Prof Dr, A. Bode, J. Krammer et al. · 2008
A main research area of the Lehrstuhl für Rechnertechnik and Rechnerorganisation at Technische Universität München (LRR-TUM) are tools supporting the development of parallel programs. Within a couple of funded projects (German Science Foundation: SFB 342, Esprit: PREPARE and HAMLET) tools supporting parallel I/O, load balancing, resource management and checkpointing, performance analysis, debugging, deterministic execution, and visualization are being developed. The debugging tool DETOP, the performance analysis tool PATOP, and the underlying monitoring system have recently been ported to the PowerXplorer and GC/PowerPlus parallel computers. This work has been carried out in cooperation with Parsytec, which distributes both tools since Dec. 1994. While LRR-TUM has ported and improved the debugger and the performance analyzer, Parsytec has implemented the necessary support in the operating system. A modified and enhanced PTrace interface is now provided by the nanokernel which is the basis of the PARIX library on the PowerPC based computers. The Parallel Debugger DETOP The parallel debugger DETOP allows to examine the run-time behavior of both data-parallel
Cited by 0
Wylie, B.J.N. · JuSER (Forschungszentrum Jülich) · 2009
Cited by 0
Keonwook Kim, Alan D. George, Priyabrata Sinha · 1999
Advancements in beamforming algorithms are exceeding the computation and communication capabilities of traditional sonar array systems. Implementing parallel beamforming algorithms in situ on distributed array systems holds the potential to provide increased performance and fault tolerance at a lower cost. This paper compares three parallel algorithms for distributed arrays in terms of execution throughput, result latency, scaled speedup, and parallel effi
Cited by 1
Wylie, B.J.N. · JuSER (Forschungszentrum Jülich) · 2011
Cited by 0
TAY SENG CHUAN · National University of Singapore · 1999
Ph.D
Cited by 0
Vikram Adve, Charles Koelbel, John M. Mellor-Crummey · 1994
Effective strategies for performance analysis and tuning will be essential for the success of data parallel languages such as High-Performance Fortran (HPF) and Fortran D. Since compilers for these languages insert all communication, they have considerable knowledge about a program's dynamic structure and the relationship between its parallelism and communication. This paper explores how this compiler knowledge can be exploited to support performance evaluation and tuning. First, the compiler itself can use parameterized models to tune the performance of individual program phases; this approach can be effective provided that the compiler can test and handle violations of the model assumptions. Second, by exploiting compiler knowledge and introducing code transformations to improve monitorability, we can collect dynamic performance information that is far more compact than full communication traces, but well suited to the needs of tuning specific communication patterns. Third, we discus...
Cited by 0
Athanasios K Grivas · 2016
Cited by 0
Manfred Jobmann, Johann Schumann · ACM SIGMETRICS Performance Evaluation Review · 1992
No abstract available.
Cited by 0
William J. Knottenbelt · 2000
Stochastic performance models provide a formal way of capturing and analysing the complex dynamic behaviour of concurrent systems. Such models can be spec-i ed by several high-level formalisms, including Stochastic Petri nets, Queueing networks and Stochastic Process Algebras. Traditionally, performance statistics for these models are derived by generating and then solving a Markov chain corresponding to the model's behaviour at the state transition level. However, workstation memory and compute power are often overwhelmed by the sheer number of states in the Markov chain and the size of their internal computer representations. This thesis presents two parallel and distributed techniques which signicantly increase the size of the models that can be analysed using Markov modelling. The techniques attack the space and time requirements of both major phases of the analysis, i.e. construction of the Markov chain from a high-level model (state space generation) and solution of the Markov chain to determine its equilibrium
Cited by 0
Allsopp, N., Bungartz, H.-J., Gerndt, M. et al. · JuSER (Forschungszentrum Jülich) · 2010
Cited by 0
Tambouris, E. · OpenGrey (Institut de l'Information Scientifique et Technique) · 2001
Cited by 0
Wagner Meira, Thomas J. Leblanc · 1997
This paper describes a general framework and several specific techniques for cause-effect analysis: an automated inference process that presents explanations for dynamic phenomena of parallel program executions in terms of underlying causes and the related source code. We illustrate the framework by describing the implementation of three analysis techniques: waiting time analysis identifies the cause of synchronization overhead as the differences in execution paths taken by synchronizing processors; protocol analysis identifies the sharing patterns to pages that produce invalidations in a DSM protocol; and transaction analysis identifies conflicts between transactions in a parallel file system that cause aborts. We present examples of how each technique can be used to understand observed performance effects, and how insights derived from these techniques suggest program modifications that lead to improved performance. 1 Introduction Many performance phenomena arise during execution ...
