50175 research outputs found
Sort by
Scheduling Page Table Walks for Irregular GPU Applications
No full textRecent studies on commercial hardware demonstrated that irregular GPU applications can bottleneck on virtual-to-physical address translations. In this work, we explore ways to reduce address translation overheads for such applications. We discover that the order of servicing a GPU's address translation requests (specifically, page table walks) plays a key role in determining the amount of translation overhead experienced by an application. We find that different SIMD instructions executed by an application require vastly different amounts of work to service their address translation needs, primarily depending upon the number of distinct pages they access. We show that better forward progress is achieved by prioritizing translation requests from the instructions that require less work to service their address translation needs. Further, in the GPU's Single-Instruction-Multiple-Thread (SIMT) execution paradigm, all threads that execute in lockstep (wavefront) need to finish operating on their respective data elements (and thus, finish their address translations) before the execution moves ahead. Thus, batching walk requests originating from the same SIMD instruction could reduce unnecessary stalls. We demonstrate that the reordering of translation requests based on the above principles improves the performance of several irregular GPU applications by 30% on average
Making and breaking of small water clusters: A combined quantum chemical and molecular dynamics approach
No full textWe present a combined quantum chemical and molecular dynamics study of cyclic and noncyclic water n-mers ((H 2 O n , n = 2�6) at four different temperatures and showcase that the dynamics of small water clusters can reproduce the known properties of bulk water reasonably well. We investigate the making and breaking of the water clusters by computing the hydrogen bond strengths, average lifetimes, and relative stabilities, which are important to understand the complex solution dynamics. We compare the behavior of water clusters in the gas phase and in the solution phase as well as the variation in the properties as a function of cluster size and highlight the notably more interesting cluster dynamics of the water trimer when compared to the other water clusters. © 2019 Wiley Periodicals, Inc. © 2019 Wiley Periodicals, Inc
Non-Local Patch-Based Regularization for Image Restoration
No full textSeveral patch-based models have been proposed for image restoration in the literature. A common feature with these models is that patches are used for filtering or optimization, where the aggregation is often performed over a non-local (NL) neighborhood. We propose a NL patch-based regularizer, where patches are used (1) for computing weights between NL neighbors, and (2) for defining a TV-type norm in patch space. A general form of the latter construction was originally proposed by Peyre et al. and later studied by other authors. In most of these proposals, both the weights and the image are treated as variables, which makes the model non-convex. In particular, the corresponding numerical solvers cannot guarantee local optimality. On the other hand, our regularizer is convex. Along with an ell-2 data fidelity term, we apply the regularizer for denoising, deblurring and super-resolution, and develop an efficient ADMM solver for computing the global minimum. An interesting finding is that, while our model is weaker than the non-convex counterparts, the minimizer of the former is generally closer to the ground truth than the reconstruction from the latter. Moreover, we demonstrate that our regularizer can outperform existing regularization techniques for deblurring and super-resolution. © 2018 IEEE
Analysis of Electrical Analogue of a Biological Cell and Its Response to External Electric Field
No full textAbstract: The use of electric field stimulation to elicit a desired cell/tissue response has become a versatile strategy in regenerative medicine. Using an array of cell types and biomaterial substrates, our group has experimentally investigated the influence of external electric field parameters on the modulation of cellular functionality in vitro. However, the mechanism of action of electric field is not clearly understood, especially in cases where cell fate processes such as differentiation and proliferation are significantly enhanced due to electric field stimulation. In order to understand these important phenomena, it is necessary to first examine the response on a single cell. In this direction, we analyze the response of an electrical analogue of a single biological cell, wherein an electrical equivalent resistor-capacitor (R-C) network has been constructed by considering membranes as capacitive and surrounding biological media (cytoplasm and nucleoplasm) as resistive components. The response of this electrical analogue of a biological cell to external electric field (E-field) is determined using analytical techniques and SPICE-based simulations. The solutions for the network provide a time constant of � 30 μs, which is higher compared to the case when membranes were considered to be purely capacitive. The above model formulation has been further extended to determine the steady-state current response under various input signals, like sinusoidal, square, and triangular pulses using SPICE simulation package. In the context of regenerative engineering, the results of the present work are perceived to be important to design electric field-based stimulation strategies to obtain desired responses of electroactive tissues. Lay Summary: The importance of the effect of electric field on cells and tissues has become evident over the last two decades. Prior studies indicate that based on the electric field parameters, it is possible to get various cellular responses. The current study is an attempt to investigate why this is the case by approximating a single cell into an equivalent electrical network with resistors and capacitors. The network response is studied using simulation tools to get current waveforms and analytical techniques to obtain time constants, which provide vital insights into the observed cell behaviors reported in the literature. © 2018, The Regenerative Engineering Society
Tau Identification at the CMS Experiment in LHC Run-2
No full textSince Run-1 of the LHC, CMS has taken the opportunity to improve further particle reconstruction. A number of improvements were made to the hadronic tau reconstruction and identification algorithms. In particular, the reconstruction of the tau decay products leaving deposits in the electromagnetic calorimeter was improved to better model signal of pi(0) from tau decays. This modification improves energy response and removes the tau footprint from isolation area. In addition to this, improvements were made to discriminators that combine isolation and tau lifetime variables, and the rejection of electrons misidentified as hadronic taus was improved using multivariate techniques. The results of these improvements using 13 TeV data at LHC Run-2 are presented and validation of tau identification using a variety of techniques has been highlighted
