Vitor Ramos, Carlos Valderrama, Samuel Xavier‐de‐Souza et al. · 2019
The analysis of application performance is essential to better exploit its potential on High-Performance Computing (HPC) architectures. Access to performance counters, available in modern processors, allows collecting key information about program behavior to provide the most appropriate HPC execution strategy. In this context, we have developed an accurate tool based on performance counters, which facilitates modeling, fingerprinting, behavior comparison and clustering of applications. It provides a high-level Python API for accessing and configuring performance counters. While the execution and counters gathering is performed by a C++ module to reduce overheads. Moreover, the accuracy of this multiplatform tool was also compared to existing alternatives. Key features, such as performance counters collection, post-processing, and comparison, enable fingerprinting of applications, an important step in understanding program behavior for later classification and optimization according to the parameters characterizing the target HPC platform. For demonstration purposes, the tool was used in the clustering of Polybench applications, a frequently used benchmark set for kernels monitoring. This clustering helped the identification of applications with similar and comparable behaviors in terms of input size, data accesses and movements, resource utilization, and computation, which facilitates the creation of test sets for a given environment, according to specific measurement parameters.
Cited by 8
A. Dinesh Babu, Hanna Mathew · 2025
The area of digital system design is evolving, hence the significance of high speed arithmetic operation is also increasing. The applications like cryptography, signal processing, high-performance computing require faster and efficient adder architectures. Since the existing adders are exhibiting delay mostly in the case of carry propagation, Parallel Prefix Adders can be used for reduced delays even in cases involving large number of bits arithmetic operations. The Approximate Parallel Prefix Adder technique utilizes the approximation method for more efficient architectures. In this work, five basic topologies of approximate parallel prefix adders are compared for the performance enhancement in three operand binary addition. The bit sizes 16 bits, 32 bits, 64 bits of the adder topologies are implemented to obtain the comparative analysis related to different performance parameters. It is simulated and implemented on Vivado Design Suite and is prototyped on Nexys 4 Artix- 7 FPGA Board for the on-chip functional verification.
Cited by 0
Guido Freckmann, Stefan Pleus, Manuela Link et al. · Journal of Diabetes Science and Technology · 2013
In this study, two sensors of each of the three CGM systems were compared in a setting following POCT05-A recommendations. The Navigator CGM system achieved more accurate results than the Guardian or the Seven Plus with respect to MARD and PARD. Performance in the hypoglycemic range was markedly worse for all CGM systems when compared with BG results.
Cited by 73Open access
Vitor Ramos Gomes Da Silva, Carlos Valderrama, Samuel Xavier De Souza et al. · ORBi UMONS · 2019
The analysis of application performance is essential to better exploit its potential on High-Performance Computing (HPC) architectures.Access to performance counters, available in modern processors, allows collecting key information about program behavior to provide the most appropriate HPC execution strategy.In this context, we have developed an accurate tool based on performance counters, which facilitates modeling, fingerprinting, behavior comparison and clustering of applications.It provides a high-level Python API for accessing and configuring performance counters; while execution and counters data gathering is performed by a C++ module to reduce overhead.Indeed, the accuracy of this multiplatform tool was also compared to existing alternatives.Key features, such as performance counters collection, post-processing, and comparison, enable fingerprinting of applications, an important step in understanding program behavior for later classification and optimization according to the parameters characterizing the target HPC platform.For demonstration purposes, the tool was used in the clustering of Polybench applications, a frequently used benchmark set for kernels monitoring.This clustering facilitated the identification of applications with similar and comparable behaviors in terms of input size, data access and transfer, resource utilization and computation, which facilitates the creation of test sets for a given environment based on specific measurement parameters.
