Hazem M. Abbas · IEEE Transactions on Neural Networks · 2004
The process of training neural networks on parallel architectures has been used to assess the performance of so many parallel machines. In this paper, we are investigating the implementation of backpropagation (BP) on the Alex AVX-2 coarse-grained MIMD machine. A host-worker parallel implementation is carried out in order to train different networks to learn the NetTalk dictionary. First, a computational model is constructed using a single processor to complete the learning process. Also, a communication model for the host-worker topology is developed in order to compute the communication overhead in the broadcasting/gathering process. Both models are then used to predict the machine performance when p processors are used and a comparison with the actual measured performance of the parallel architecture implementation is carried out. Simulation results show that both models can be used effectively to predict the machine performance for the NetTalk problem. Finally, a comparison between the AVX-2 NetTalk implementation and the performance of other parallel platforms is presented.
Cited by 1
Raja Jamilah Raja Yusof, Roziati Zainuddin, Mohd Sapiyan Baba et al. · Malaysian Journal of Computer Science · 2012
Parallel plot is a relatively new type of plot in comparison to scatter plot. The aim of this paper is to find out the effectiveness of parallel plot visualization in comparison to the scatter plot for novice users. Comparison through literature shows that parallel and scatter plot visualization techniques offer similar visualization features. A study was conducted based on the parallel plot in Parallel Coordinate for the Qur’an (PaCQ) interface using a relational model, Goals Operators Methods and Selection Rules (GOMS) analysis and user performance evaluation. Based on the GOMS analysis, the subtasks and operation time involved in exploring data in scatter plot exceeds the subtasks and operation time in parallel plot in relation to the PaCQ interface developed for novice user even though, earlier studies found that scatter plot is easier to use. The GOMS analysis shows that the average time taken to complete tasks for the parallel plot is shorter then the scatter plot. A user performance evaluation shows no significant difference in the scores of interpreting the parallel plot within three levels of educational level of novice users. These findings provide evidences that the parallel plot is more effective for PaCQ designed for novice users.
Cited by 2Open access
Raja Jamilah, Raja Yusof, Roziati Zainuddin et al. · 2012
Parallel plot is a relatively new type of plot in comparison to scatter plot. The aim of this paper is to find out the effectiveness of parallel plot visualization in comparison to the scatter plot for novice users. Comparison through literature shows that parallel and scatter plot visualization techniques offer similar visualization features. A study was conducted based on the parallel plot in Parallel Coordinate for the Qur’an (PaCQ) interface using a relational model, Goals Operators Methods and Selection Rules (GOMS) analysis and user performance evaluation. Based on the GOMS analysis, the subtasks and operation time involved in exploring data in scatter plot exceeds the subtasks and operation time in parallel plot in relation to the PaCQ interface developed for novice user even though, earlier studies found that scatter plot is easier to use. The GOMS analysis shows that the average time taken to complete tasks for the parallel plot is shorter then the scatter plot. A user performance evaluation shows no significant difference in the scores of interpreting the parallel plot within three levels of educational level of novice users. These findings provide evidences that the parallel plot is more effective for PaCQ designed for novice users.
Cited by 0
Ruben Laso, Diego Krupitza, Sascha Hunold · 2024
Since the advent of parallel algorithms in the C++17 Standard Template Library (STL), the STL has become a viable framework for creating performance-portable applications. Given multiple existing implementations of the parallel algorithms, a systematic, quantitative performance comparison is essential for choosing the appropriate implementation for a particular hardware configuration.
