Improving GPGPU Concurrency with Elastic Kernels

Each new generation of GPUs vastly increases the resources available to GPGPU programs. GPU programming models (like CUDA) were designed to scale to use these resources. However, we find that CUDA programs actually do not scale to utilize all available resources, with over 30% of resources going unused on average for programs of the Parboil2 suite that we used in our work. Current GPUs therefore allow concurrent execution of kernels to improve utilization. In this work, we study concurrent execution of GPU kernels using multiprogram workloads on current NVIDIA Fermi GPUs. On two-program workloads from the Parboil2 benchmark suite we find concurrent execution is often no better than serialized execution. We identify that the lack of control over resource allocation to kernels is a major serialization bottleneck. We propose transformations that convert CUDA kernels into elastic kernels which permit fine-grained control over their resource usage. We then propose several elastic-kernel aware concurrency policies that offer significantly better performance and concurrency compared to the current CUDA policy. We evaluate our proposals on real hardware using multiprogrammed workloads constructed from benchmarks in the Parboil 2 suite. On average, our proposals increase system throughput (STP) by 1.21x and improve the average normalized turnaround time (ANTT) by 3.73x for two-program workloads when compared to the current CUDA concurrency implementation.

Synconset Waves and Chains: Spiking Onsets in Synchronous Populations Predict and Are Predicted by Network Structure

Synfire waves are propagating spike packets in synfire chains, which are feedforward chains embedded in random networks. Although synfire waves have proved to be effective quantification for network activity with clear relations to network structure, their utilities are largely limited to feedforward networks with low background activity. To overcome these shortcomings, we describe a novel generalisation of synfire waves, and define `synconset wave' as a cascade of first spikes within a synchronisation event. Synconset waves would occur in `synconset chains', which are feedforward chains embedded in possibly heavily recurrent networks with heavy background activity. We probed the utility of synconset waves using simulation of single compartment neuron network models with biophysically realistic conductances, and demonstrated that the spread of synconset waves directly follows from the network connectivity matrix and is modulated by top-down inputs and the resultant oscillations. Such synconset profiles lend intuitive insights into network organisation in terms of connection probabilities between various network regions rather than an adjacency matrix. To test this intuition, we develop a Bayesian likelihood function that quantifies the probability that an observed synfire wave was caused by a given network. Further, we demonstrate it's utility in the inverse problem of identifying the network that caused a given synfire wave. This method was effective even in highly subsampled networks where only a small subset of neurons were accessible, thus showing it's utility in experimental estimation of connectomes in real neuronal-networks. Together, we propose synconset chains/waves as an effective framework for understanding the impact of network structure on function, and as a step towards developing physiology-driven network identification methods. Finally, as synconset chains extend the utilities of synfire chains to arbitrary networks, we suggest utilities of our framework to several aspects of network physiology including cell assemblies, population codes, and oscillatory synchrony.

Elastic-plastic contact conditions for frictionally constrained bodies under cyclic tangential loading

A Frictionally constrained condition implies dependence of friction force on tangential displacement amplitude. The condition may occur due to chemical, physical, and/or mechanical interaction between the contacting surfaces. The condition, sometimes also referred to as the presliding condition or partial slip condition, is characterized under fretting. Under such conditions, various experimental studies indicate the existence of two distinguishable regions, that is, stick region and slip region. In the present study, frictionally constrained conditions are identified and the evolutions of stick-slip regions are investigated in detail. Investigations have been performed on self-mated stainless steel and chromium carbide coated surfaces mated against stainless steel, under both vacuum and ambient conditions. Contact conditions prevailing at the contact interface were identified based on the mechanical responses and were correlated with the surface damage observed. Surface degradation has been observed in the form of microcracks and material transfer. Detailed numerical analysis has also been performed in order to understand the energy dissipation and the damage mode involved in the surface or subsurface damage. It has been observed that under frictionally constrained conditions, the occurrence of annular slip features are mainly due to the junction growth, resulting from elastic-plastic deformation at the contact interface. Ratcheting has been observed as the governing damage mode under cyclic tangential loading condition.

Elastic thickness structure of the Andaman subduction zone: Implications for convergence of the Ninetyeast Ridge

We use the Bouguer coherence (Morlet isostatic response function) technique to compute the spatial variation of effective elastic thickness (T-e) of the Andaman subduction zone. The recovered T-e map resolves regional-scale features that correlate well with known surface structures of the subducting Indian plate and the overriding Burma plate. The major structure on the India plate, the Ninetyeast Ridge (NER), exhibits a weak mechanical strength, which is consistent with the expected signature of an oceanic ridge of hotspot origin. However, a markedly low strength (0< T-e <3 km) in that region, where the NER is close to the Andaman trench (north of 10 N), receives our main attention in this study. The subduction geometry derived from the Bouguer gravity forward modeling suggests that the NER has indented beneath the Andaman arc. We infer that the bending stresses of the viscous plate, which were reinforced within the subducting oceanic plate as a result of the partial subduction of the NER buoyant load, have reduced the lithospheric strength. The correlation, T-e < T-s (seismogenic thickness) reveals that the upper crust is actively deforming beneath the frontal arc Andaman region. The occurrence of normal-fault earthquakes in the frontal arc, low Te zone, is indicative of structural heterogeneities within the subducting plate. The fact that the NER along with its buoyant root is subducting under the Andaman region is inhibiting the subduction processes, as suggested by the changes in trench line, interrupted back-arc volcanism, variation in seismicity mechanism, slow subduction, etc. The low T-e and thinned crustal structure of the Andaman back-arc basin are attributed to a thermomechanically weakened lithosphere. The present study reveals that the ongoing back-arc spreading and strike-slip motion along the West Andaman Fault coupled with the ridge subduction exerts an important control on the frequency and magnitude of seismicity in the Andaman region. (C) 2013 Elsevier Ltd. All rights reserved.

