multithreaded Architectures

Graph matching is a prototypical combinatorial problem with many applications in computer science and scientific computing, but algorithms for computing optimal matchings are challenging to parallelize. Approximate matching algorithms provide an alternate route for parallelization, and in many contexts compute near-optimal matchings for large-scale graphs. We present sharedmemory parallel implementations for computing half-approximate weighted matching on state-of-the-art multicore (Intel Nehalem and AMD Magny-Cours), manycore (Nvidia Tesla and Nvidia Fermi) and massively multithreaded (Cray XMT) platforms. We provide two implementations: the first implementation uses shared work queues, and is suited to all these platforms; the second implementation is based on dataflow principles, and exploits the architectural features of the Cray XMT. Using a carefully chosen dataset that exhibits characteristics from a wide range of real-world applications, we show scalable performance across different platforms. In particular, for one instance of the input, an R-MAT graph (RMAT-G), we show speedups of: about 32 on 48 cores of an AMD Magny-Cours; 7 on 8 cores of Intel Nehalem; 3 on Nvidia Tesla and 10 on Nvidia Fermi relative to one core of Intel Nehalem; and 60 on 128 processors of Cray XMT. We demonstrate good weak and strong scaling for graphs with up to a billion edges using up to 12, 800 threads. Given the breadth of this work, we focus on simplicity and portability of software rather than excessive fine-tuning for each platform. To the best of our knowledge, this is the first such large-scale study of the half-approximate weighted matching problem on shared-memory platforms. Driven by the critical enabling role of combinatorial algorithms such as matching in scientific computing and the emergence of informatics applications, there is a growing demand to support irregular computations on current and future computing platforms. In this context, we evaluate the capability of emerging multithreaded platforms to tolerate latency induced by irregular memory access patterns, and to support fine-grained parallelism via light-weight synchronization mechanisms. By contrasting the architectural features of these platforms against the Cray XMT, which is specifically designed to support irregular memory-intensive applications, we delineate the impact of these choices on performance.

Halappanavar, Mahantesh↗

A parallel finite-difference method for computational aerodynamics

A finite-difference scheme for solving complex three-dimensional aerodynamic flow on parallel-processing supercomputers is presented. The method consists of a basic flow solver with multigrid convergence acceleration, embedded grid refinements, and a zonal equation scheme. Multitasking and vectorization have been incorporated into the algorithm. Results obtained include multiprocessed flow simulations from the Cray X-MP and Cray-2. Speedups as high as 3.3 for the two-dimensional case and 3.5 for segments of the three-dimensional case have been achieved on the Cray-2. The entire solver attained a factor of 2.7 improvement over its unitasked version on the Cray-2. The performance of the parallel algorithm on each machine is analyzed.

Swisshelm, Julie M.↗

Parallel computation in a three-dimensional elastic-plastic finite-element analysis

CRAY 2 and CRAY Y-MP evaluation runs are undertaken for a three-dimensional, elastoplastic FEM code implementation of the 'autotasking' parallel-processing technique, which is used in all code components except the matrix-equation solver. For a typical example problem in the case of a four-processor CRAY 2, a speedup factor of 2.1 was found achievable in a dedicated environment, and 1.7 in a multiuser environment; for a three-processor CRAY Y-MP, speedup in a multiuser environment was about 2.4. Only minimal effort was required to implement autotasking in this implementation code.

Shivakumar, K. N.↗

Fluid dynamics of multiphase flow in volcanic environments on the Earth, Io, and other planets

Data was compiled on H2O, CO2, S, and SO2 and it was analyzed by various techniques to prepare phase diagrams relevant to conditions on the Earth, Io, Mars, and other planets. The use of Cray models for volcanic eruptions was examined. Attempts are being made to develop equations of state of the compounds listed above to be compatible with the Cray code so that they can be called as subroutines for the volcanic models. This is a major project, as the current code uses only the equation of state for a perfect gas, and requires that the atmosphere be the same composition as the volcanic gas. The analysis of the thermodynamic data is time consuming, and was complicated by changes from a VAX/VMS system to Macintosh's, Crays using CTSS, Crays using Unicos, and Sun workstations.

