Analytical Performance Estimation Research Articles

Analytical performance estimation during code generation on modern GPUs

Automatic code generation is frequently used to create implementations of algorithms specifically tuned to particular hardware and application parameters. The code generation process involves the selection of adequate code transformations, tuning parameters, and parallelization strategies. We propose an alternative to time-intensive autotuning, scenario-specific performance models, or black-box machine learning to select the best-performing configuration.This paper identifies the relevant performance-defining mechanisms for memory-intensive GPU applications through a performance model coupled with an analytic hardware metric estimator. This enables a quick exploration of large configuration spaces to identify highly efficient code candidates with high accuracy.We examine the changes of the A100 GPU architecture compared to the predecessor V100 and address the challenges of how to model the data transfer volumes through the new memory hierarchy.We show how our method can be coupled to the “pystencils” stencil code generator, which is used to generate kernels for a range-four 3D-25pt stencil and a complex two-phase fluid solver based on the Lattice Boltzmann Method. For both, it delivers a ranking that can be used to select the best-performing candidate.The method is not limited to stencil kernels but can be integrated into any code generator that can generate the required address expressions.

Inductance Measurement Methods for Surface-Mount Permanent Magnet Machines

Analytical performance estimation of a permanent magnet (PM) motor requires an accurate equivalent circuit model. In a lumped electrical model of a PM motor, resistance and inductances appear as passive elements, which are used to represent the phase winding resistance, inductance, core loss, etc. There is currently no available standard for parameter measurement of permanent magnet (PM) motors. In the literature, there are many studies on inductance measurement. However, they are applied to different types of motors. The purpose of this study is to evaluate those different inductance measurement methods, on the same motors, to identify whether they lead to the same result. Also, it was aimed to find out the difficulties involved in the measurement process. This study concentrates on determining the d-axis and q-axis inductances of two different surface-mount PM motors at standstill and under running conditions. The standstill measurement methods evaluated include the “current decay” method and the “DC inductance bridge” method as well as more common methods. The dependence of the inductances on the current magnitude, frequency, and excitation signal waveform is investigated. Measurements with PWM and sinusoidal AC voltage excitation are found to give similar results. The tests indicated that the “current decay” method is prone to measurement errors especially when the phase resistance is low. It is discovered that inductance measurements from standstill tests are independent of frequency for all practical purposes. Next, the same inductances are measured, while the test motors are running. The methods considered include; inductance measurement from no-load test, zero power factor load test, and unity power factor load test; while the machine is in generating mode. Furthermore, a new inductance measurement method is introduced where the measurement is made while the test motor is driven with a vector-controlled drive. Finally, inductance measurement results from different standstill tests and running tests are compared and evaluated.

Analytical and experimental investigation on heat transfer and flow parameters of Multichannel louvered fin cross flow heat exchanger using iterative LMTD and ∊-NTU method

Automotive Heat exchangers plays vital role by ensuring human comfort in cabin by maintaining optimum temperature and maintains operating temperature of Internal Combustion(IC) engine within permissible limits. In this work, iteration based mathematical model to solve for analytical thermo-hydraulic performance estimation of a typical heat exchanger is adopted with Log Mean Temperature Difference (LMTD) method by incorporating the variation in properties of fluids during each iteration, since the overall thermal resistance ratio between targeted & calculated Heat Rejection Rate (HRR) has to be converged to unity to solve for the actual thermal performance of the heat exchanger. This condition ensures convergence and there is no deviation between the two adjacent iterations in terms of HRR value. Additionally, the above condition ensures that deviation between analytical value calculated by Effectiveness-NTU (∊-NTU) method and experimental data will be reduced significantly due to the proposed iterative LMTD approach adopted in this work. Along with iterative LMTD methodology, ∊-NTU method is adopted for last iteration to cross-verify the obtained value of the HRR for same operating conditions. In order to evaluate heat transfer coefficients for air and coolant side, standard correlations are used from well-established and validated literature data. For coolant side pressure drop calculations, network solution methodology is adopted where fluid flow behavior in each segment of heat exchanger is studied in detail and analytical relationships are used to predict net pressure drop in entire heat exchanger. In our mathematical model, louvered fin effects on heat transfer is taken into account by adopting colburn approach and the geometrical effects of internal fins on flow channels is also included. Experimental validations tests are carried out to compare with the analytical results of HRR, Coolant Side Pressure Drop (CSPD) and Air Side Pressure Drop (ASPD) across heat exchanger. A good agreement is found between experimental and analytical results which validates the accuracy of present mathematical model and methodology, the maximum deviation for HRC is 4% on higher side, for CSPD and ASPD it is 15.4%,8.7% on higher side respectively.