Cited by 0Open access
Norma Alias, Hafizah Farhah Saipan Saipol, Asnida Che Abd. Ghani et al. · Advanced Science Letters · 2013
This paper presents the parallel implementation and performance analysis of a large sparse matrix of one dimensional parabolic Partial Differential Equation (PDE) between Parallel Virtual Machine (PVM) and Matlab Distributed Computing (MDC) platform. PVM is a software to perform parallel programming using heterogeneous or homogeneous clusters of parallel computing system connected by a logical network system to users as virtual memory parallel computer system. MDC is a scientific library programming language. Alternating Group Explicit (AGE) method is used to solve the parabolic equation with large sparse matrix. The distributed parallel algorithm of AGE method is implemented on the PVM with RedHat Linux operation and MDC Version 2011a. The parallel performance evaluation (PPE) is compared between the two platforms in terms of time execution, speedup, efficiency, effectiveness, and temporal performance. From the performance analysis, we can conclude that PVM is better compared to MDC.
Cited by 10
E. S. G. Hettiaratchi · Medical Education · 1978
Two parallel tests in physiology, one in multiple choice answer form and one in short answer form, were administered consecutively to a group of 104 students. The construction of the questions in the two papers and marking schedules were such that if the form of examination did not influence the students' performance, the scores obtained in the two papers could be expected to be closely similar. The results indicate that the form of the question influences the students' performance significantly. The short answer format offers certain advantages over the multiple choice format and it would appear that greater use should be made of this type of question in testing at medical schools.
Cited by 10
Agostino Ponzetti, R. Laface, E. Milanesi et al. · Annals of Oncology · 2020
Cited by 0Open access
Simplice Donfack, Jack Dongarra, Mathieu Faverge et al. · HAL (Le Centre pour la Communication Scientifique Directe) · 2013
Gaussian elimination is a canonical linear algebra procedure for solving linear systems of equations. In the last few years, the algorithm received a lot of attention in an attempt to improve its parallel performance. This article surveys recent developments in parallel implementations of the Gaussian elimination. Five different flavors are investigated. Three of them are based on different strategies for pivoting: partial pivoting, incremental pivoting, and tournament pivoting. The fourth one replaces pivoting with the Random Butterfly Transformation, and finally, an implementation without pivoting is used as a performance baseline. The technique of iterative refinement is applied to recover numerical accuracy when necessary. All parallel implementations are produced using dynamic, superscalar, runtime scheduling and tile matrix layout. Results on two multi-socket multicore systems are presented. Performance and numerical accuracy is analyzed.
Cited by 8Open access
Hettiaratchi, E. S. G. · University of Nairobi Research Archive (University of Nairobi) · 2014
Two parallel tests in physiology, one in multiple choice answer form and one in short answer form, were administered consecutively to a group of 104 students. The construction of the questions in the two papers and marking schedules were such that if the form of examination did not influence the students' performance, the scores obtained in the two papers could be expected to be closely similar. The results indicate that the form of the question influences the students' performance significantly. The short answer format offers certain advantages over the multiple choice format and it would appear that greater use should be made of this type of question in testing at medical schools.
Cited by 0
Bin Zhu, Liping Wang, Jun Wu et al. · International Journal of Robotics and Automation · 2022
The reliability study of machine tools draws more and more attentionboth in academia and industry, and the loading device is very impor-tant for the reliability test as it can provide force and torque to simu-late the cutting force.