Transverse elastic response of bundled coated cylinders

The overall elastic response of a bundle of coated cylinders is a major aspect of thermal, nuclear and automotive engineering designs. This paper extends the previous work on tubular bundles to assess the effect of coating material and thickness. A major contribution from this paper is determining the overall transverse elastic response of coated thick cylinders by extending the Michell stress function approach in conjunction with contact mechanics. Finite element results using contact elements pave the way for applying the contact stress boundary conditions for Michell analysis. Theoretical and finite element analyses overall give results consistent with the previous work, and the results also fall within the well-established Voigt-Reuss bounds. (C) 2013 Elsevier Ltd. All rights reserved.

Spatial variations of effective elastic thickness over the Ninetyeast Ridge and implications for its structure and tectonic evolution

We present new data on the strength of oceanic lithosphere along the Ninetyeast Ridge (NER) from two independent methods: spectral analysis (Bouguer coherence) using the fan wavelet transform technique, and spatial analysis (flexure inversion) with the convolution method. The two methods provide effective elastic thickness (T-e) patterns that broadly complement each other, and correlate well with known surface structures and regional-scale features. Furthermore, our study presents a new high resolution database on the Moho configuration, which obeys flexural isostasy, and exhibit regional correlations with the T-e variations. A continuous ridge structure with a much lower T-e value than that of normal oceanic lithosphere provides strong support for the hotspot theory. The derived T-e values vary over the northern (higher T-e similar to 10-20 km), central (anomalously low T-e similar to 0-5 km), and southern (low T-e similar to 5 km) segments of the NER. The lack of correlation of the T-e value with the progressive aging of the lithosphere implies differences in thermo-mechanical setting of the crust and underlying mantle in different parts of the NER, again indicating diversity in their evolution. The anomalously low T-e and deeper Moho (similar to 22 km) estimates of the central NER (between 0.5 degrees N and 17 degrees S) are attributed to the interaction of a hotspot with the Wharton spreading ridge that caused significant thermal rejuvenation and hence weakening of the lithosphere. The higher mechanical strength values in the northern NER (north of 0.5 degrees N) may support the idea of off-ridge emplacement and a relatively large plate motion at the time of volcanism. The low T-e and deeper Moho (similar to 22 km) estimates in the southern part (south of 17 degrees S) suggest that the lithosphere was weak and therefore younger at the time of volcanism, and this supports the idea that the southern NER was emplaced on the edge of the Indian plate. (C) 2013 Elsevier B.V. All rights reserved.

Chirality dependent elastic properties of single-walled boron nitride nanotubes under uniaxial and torsional loading

The elastic behavior of single-walled boron nitride nanotubes is studied under axial and torsional loading. Molecular dynamics simulation is carried out with a tersoff potential for modeling the interatomic interactions. Different chiral configurations with similar diameter are considered to study the effect of chirality on the elastic and shear moduli. Furthermore, the effects of tube length on elastic modulus are also studied by considering different aspects ratios. It is observed that both elastic and shear moduli depend upon the chirality of a nanotube. For aspect ratios less than 15, the elastic modulus reduces monotonically with an increase in the chiral angle. For chiral nanotubes, the torsional response shows a dependence on the direction of loading. The difference between the shear moduli against and along the chiral twist directions is maximum for chiral angle of 15 degrees, and zero for zigzag (0 degrees) and armchair (30 degrees) configurations. (C) 2014 AIP Publishing LLC.

Tectonic inheritance of the Indian Shield: New insights from its elastic thickness structure