Cited by 1Open access
Yingying Li, Hao Liu, Peijian Chen et al. · Wear · 2024
Cited by 15
Mark Geldart, Phil Webb, Hans C. E. Larsson et al. · International Journal of Machine Tools and Manufacture · 2003
Cited by 69
M. Kh. Aswad, P. W. Trinder, A. D. AlZain et al. · 2012
Multicores are becoming the dominant processor technology and functional languages are theoretically well suited to exploit them. In practice, however, implementing effective high level parallel functional languages is extremely challenging. This paper is the first programming and performance comparison of functional multicore technologies and reports some of the first ever multicore results for two languages. As such it reflects the growing maturity of the field by systematically evaluating four parallel Haskell implementations on a common multicore architecture. The comparison contrasts the programming effort each language requires with the parallel performance delivered. The study uses 15 ’typical ’ programs to compare a ’no pain’, i.e. entirely implicit, parallel language with three ’low pain’, i.e. semi-explicit languages. The parallel Haskell implementations use different versions of GHC compiler technology, and hence the comparative performance metric is speedup which normalises against sequential performance. We ground the speedup comparisons by reporting both sequential and parallel runtimes and efficiencies for three of the languages. Our experiments focus on the number of programs improved, the absolute speedups delivered, the parallel scalability, and the program changes required to coordinate parallelism. The results are encouraging and, on occasion, surprising.
Cited by 0
M. KH. Aswad, et al. · 2013
Multicore and NUMA architectures are becoming the dominant processor technology and functional languages are theoretically well suited to exploit them. In practice, however, implementing effective high level parallel functional languages is extremely challenging. This paper is a systematic programming and performance comparison of four parallel Haskell implementations on a common multicore architecture. It provides a detailed analysis of the performance, and contrasts the programming effort that each language requires with the parallel performance delivered. The study uses 15 ’typical’ programs to compare a ‘no pain’, i.e. entirely implicit, parallel implementation with three ‘low pain’, i.e. semi-explicit, language implementations. We report detailed studies comparing the parallel performance delivered. The comparative performance metric is speedup which normalises against sequential performance. We ground the speedup comparisons by reporting both sequential and parallel runtimes and efficiencies for three of the languages. To measure the programming effort
Cited by 0
Ying, Yang · Open Science Framework · 2025
Self-deception is a positive self-perception bias profoundly influenced by social context. Two experiments respectively investigated self-deception within two distinct social comparison contexts: one involving comparison with “average others” (aggregate performance) and the other with “specific others” (a single individual). In Experiment 1, participants were falsely informed that their ability scores were higher (downward comparison) or lower (upward comparison) than the “average others” score. Self-deception was then assessed by asking them to predict the number of questions they would answer correctly on a subsequent test of the same type, with more inflated predictions indicating greater self-deception. Results showed that prediction scores were highest following downward comparison. Furthermore, after controlling for actual ability, the discrepancy between predicted and actual scores was significantly greater in the downward and no-comparison groups than in the upward and parallel comparison groups. In Experiment 2, participants were falsely informed that their performance-based monetary rewards were higher or lower than those of a “specific other”. Prediction scores following downward and parallel comparisons were significantly higher than after upward comparisons. However, on the prediction–actual discrepancy index,, the discrepancy was greatest after downward comparison, intermediate after parallel comparison, and smallest after upward comparison. For the number of correct predictions, downward comparison also elicited the highest scores, followed by parallel and then upward comparison. In summary, downward comparison induces the highest self-deception, regardless of comparison context. Moreover, the effects were more pronounced in comparisons with a “specific other” than with “average others,” suggesting that concrete, vivid social information more potently activates self-enhancement motivation.