Kieffer, Susan Werner↗

PAN AIR: A computer program for predicting subsonic or supersonic linear potential flows about arbitrary configurations using a higher order panel method. Volume 4: Maintenance document (version 3.0)

The Maintenance Document Version 3.0 is a guide to the PAN AIR software system, a system which computes the subsonic or supersonic linear potential flow about a body of nearly arbitrary shape, using a higher order panel method. The document describes the overall system and each program module of the system. Sufficient detail is given for program maintenance, updating, and modification. It is assumed that the reader is familiar with programming and CRAY computer systems. The PAN AIR system was written in FORTRAN 4 language except for a few CAL language subroutines which exist in the PAN AIR library. Structured programming techniques were used to provide code documentation and maintainability. The operating systems accommodated are COS 1.11, COS 1.12, COS 1.13, and COS 1.14 on the CRAY 1S, 1M, and X-MP computing systems. The system is comprised of a data base management system, a program library, an execution control module, and nine separate FORTRAN technical modules. Each module calculates part of the posed PAN AIR problem. The data base manager is used to communicate between modules and within modules. The technical modules must be run in a prescribed fashion for each PAN AIR problem. In order to ease the problem of supplying the many JCL cards required to execute the modules, a set of CRAY procedures (PAPROCS) was created to automatically supply most of the JCL cards. Most of this document has not changed for Version 3.0. It now, however, strictly applies only to PAN AIR version 3.0. The major changes are: (1) additional sections covering the new FDP module (which calculates streamlines and offbody points); (2) a complete rewrite of the section on the MAG module; and (3) strict applicability to CRAY computing systems.

Purdon, David J.↗

Performance of a Bounce-Averaged Global Model of Super-Thermal Electron Transport in the Earth's Magnetic Field

In this paper, we report the results of our recent research on the application of a multiprocessor Cray T916 supercomputer in modeling super-thermal electron transport in the earth's magnetic field. In general, this mathematical model requires numerical solution of a system of partial differential equations. The code we use for this model is moderately vectorized. By using Amdahl's Law for vector processors, it can be verified that the code is about 60% vectorized on a Cray computer. Speedup factors on the order of 2.5 were obtained compared to the unvectorized code. In the following sections, we discuss the methodology of improving the code. In addition to our goal of optimizing the code for solution on the Cray computer, we had the goal of scalability in mind. Scalability combines the concepts of portabilty with near-linear speedup. Specifically, a scalable program is one whose performance is portable across many different architectures with differing numbers of processors for many different problem sizes. Though we have access to a Cray at this time, the goal was to also have code which would run well on a variety of architectures.

McGuire, Tim↗

NAS Parallel Benchmarks Results 3-95

The NAS Parallel Benchmarks (NPB) were developed in 1991 at NASA Ames Research Center to study the performance of parallel supercomputers. The eight benchmark problems are specified in a "pencil and paper" fashion, i.e., the complete details of the problem are given in a NAS technical document. Except for a few restrictions, benchmark implementors are free to select the language constructs and implementation techniques best suited for a particular system. In this paper, we present new NPB performance results for the following systems: (a) Parallel-Vector Processors: CRAY C90, CRAY T90 and Fujitsu VPP500; (b) Highly Parallel Processors: CRAY T3D, IBM SP2-WN (Wide Nodes), and IBM SP2-TN2 (Thin Nodes 2); and (c) Symmetric Multiprocessors: Convex Exemplar SPPIOOO, CRAY J90, DEC Alpha Server 8400 5/300, and SGI Power Challenge XL (75 MHz). We also present sustained performance per dollar for Class B LU, SP and BT benchmarks. We also mention future NAS plans for the NPB.

Saini, Subhash↗

NAS Parallel Benchmarks Results

The NAS Parallel Benchmarks (NPB) were developed in 1991 at NASA Ames Research Center to study the performance of parallel supercomputers. The eight benchmark problems are specified in a pencil and paper fashion i.e. the complete details of the problem to be solved are given in a technical document, and except for a few restrictions, benchmarkers are free to select the language constructs and implementation techniques best suited for a particular system. In this paper, we present new NPB performance results for the following systems: (a) Parallel-Vector Processors: Cray C90, Cray T'90 and Fujitsu VPP500; (b) Highly Parallel Processors: Cray T3D, IBM SP2 and IBM SP-TN2 (Thin Nodes 2); (c) Symmetric Multiprocessing Processors: Convex Exemplar SPP1000, Cray J90, DEC Alpha Server 8400 5/300, and SGI Power Challenge XL. We also present sustained performance per dollar for Class B LU, SP and BT benchmarks. We also mention NAS future plans of NPB.