Performance limits of adaptive-optics/high-contrast imagers with pyramid wavefront sensors

ABSTRACT Advanced adaptive-optics (AO) systems will likely utilize pyramid wavefront sensors (PWFSs) over the traditional Shack–Hartmann sensor in the quest for increased sensitivity, peak performance and ultimate contrast. Here, we explain and quantify the PWFS theoretical limits as a means to highlight its properties and applications. We explore forward models for the PWFS in the spatial-frequency domain: these prove useful because (i) they emanate directly from physical-optics (Fourier) diffraction theory; (ii) they provide a straightforward path to meaningful error breakdowns; (iii) they allow for reconstruction algorithms with $O (n\, \log(n))$ complexity for large-scale systems; and (iv) they tie in seamlessly with decoupled (distributed) optimal predictive dynamic control for performance and contrast optimization. All these aspects are dealt with here. We focus on recent analytical PWFS developments and demonstrate the performance using both analytic and end-to-end simulations. We anchor our estimates on observed on-sky contrast on existing systems, and then show very good agreement between analytical and Monte Carlo performance estimates on AO systems featuring the PWFS. For a potential upgrade of existing high-contrast imagers on 10-m-class telescopes with visible or near-infrared PWFSs, we show, under median conditions at Paranal, a contrast improvement (limited by chromatic and scintillation effects) of 2×–5× when just replacing the wavefront sensor at large separations close to the AO control radius where aliasing dominates, and of factors in excess of 10× by coupling distributed control with the PWFS over most of the AO control region, from small separations starting with an inner working angle of typically 1–2 λ/D to the AO correction edge (here 20 λ/D).

Corrigendum: Analytical performance estimation of network‐assisted D2D communications in urban scenarios with rectangular cells

Corrigendum: Analytical performance estimation of network‐assisted D2D communications in urban scenarios with rectangular cells

Analytical performance estimation of network‐assisted D2D communications in urban scenarios with rectangular cells

AbstractThe aggressive spatial reuse of radio resources is considered today as one of the most promising avenues to achieve significant cellular capacity improvements in future fifth‐generation networks. Accordingly, device‐to‐device (D2D) communications are an emerging paradigm that promises to offer these much expected gains without the need for additional investments into the network infrastructure. However, before this attractive technology can be deployed ubiquitously, the research community has to fully understand the extent of its potential performance benefits across typical scenarios of interest. In this work, we consider one such use case of rectangular cells (common for offices, shopping malls, dormitories, etc.) and develop the corresponding analytical methodology for D2D performance evaluation. As our target metric, we employ the signal‐to‐interference (SIR) ratio experienced by a D2D user. To this end, we propose two relatively simple approximations for SIR distribution and hence capture the related parameters, including user throughput. Further, we carefully investigate the most interesting numerical results by making important conclusions on the envisioned operation of our chosen scenario. In particular, we demonstrate that under certain conditions, the SIR behaviour is insensitive to the dimensions of cells, while different propagation exponents ‘scale’ its density function thus allowing to simplify the characterisation of SIR in a wide range of input parameters. Copyright © 2015 John Wiley & Sons, Ltd.

On-orbit verification of fuel-free attitude control system for spinning solar sail utilizing solar radiation pressure

This paper introduces a new attitude control system for a solar sail, which leverages solar radiation pressure. This novel system achieves completely fuel-free and oscillation-free attitude control of a flexible spinning solar sail. This system consists of thin-film-type devices that electrically control their optical parameters such as reflectivity to generate an imbalance in the solar radiation pressure applied to the edge of the sail. By using these devices, minute and continuous control torque can be applied to the sail to realize very stable and fuel-free attitude control of the large and flexible membrane. The control system was implemented as an optional attitude control system for small solar power sail demonstrator named IKAROS (Interplanetary Kite-craft Accelerated by Radiation Of the Sun). In-orbit attitude control experiments were conducted, and the performance of the controller was successfully verified in comparison with the ground-based analytical performance estimation.

A Reactive Cross Collision Exclusionary Backoff Algorithm in IEEE 802.11 Network

An inseparable challenge associated with every random access network is the design of an efficient Collision Resolution Algorithm (CRA), since collisions cannot be completely avoided in such network. To maximize the collision resolution efficiency of a popular CRA, namely Binary Exponential Backoff (BEB), we propose a reactive backoff algorithm. The proposed backoff algorithm is reactive in the sense that it updates the contention window based on the previously selected backoff value in the failed contention stage to avoid a typical type of collision, referred as cross-collision. Cross-collision would occur if the contention slot pointed by the currently selected backoff value appeared to be present in the overlapped portion of the adjacent (the previous and the current) windows. The proposed reactive algorithm contributes to significant performance improvements in the network since it offers a supplementary feature of Cross Collision Exclusion (XCE) and also retains the legacy collision mitigation features. We formulate a Markovian model to emulate the characteristics of the proposed algorithm. Based on the solution of the model, we then estimate the throughput and delay performances of WLAN following the signaling mechanisms of the Distributed Coordination Function (DCF) considering IEEE 802.11b system parameters. We validate the accuracy of the analytical performance estimation framework by comparing the analytically obtained results with the results that we obtain from the simulation experiments performed in ns-2. Through the rigorous analysis, based on the validated model, we show that the proposed reactive cross collision exclusionary backoff algorithm significantly enhances the throughput and reduces the average packet delay in the network.