Cited by 0Open access
Ying, Yang · Open Science Framework · 2025
Self-deception is a positive self-perception bias profoundly influenced by social context. Two experiments respectively investigated self-deception within two distinct social comparison contexts: one involving comparison with “average others” (aggregate performance) and the other with “specific others” (a single individual). In Experiment 1, participants were falsely informed that their ability scores were higher (downward comparison) or lower (upward comparison) than the “average others” score. Self-deception was then assessed by asking them to predict the number of questions they would answer correctly on a subsequent test of the same type, with more inflated predictions indicating greater self-deception. Results showed that prediction scores were highest following downward comparison. Furthermore, after controlling for actual ability, the discrepancy between predicted and actual scores was significantly greater in the downward and no-comparison groups than in the upward and parallel comparison groups. In Experiment 2, participants were falsely informed that their performance-based monetary rewards were higher or lower than those of a “specific other”. Prediction scores following downward and parallel comparisons were significantly higher than after upward comparisons. However, on the prediction–actual discrepancy index,, the discrepancy was greatest after downward comparison, intermediate after parallel comparison, and smallest after upward comparison. For the number of correct predictions, downward comparison also elicited the highest scores, followed by parallel and then upward comparison. In summary, downward comparison induces the highest self-deception, regardless of comparison context. Moreover, the effects were more pronounced in comparisons with a “specific other” than with “average others,” suggesting that concrete, vivid social information more potently activates self-enhancement motivation.
Cited by 0Open access
Richard J. Priem, M. Hersch · University of North Texas Digital Library (University of North Texas) · 1956
A study of the effect of fuel-orifice diameter on the performance of heptane-oxygen rocket engines with single-element injectors. Five engines with a range of thrust were tested with parallel-sheet, triplet, and parallel-jet injectors. Decreasing the fuel-orifice diameter was found to increase the performance for a fixed chamber length. Results regarding injector performance and a comparison of experimental and analytical results are provided.
Cited by 0Open access
R.E. Beighley, C.A. Gougler, James R. Johnson · 2003
Parallel active harmonic filters are just emerging as a viable technology. This paper discusses: (1) what a parallel active harmonic filter is; (2) a comparison of current product design/application emphases; (3) application comparisons to passive harmonic filters; and (4) performance of an active filter.
Cited by 4
Haluk Rahmi Topcuoglu, Alok Choudhary, Haluk Topcuoglu et al. · 1995
Parallel I/0 subsystems are added to massively parallel computers in order to lessen I/0 bottleneck to some extent. Up to now, a few number of parallel software systems have been designed and implemented to assist programmers in 1/0 intensive applications; PASSION is one of them. By providing parallel I/0 support at the language, compiler and run-time level, PASSION system explores the large design space of parallel systems. The target of this paper is to show the performance benefits of using PASSION I/0 libraries at runtime in comparison with using conventional parallel 1/0 primitives for high performance parallel 1/0 in LU factorization kernel, a very widely used scientific kernel. 1 1
Cited by 0
Haluk Rahmi Topcuoglu, Alok Choudhary · Syracuse University Libraries (Syracuse University) · 1995
Parallel I/0 subsystems are added to massively parallel computers in order to lessen I/0 bottleneck to some extent. Up to now, a few number of parallel software systems have been designed and implemented to assist programmers in I/0 intensive applications; PASSION is one of them. By providing parallel I/0 support at the language, compiler and run-time level, PASSION system explores the large design space of parallel systems. The target of this paper is to show the performance benefits of using PASSION I/0 libraries at runtime in comparison with using conventional parallel I/0 primitives for high performance parallel I/0 in LU factorization kernel, a very widely used scientific kernel.
Cited by 0Open access
Tom Deakin, James Price, Matt Martineau et al. · Bristol Research (University of Bristol)
We present a major update to the GPU-STREAM benchmark, first shown at SC'15.The original benchmark allowed comparison of achievable memory bandwidth performance through the STREAM kernels on OpenCL devices.GPU-STREAM v2.0 extends the benchmark to another dimension: the kernels are implemented in a wide range of popular stateof-the-art parallel programming models.This allows an intuitive comparison of performance across a diverse set of programming models and devices, investigating whether choice of model matters to performance and performance portability.In particular we investigate 7 parallel programming languages (OpenMP 4.x, OpenACC, Kokkos, RAJA, SYCL, CUDA and OpenCL) across 12 devices (6 GPUs from NVIDIA and AMD, Intel Xeon Phi (Knights Landing), 4 generations of Intel Xeon CPUs, and IBM Power 8).