Circuit-Parameter-Based Audiosusceptibility Model for Series Resonant Converter
No full textModels that accurately predict the output voltage ripple magnitude are essential for applications with stringent performance target for it. Impact of the dc input ripple on the output ripple for a frequency-controlled series resonant converter (SRC) using a discrete-domain exact discretization modeling method is analyzed in this paper. A novel discrete state-space model along with a small-signal model for the SRC considering three state variables is presented. The audiosusceptibility (AS) transfer function that relates the input-to-output ripple is derived from the small-signal model. Analysis of the AS transfer function indicates a resonance peak and an expression is derived connecting the AS resonance frequency for the input ripple based on the SRC component values. Further analysis is done to show that a set of values for the SRC parameter exists, which forms a design space, for which the normalized gain offered by the SRC for the input ripple is less than unity at any input ripple frequency. A test setup to introduce the variable frequency ripple at the input of the SRC for the experimental evaluation of the AS transfer function is developed. Influence of stray parameters on the AS gain, AS resonance frequency, and on the SRC tank resonance frequency is evaluated. An SRC is designed at a power level of 10 kW. The analysis using the derived analyticalmodel, simulations, and experimental results are found to be closely matching
Monsoon rainfall over India in June and link with northwest tropical pacific: June ISMR and link with northwest tropical pacific
No full textRecent years have witnessed large interannual variation of all-India rainfall (AIR) in June, with intermittent large deficits and excesses. Variability of June AIR is found to have the strongest link with variation of rainfall over northwest tropical Pacific (NWTP), with AIR deficit (excess) associated with enhancement (suppression) of NWTP rainfall. This association is investigated using high-resolution Meteorological Research Institute model which shows high skill in simulating important features of Asian summer monsoon, its variability and the inverse relationship between NWTP rainfall and AIR. Analysis of the variation of NWTP rainfall shows that it is associated with a change in the latitudinal position of subtropical westerly jet over the region stretching from West of Tibetan Plateau (WTP) to NWTP and the phase of Rossby wave steered in it with centres over NWTP and WTP. In years with large rainfall excess/deficit, the strong link between AIR and NWTP rainfall exists through differences in Rossby wave phase steered in the jet. The positive phase of the WTP-NWTP pattern, with troughs over WTP and west of NWTP, tends to be associated with increased rainfall over NWTP and decreased AIR. This scenario is reversed in the opposite phase. Thus, the teleconnection between NWTP rainfall and AIR is a manifestation of the difference in the phase of Rossby wave between excess and deficit years, with centres over WTP and NWTP. This brings out the importance of prediction of phase of Rossby waves over WTP and NWTP in advance, for prediction of June rainfall over India
Lip reading using simple dynamic features and a novel ROI for feature extraction
No full textDeaf or hard-of-hearing people mostly rely on lip-reading to understand speech. They demonstrate the ability of humans to understand speech from visual cues only. Automatic lip reading systems work in a similar fashion - by obtaining speech or text from just the visual information, like a video of a person's face. In this paper, an automatic lip reading system for spoken digit recognition is presented. The system uses simple dynamic features by creating difference images between consecutive frames of the video input. Using this technique, word recognition rates of 83.79 and 65.58 are achieved in speaker-dependent and speaker-independent testing scenarios, respectively. A novel, extended region-of-interest (ROI) which includes lower jaw and neck region is also introduced. Most lip-reading algorithms use only the mouth/lip region for relevant feature extraction. Over simple mouth as the ROI, the proposed ROI improves the performance by 4 in speaker-dependent tests and by 11 in speaker-independent tests. © 2018 Association for Computing Machinery
CONSENSUS OPTIMIZATION FOR DISTRIBUTED REGISTRATION
No full textWe consider the problem of jointly registering multiple point sets using rigid transforms. We propose a distributed algorithm based on consensus optimization for the least-squares formulation of this problem. In each iteration, the computation is distributed among the point sets and the results are averaged. For each point set, the dominant cost per iteration is the SVD of a square matrix of size d, where d is the ambient dimension. Existing methods for joint registration are either centralized or perform the optimization sequentially. The proposed algorithm is naturally more scalable than these methods. As an application, we integrate the proposed algorithm within a divide-and-conquer approach for sensor network localization. In particular, we are able to localize very large networks, which are beyond the scope of most existing localization methods
Emulation of Wind Turbine System using Vector Controlled Induction Motor Drive
No full textDepleting fossil fuels and concerns about global warming have forced a paradigm shift towards renewable energy resources in meeting our energy requirements. Wind energy utilization has been growing at a rapid rate fuelling research and development in high power wind turbines. To advance research and education in wind energy conversion systems, a controlled test bed is necessary that does not depend on wind availability. In this paper, a control structure is proposed to emulate high power, large inertia wind turbines using low power, small inertia machines. Majority of the existing methods for wind turbine emulation employ DC or induction motors with torque compensation loop. However, it suffers from noise and instability issues. DC motors have several disadvantages compared to induction machines in terms of cost, maintenance, speed and ruggedness, even though their control is simple. Hence, in this paper, Squirrel Cage Induction Machine (SUM) in speed controlled mode is employed for emulation. The performance of the control structure is demonstrated through an experiment involving step change in wind velocity, where the dynamics of a high inertia wind turbine system are emulated. Simulations and experiments are conducted on a 7.5 kW Doubly Fed Induction Generator (DFIG) driven by a 5.5 kW SCIM, emulating a wind turbine