Cited by 0
Carlos Garre, Domenico Mundo, Marco Gubitosa et al. · SAE technical papers on CD-ROM/SAE technical paper series · 2014
<div class="section abstract"><div class="htmlview paragraph">Real-time simulation is a valuable tool in the design and test of vehicles and vehicle parts, mainly when interfacing with hardware modules working at a given rate, as in hardware-in-the-loop testing. Real-time operating-systems (RTOS) are designed for minimizing the latency of critical operations such as interrupt dispatch, task switch or inter-process communication (IPC). General-purpose operating-systems (GPOS), instead, are designed for maximizing throughput in heavy-load systems. In complex simulations where the amount of work to do in one step is high, achieving real-time depends not only in the latency of the event starting the step, but also on the capacity of the system for computing one step in the available time. While it is demonstrated that RTOS present lower latencies than GPOS, the choice is not clear when maximizing throughput is also critical.</div><div class="htmlview paragraph">In this paper, the performance of RTOS and GPOS running complex real-time simulations is compared, focusing on the computation of large simulation steps. GNU/Linux has been chosen as GPOS. A RTOS is chosen with a micro-benchmark comparing the major choices of Linux-based RTOS. Once chosen the systems, the simulation of a tire model is used as application case for benchmarking, comparing within 52470 different configurations (with different number of elements and threads). The benchmark measures which configurations miss a single deadline, and demonstrates that even in simulations with a high number of elements and large time steps, RTOS are a better choice due mainly to the latency associated to IPC and task switch when the simulation is parallelized.</div></div>
Cited by 8
Stuti Moitra · 2008
Various tridiagonal solvers have been proposed in recent years for di�erent par� allel platforms. In this paper � the performance of three tridiagonal solvers � namely� the parallel partition LU algorithm � the parallel diagonal dominant algorithm � and the reduced diagonal dominant algorithm � is studied. These algorithms are designed for distributed�memory machines and are tested on an Intel Paragon and an IBM SP2 machines. Measured results are reported in terms of execution time and speedup. An� alytical study are conducted for di�erent communication topologies and for di�erent tridiagonal systems. The measured results match the analytical results closely. In ad� dition to address implementation issues � performance considerations such as problem sizes and models of speedup are also discussed.
Cited by 0
Shambhavi Balamuthu Sampath, B. R. Nanjesh, Bharat Bhushan Sagar et al. · 2015
Parallel computing operates on the principle that large problems can often be divided into smaller ones, which are then solved concurrently to save time by taking advantage of non-local resources and overcoming memory constraints. Multiplication of larger matrices requires a lot of computation time. This paper deals with the two methods for handling Parallel Matrix Multiplication. First is, dividing the rows of one of the input matrices into set of rows based on the number of slaves and assigning one rows set for each slave for computation. Second method is, assigning one row of one of the input matrices at a time for each slave starting from first row to first slave and second row to second slave and so on and loop backs to the first slave when last slave assignment is finished and repeated until all rows are finished assigning. These two methods are implemented using Parallel Virtual Machine and the computation is performed for different sizes of matrices over the different number of nodes. The results show that the row per slave
Cited by 0
Hasan Ozkaya, Osman S. Senturk, Ahmet M. Hava · 2007
In this work discrete time current regulation methods for three-phase parallel active filters (PAF) are investigated. With emphasis on PAF applications, linear and hysteresis current regulation methods are studied and improvements on the discrete time implementation of the latter type without any additional analog hardware are provided. The dependency of the switching ripple filter (SRF) type on the PAF current regulator (CR) type is considered and design methods provided. A PAF is designed for a three-phase 10 kW diode rectifier load and studies are investigated through this system. The CR and PAF performances are studied by means of theory, simulations, and laboratory experiments. A DSP controlled PAF system is built and all the algorithms are experimentally verified. The study shows that the resonant filter linear current regulator and the improved discrete time hysteresis current regulator with appropriate SRF types perform satisfactorily.
Cited by 3
Rin MIKAMI, Toshinari TAKESHITA, Satoru Ikeda et al. · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2025
SAP-net, based on an activation-spreading model in cognitive psychology, selects knowledge for reinforcement learning and has successfully avoided dynamic obstacles. However, its parallel computation model (P-SAP) is difficult to analyze because of parallel computation. To address this, we compared the activity values of P-SAP and the sequential computation model (S-SAP). In this paper, P-SAP and S-SAP produce similar outputs when activation order and weighted adjacency matrices are identical. This suggests that P-SAP’s behavior can be approximated using S-SAP, improving P-SAP’s reliability.