Subhash, Saini↗

Optimizing Blocking and Nonblocking Reduction Operations for Multicore Systems: Hierarchical Design and Implementation

Many scientific simulations, using the Message Passing Interface (MPI) programming model, are sensitive to the performance and scalability of reduction collective operations such as MPI Allreduce and MPI Reduce. These operations are the most widely used abstractions to perform mathematical operations over all processes that are part of the simulation. In this work, we propose a hierarchical design to implement the reduction operations on multicore systems. This design aims to improve the efficiency of reductions by 1) tailoring the algorithms and customizing the implementations for various communication mechanisms in the system 2) providing the ability to configure the depth of hierarchy to match the system architecture, and 3) providing the ability to independently progress each of this hierarchy. Using this design, we implement MPI Allreduce and MPI Reduce operations (and its nonblocking variants MPI Iallreduce and MPI Ireduce) for all message sizes, and evaluate on multiple architectures including InfiniBand and Cray XT5. We leverage and enhance our existing infrastructure, Cheetah, which is a framework for implementing hierarchical collective operations to implement these reductions. The experimental results show that the Cheetah reduction operations outperform the production-grade MPI implementations such as Open MPI default, Cray MPI, and MVAPICH2, demonstrating its efficiency, flexibility and portability. On Infini- Band systems, with a microbenchmark, a 512-process Cheetah nonblocking Allreduce and Reduce achieves a speedup of 23x and 10x, respectively, compared to the default Open MPI reductions. The blocking variants of the reduction operations also show similar performance benefits. A 512-process nonblocking Cheetah Allreduce achieves a speedup of 3x, compared to the default MVAPICH2 Allreduce implementation. On a Cray XT5 system, a 6144-process Cheetah Allreduce outperforms the Cray MPI by 145%. The evaluation with an application kernel, Conjugate Gradient solver, shows that the Cheetah reductions speeds up total time to solution by 195%, demonstrating the potential benefits for scientific simulations.

Gorentla Venkata, Manjunath↗

MapReduceXMT v. Beta 0.1

The MapReduceXMT library ports the MapReduce framework onto the Cray XMT. MapReduce is a programming paradigm and an approach to data management for unstructured problems. It has gained relevance due to its ability to map serial operations onto parallel distributed architectures, significantly improving developer/analyst productivity. The MapReduceXMT implements the key aspects of MapReduce for the Cray XMT, a massively threaded system that is inherently difficult to program. MapReduceXMT allows users to utilize the machine effectively and efficiently without extensive training in multi-threaded programming. The MapReduceXMT library ports the MapReduce framework onto the Cray XMT. MapReduce is a programming paradigm and an approach to data management for unstructured problems. It has gained relevance due to its ability to map serial operations onto parallel distributed architectures, significantly improving developer/analyst productivity. The MapReduceXMT implements the key aspects of MapReduce for the Cray XMT, a massively threaded system that is inherently difficult to program. MapReduceXMT allows users to utilize the machine effectively and efficiently without extensive training in multi-threaded programming.

Goodman, Eric↗

OpenSHMEM Implementation of HPCG Benchmark

We describe the effort to implement the HPCG benchmark using OpenSHMEM and MPI one-sided communication. Unlike the High Performance LINPACK (HPL) benchmark that places em- phasis on large dense matrix computations, the HPCG benchmark is dominated by sparse operations such as sparse matrix-vector product, sparse matrix triangular solve, and long vector operations. The MPI one-sided implementation is developed using the one-sided OpenSHMEM implementation. Pre- liminary results comparing the original MPI, OpenSHMEM, and MPI one-sided implementations on an SGI cluster, Cray XK7 and Cray XC30 are presented. The results suggest the MPI, OpenSHMEM, and MPI one-sided implementations all obtain similar overall performance but the MPI one-sided im- plementation seems to slightly increase the run time for multigrid preconditioning in HPCG on the Cray XK7 and Cray XC30.

OpenSHMEM HPCG↗

A Novel Application of Parallel Betweenness Centrality to Power Grid Contingency Analysis

In Energy Management Systems, contingency analysis is commonly performed for identifying and mitigating potentially harmful power grid component failures. The exponentially increasing combinatorial number of failure modes imposes a significant computational burden for massive contingency analysis. It is critical to select a limited set of high-impact contingency cases within the constraint of computing power and time requirements to make it possible for real-time power system vulnerability assessment. In this paper, we present a novel application of parallel betweenness centrality to power grid contingency selection. We cross-validate the proposed method using the model and data of the western US power grid, and implement it on a Cray XMT system - a massively multithreaded architecture - leveraging its advantages for parallel execution of irregular algorithms, such as graph analysis. We achieve a speedup of 55 times (on 64 processors) compared against the single-processor version of the same code running on the Cray XMT. We also compare an OpenMP-based version of the same code running on an HP Superdome shared-memory machine. The performance of the Cray XMT code shows better scalability and resource utilization, and shorter execution time for large-scale power grids. This proposed approach has been evaluated in PNNL’s Electricity Infrastructure Operations Center (EIOC). It is expected to provide a quick and efficient solution to massive contingency selection problems to help power grid operators to identify and mitigate potential widespread cascading power grid failures in real time.

24 POWER TRANSMISSION AND DISTRIBUTION↗