RaPTEX

Advances in microprocessors, memory, and radio technology have enabled the emergence of embedded systems that rely on communication systems to exchange information and coordinate their activities in spatially distributed applications. However, developing embedded communication systems that satisfy specific application requirements is a challenge due to the many tradeoffs imposed by different choices of underlying protocols and their parameters. Furthermore, evaluating the correctness and performance of the design and implementation before deploying it is a nontrivial task due to the complexity of the resulting system. This article presents the design and implementation of RaPTEX, a rapid prototyping tool for embedded communication systems, especially well suited for wireless sensor networks (WSNs), consisting of three major subsystems: a toolbox, an analytical performance estimation framework, and an emulation environment. We use a hierarchical approach in the design of the toolbox to facilitate the composition of the network stack. For fast exploration of the tradeoff space at design time, we build an analytical performance estimation model for energy consumption, delay, and throughput. For realistic performance evaluation, we design and implement a hybrid, accurate, yet scalable, emulation environment. Through three use cases, we study the tradeoff space for different protocols and topologies, and highlight the benefits of using RaPTEX for designing and evaluating embedded communication systems for WSNs.

Seizure prediction: Any better than chance?

To test whether epileptic seizure prediction algorithms have true predictive power, their performance must be compared with the one expected under well-defined null hypotheses. For this purpose, analytical performance estimates and seizure predictor surrogates were introduced. We here extend the Monte Carlo framework of seizure predictor surrogates by introducing alarm times surrogates. We construct artificial seizure time sequences and artificial seizure predictors to be consistent or inconsistent with various null hypotheses to determine the frequency of null hypothesis rejections obtained from analytical performance estimates and alarm times surrogates under controlled conditions. Compared to analytical performance estimates, alarm times surrogates are more flexible with regard to the testable null hypotheses. Both approaches have similar, high statistical power to indicate true predictive power. For Poisson predictors that fulfill the null hypothesis of analytical performance estimates, the frequency of false positive null hypothesis rejections can exceed the significance level for long mean inter-alarm intervals, revealing an intrinsic bias of these analytical estimates. Alarm times surrogates offer important advantages over analytical performance estimates. The key question in the field of seizure prediction is whether seizures can in principle be predicted or whether algorithms which have been presumed to perform better than chance actually are unable to predict seizures and simply have not yet been tested against the appropriate null hypotheses. Alarm times surrogates can help to answer this question.

Statistical Performance Analysis and Estimation for Parallel Multimedia Processing

When parallelizing complex multimedia processing on multiple processors, the stochastic timing behavior should be carefully studied. Although there are already many papers on the performance analysis of stochastic parallel system, they are not targeted on multimedia processing. In this paper, first we study H.264/AVC encoder (running on x86) and QSDPCM encoder (running on TI TMS32C62) to characterize important aspects of the stochastic timing behavior in complex multimedia processing applications. It is shown that the variation and correlation are indeed very significant. In order to make systematic analysis feasible, we apply Stochastic Timed Marked Graph (STMG) as a formal model to capture essential timing related behaviors of parallel multimedia processing systems. Then, we show how the local timing variations and correlations interact and propagate to the global timing behavior; from this we conclude general parallelization guidelines. Furthermore, we develop an analytical performance estimation technique to derive the probability distribution of timing behavior for parallel multimedia processing systems that have correlated stochastic timing behaviors inside. The estimation technique is based on principal component analysis and approximations.

A Stochastic Approach to Predicting Performance of Web Service Composition

In this paper, we propose an analytical approach to predict the performance of web service composition built on BPEL. The approach first translates web service composition specification into Stochastic Petri Nets. From the SPN model and its corresponding continuous-time Markov chain, we derive the analytical performance estimates of process-completion-time. In the case study, we also use computer simulation techniques to validate our analytical model.

Emulating switch-level models of CMOS circuits

This paper presents a method for emulating switch-level models of CMOS circuits using FPGAs. In this method, logic gates are used to model switch-level circuits without any abstraction. In contrast to the abstraction methods for which transistors are grouped together to form gates, in this method, gates are grouped together to form the switch models of transistors. The method presented in this paper, unlike the abstraction methods, can emulate many important features of switch-level models, such as bi-directional signal propagation and variations in driving strength. In order to attain a better utilization of FPGA resources a mixed-mode emulation approach has been used. In this approach parts of the circuit are emulated at the switch-level while the remaining parts of the circuit are emulated at the gate-level. The experimental results show that the presented emulation-based approach could be significantly faster than existing simulation-based approaches. The analytical performance estimation shows that the speed-up grows with the circuit size and is workload dependent.

Analytical Performance Estimation of a Wireless TDMA System with TrafficClassification

This paper describes an analytical model for performance estimation of a mobile communications system based on a TDMA/TDD access scheme. In accordance with the ATM classification of services and its terminology, three types of traffic have been taken into account: CBR for voice connections, VBR for real-time data applications and ABR for non real-time data services. Both voice and data traffic is modeled following Poisson processes. Moreover, three different policies for data resource assignment are evaluated: FCFS, NPPS and silence exploitation (SAD). Theoretical versus simulated results show the suitability of the model when the real-time traffic load is relatively low compared to non real-time traffic.