Cited by 0
Xing Guo, Jiaji Wu, Zhensen Wu et al. · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · 2016
The infrared (IR) signature of an aerial target due to the reflection of radiation from the Sun, the Earth's surface and atmosphere plays an important role in aerial target detection and tracking. As the background radiation from the Earth's surface, and atmosphere is distributed in the entire space and in a wide spectrum, it is time-consuming to obtain an aerial target's reflected radiation. This problem is suitable for parallel implementation to run on multicore CPU or many-core GPU because the reflection of background radiation incident from different directions in each spectral wavelength can be calculated in parallel. We consider three different parallel approaches: 1) CPU implementation using OpenMP (open multiprocessing); 2) GPU implementation using OpenACC (open accelerators); and 3) GPU implementation using CUDA (compute unified device architecture). An NVIDIA K20c GPU (with 2496 cores) and two Intel Xeon E5-2690 CPU (with 8 cores each) are used in our experiment. Compared to their single-threaded CPU counterpart, speedups obtained by OpenMP, OpenACC, and CUDA implementations are 15x, 140x, 426x, respectively. The result shows that GPU implementations are promising in our problem.
Cited by 27
M ghanbarzadeh, H Nasiri · SHILAP Revista de lepidopterología · 2017
Cited by 0Open access
Sarwo Edhy Sofyan, Teuku Meurah Indra Riayatsyah, Khairil KHAİRİL et al. · Results in Engineering · 2024
This study uses Ansys Fluent to compare the thermal performance of series and parallel earth air heat exchanger (EAHE) systems for cooling. The model was validated using experimental data from published literature and matched simulation results. The sensitivity study examined how length, diameter, and ground surface coverings affected EAHE performance. The relationship between effectiveness and the Number Transfer Unit (NTU) of EAHE was explored along with the soil thermal regime. The simulation results show that the series EAHE can achieve a lower temperature drop than parallel. With an input air temperature of 32 °C and EAHE lengths varying from 10 m to 50 m (in 10 m increments), the 50 m EAHE produced the lowest outlet air temperatures: 27.2 °C for the series configuration and 28.8 °C for the parallel arrangement. Changing the EAHE diameter (4–6 in) results in a ±0.2 °C outlet air temperature drop for both setups. EAHE performed best under short grass soil cover, yielding 1.4 °C and 0.5 °C lower outlet air temperatures, for series and parallel arrangements than the asphalt cover. The pressure drop increased proportionally with EAHE's length. Simulation results indicate a ±6 times larger pressure loss in the series design compared to the parallel setup. The effectiveness-NTU relationship shows that parallel EAHEs are 15 % more effective than series ones. • 3-D CFD simulation models were developed to investigate the thermal performance of both series and parallel EAHEs. • Sensitivity analysis has been conducted to investigate the effect of design parameters and soil surface covers. • Prediction of the soil temperature profile next to EAHE is presented. • The pressure drop is six times higher in series pipes than in parallel pipes. • The effectiveness-NTU study found parallel EAHEs 15 % more effective than series.
Cited by 11Open access
Juan Wei, Peng Qu, Gary X. Shen · Concepts in Magnetic Resonance Part B · 2006
Parallel imaging performance of overlapped, nonoverlapped loop arrays, and planar strip arrays are investigated at 220 MHz quantitatively based on simulations with finite difference time domain (FDTD) method. Their signal-to-noise ratio (SNR) and geometry factors (g-factors) in different axial and coronal slices are compared. The average SNR of planar strip arrays is up to 6% better than loop coil arrays in the vicinity of coil array plane. The g-factors of planar strip arrays are significantly better when the acceleration factor approaches the number of array elements. Specifically, the four-channel planar strip array modeled in this work gains 46% improvement at acceleration factor four compared with its loop array counterpart. Based on the simulation at 220 MHz, it is demonstrated that planar strip arrays have better parallel imaging performance than loop arrays in the vicinity of the coil array and for high acceleration factors. © 2006 Wiley Periodicals, Inc. Concepts Magn Reson Part B (Magn Reson Engineering) 29B: 84–94, 2006.
Cited by 1
Demirtaş, Uğurtan, Sezer Uzol, Nilay · OpenMETU (Middle East Technical University) · 2022
Cited by 0Open access