Cited by 0
Wellington S. Martins · 1996
In the last few years, there has been considerable interest in general purpose computational models of parallel computation to permit independent development of hardware and software. The BSP and related models represent an important step in this direction. This paper presents a methodology for the performance analysis of bulk synchronous parallel algorithms based on parameters which reflect the two-level memory hierarchy advocated by these models. A parallel sorting algorithm is taken as a case study where it is shown a close agreement between theoretical and experimental results.
Cited by 0Open access
Qinchuan Li, Yang Chao, Lingmin Xu et al. · Research on intelligent manufacturing · 2023
Cited by 18Open access
2003
Abstract The significant gap between peak and realized performance of parallel systems motivates the need for performance analysis. In order to predict the performance of a class of parallel multilevel ILU preconditioner (PBILUM), we build two performance prediction models for both the preconditioner construction phase and the solution phase. These models combine theoretical features of the preconditioners with estimates on computation cost, communications overhead, etc. Experimental simulations show that our model predication based on certain reasonable assumptions is close to the simulation results. The models may be used to predict the performance of this class of parallel preconditioners. 1 Introduction Parallel processing is making a tremendous impact on many areas of computer applications. Among the various applications, parallel processing and optimization algorithms for solving large sparse linear systems have drawn a considerable amount of attention from researchers in high performance scientific computing area. With the high raw computing power of parallel computers, it is now possible to address many large scale computation intensive applications that were until recently beyond the capability of conventional computing techniques [6].
Cited by 0
Junko Nakajima, Mitsuru Matsui · 2003
Abstract. This paper shows an extensive software performance analysis of dedicated hash functions, particularly concentrating on Pentium III, which is a current dominant processor. The targeted hash functions are MD5, RIPEMD-128-160, SHA-1-256-512 and Whirlpool, which fully cover currently used and future promising hashing algorithms. We try to optimize hashing speed not only by carefully arranging pipeline scheduling but also by processing two or even three message blocks in parallel using MMX registers for 32-bit oriented hash functions. Moreover we thoroughly utilize 64-bit MMX instructions for maximizing performance of 64-bit oriented hash functions, SHA-512 and Whirlpool. To our best knowledge, this paper gives the first detailed measured performance analysis
Cited by 0
Jidong Zhai, Yuyang Jin, Wenguang Chen et al. · 2023
Offers an approach for performance analysis of parallel applications on HPC systems, including communication, memory, noise and scalability analysis.
Cited by 0Open access
Boyao Zhang, Jue Wang, Yangde Feng et al. · AIP conference proceedings · 2018
Molecular dynamics simulation is an important tool for studying materials microstructure evolution under radiation effects. It is difficult to simulate atomic diffusion in materials owing to the time-scale limitations of the molecular dynamics method. Accelerated molecular dynamics has been developed as a solution and the parallel replica method is the simplest and most accurate of the accelerated dynamics techniques. The simulation time scale can reach several orders more than the direct molecular dynamics, while retaining the complete atomic detail. In parallel replica method, the entire system is replicated on each of M available parallel or distributed processors, it is in line with the characteristics of high-performance parallel computing and can be optimized from the program level.
Cited by 0
Correrini, F., Nazzarelli, M.R., Setti, L et al. · OpenGrey (Institut de l'Information Scientifique et Technique) · 1994
Consiglio Nazionale delle Ricerche (CNR). Biblioteca Centrale / CNR - Consiglio Nazionale delle Richerche
Cited by 0
Wylie, B.J.N. · JuSER (Forschungszentrum Jülich) · 2011
Cited by 0