Cited by 0Open access
Ke Xu, Xin Guo, Xinjing Zhang et al. · Journal of Computer Assisted Learning · 2025
ABSTRACT Background Previous studies have indicated that encouraging learners to teach to their peers has been proven effective in second language learning. However, to date, no research has focused on how learners teaching to peers of varying proficiency levels impacts video learning performance and the cognitive neuroscience mechanisms of learners. Objectives The study investigated the effect of social comparison (upward comparison vs. parallel comparison vs. downward comparison) in video learning using learning by teaching. Methods The study employed a within‐subjects design, wherein participants taught English vocabulary to peers of three different ability levels in a second language setting. Students ( N = 36 undergraduate and graduate students) were assigned to watch three videos in balanced order, then generate and speak out sentences to the peer. Results and Conclusions The results of Repeated measures ANOVAs showed that learners in the upward and parallel comparison conditions showed higher learning performance and reported higher pressure than in the downward comparison condition. The results of EEGs showed that compared with downward comparison, upward and parallel comparison not only reduced the learners' theta oscillations in the left temporoparietal, occipital, or front‐central but also manifested in the lower ratio of theta/alpha at the left temporal–parietal region, suggesting learners reduced mental workload. The findings have practical implications: teaching to peers with different levels can cause social comparison, and in video learning, learners should be encouraged to teach peers with higher‐level or the same‐level rather than the lower‐level in order to improve their learning performance.
Cited by 0
Hans-wolfgang Loidl · 1997
. This paper discusses the parallelisation and performance tuning of a typical computer algebra algorithm, LinSolv, using evaluation strategies. We present three steps in the parallelisation process starting with a naive parallel version. As this algorithm uses infinite data structures as intermediate values it is necessary to define very sophisticated strategies in order to improve parallel performance. We also compare the strategic parallel code with pre-strategy code. This comparison shows how evaluation strategies help to localise changes needed for parallelisation. In particular, the separation between algorithmic and parallel code makes the structure of the parallelism much clearer. 1 Introduction Tuning the performance of a parallel algorithm can be a long, tiresome process. A parallel programming model should aid the programmer especially in this stage, allowing him to experiment with different patterns of parallel behaviour. Based on our experiences with developing p...
Cited by 0
Alexander Reinefeld, Volker Schnecke · 1994
Although PVM was developed to use a network of heterogeneous UNIX computers as a single large parallel computer, it has become an interface for portable programming even on MPP's. We present PVM performance results for a massively parallel transputer system with up to 512 processors. In comparison to an implementation of the same application in the native transputer operating system Parix, we realized about 30% performance loss for the PVM application. This is mainly caused by restrictions in the process model and the less efficient communication of PVM in comparison to Parix. 1 Introduction Analogous to the shift from assembler language programming to the third-generation languages in the early years of computer science, we are currently witnessing a paradigm change towards the use of portable programming models in parallel high-performance computing. Like before, the high-level programming environment is paid for by a lower system performance. But how much does portability cost in p...
Cited by 0
Hong-Jun Lee, Kwang‐Kyo Oh · 2017
Parallel operation of multiple motor drives has been recently used in a variety of applications. In this paper, we present a comparison study result on the existing parallel operation schemes We first briefly discuss the configuration of each parallel operation scheme. Then we provide simulation results of the operation schemes and discuss the performance characteristics. Advantages and disadvantages of schemes are also discussed.
Cited by 3
Hongwei Bai (1447999), Weiyan Tong (22032834), Zhenkun Geng (22032837) et al. · Figshare · 2025
The matrices show classification performance for four models: (a) Improved Parallel 1D-CNN, (b) Single-Channel Large-Kernel CNN, (c) Parallel 1D-CNN, and (d) Single-Channel Small-Kernel CNN. Each matrix presents the counts of correct and incorrect predictions across classes, enabling direct comparison of model accuracy.
Cited by 0Open access