A new evaluation of the elastic thickness (Te) structure of the Indian Shield, derived from isotropic fan wavelet methodology, documents spatial variations of lithospheric deformation in different tectonic provinces correlated with episodic tectono-thermal events. The Te variations corroborated by shear velocity, crustal thickness, and seismogenic thickness reveal the heterogeneous rheology of the Indian lithosphere. The thinned, attenuated lithosphere beneath Peninsular India is considered to be the reason for its mechanically weak strength (<30 km), where a decoupled crust-mantle rheology under different surface/subsurface loading structures may explain the prominent low Te patterns. The arcuate Te structure of the Western Dharwar province and a NNE-trending band of low Te anomaly in the Southern Granulite Terrane are intriguing patterns. The average Te values (40-50 km) of the Central Indian Tectonic Zone, the Bastar Craton, and the northern Eastern Ghats Mobile Belt are suggestive of old, stable, Indian lithosphere, which was not affected by any major tectono-thermal events after cratonic stabilization. We propose that the anomalously high Te (60-85 km) and high S-wave velocity zone to the north of the Narmada-Son Lineament, mainly in NW Himalaya, and the northern Aravalli and Bundelkhand Cratons, suggest that Archean lithosphere characterized by a high velocity mantle keel supports the orogenic topographic loads in/near the Himalaya. The Te map clearly segments the volcanic provinces of the Indian Shield, where the signatures of the Reunion, Marion, and Kerguelen hotspots are indicated by significantly low Te patterns that correlate with plume- and rift-related thermal and mechanical rejuvenation, magmatic underplating, and crustal necking. The correlations between Te variations and the occurrence of seismicity over seismically active zones reveal different causal relationships, which led to the current seismogenic zonation of the Indian Shield. (C) 2013 Elsevier B.V. All rights reserved.

Large-Scale Elastic Net Regularized Linear Classification SVMs and Logistic Regression

Elastic Net Regularizers have shown much promise in designing sparse classifiers for linear classification. In this work, we propose an alternating optimization approach to solve the dual problems of elastic net regularized linear classification Support Vector Machines (SVMs) and logistic regression (LR). One of the sub-problems turns out to be a simple projection. The other sub-problem can be solved using dual coordinate descent methods developed for non-sparse L2-regularized linear SVMs and LR, without altering their iteration complexity and convergence properties. Experiments on very large datasets indicate that the proposed dual coordinate descent - projection (DCD-P) methods are fast and achieve comparable generalization performance after the first pass through the data, with extremely sparse models.

Elastic Resources Framework in IaaS, preserving performance SLAs

Elasticity in cloud systems provides the flexibility to acquire and relinquish computing resources on demand. However, in current virtualized systems resource allocation is mostly static. Resources are allocated during VM instantiation and any change in workload leading to significant increase or decrease in resources is handled by VM migration. Hence, cloud users tend to characterize their workloads at a coarse grained level which potentially leads to under-utilized VM resources or under performing application. A more flexible and adaptive resource allocation mechanism would benefit variable workloads, such as those characterized by web servers. In this paper, we present an elastic resources framework for IaaS cloud layer that addresses this need. The framework provisions for application workload forecasting engine, that predicts at run-time the expected demand, which is input to the resource manager to modulate resource allocation based on the predicted demand. Based on the prediction errors, resources can be over-allocated or under-allocated as compared to the actual demand made by the application. Over-allocation leads to unused resources and under allocation could cause under performance. To strike a good trade-off between over-allocation and under-performance we derive an excess cost model. In this model excess resources allocated are captured as over-allocation cost and under-allocation is captured as a penalty cost for violating application service level agreement (SLA). Confidence interval for predicted workload is used to minimize this excess cost with minimal effect on SLA violations. An example case-study for an academic institute web server workload is presented. Using the confidence interval to minimize excess cost, we achieve significant reduction in resource allocation requirement while restricting application SLA violations to below 2-3%.

Diffusion in an elastic medium: A model for macromolecule transport across the nuclear pore complex

Nuclear pore complexes (NPCs) are very selective filters that sit on the membrane of the nucleus and monitor the transport between the cytoplasm and the nucleoplasm. For the central plug of NPC two models have been suggested in the literature. The first suggests that the plug is a reversible hydrogel while the other suggests that it is a polymer brush. Here we propose a model for the transport of a protein through the plug, which is general enough to cover both the models. The protein stretches the plug and creates a local deformation, which together with the protein, we refer to as the bubble. We start with the free energy for creation of the bubble and consider its motion within the plug. The relevant coordinate is the center of the bubble which executes random walk. We find that for faster relaxation of the gel, the diffusion of the bubble is greater. (C) 2014 Elsevier-B.V. All rights reserved.

SODAR observations of inertia- gravity waves in the atmospheric boundary layer during the passage of tropical cyclone

This study reports characteristics of inertia-gravity waves (IGWs) in the atmospheric boundary layer during the passage of Tropical Cylone-03B, using the Doppler Sound Detection and Ranging (SODAR) observations at the Indian tropical station of Gadanki (13.45 degrees N, 79.2 degrees E; near the east coast of India). Wavelet analysis of horizontal winds indicates significant wave motion (60h) near the characteristic inertial period. The hodograph analysis of the filtered winds shows an anti-cyclonic turning of horizontal wind with height and time, indicating the presence of IGW. This study finds important implications in boundary layer dynamics during the passage of tropical cyclones.

Two-step Synthesis of MoS2 Nanotubes using Shock Waves with Lead as Growth Promoter

A new two-step procedure for the synthesis of MoS2 nanotubes using lead as a growth promoter is reported. In the first step, molybdenum suboxide nanowhiskers containing a small amount of lead atoms were created by exposing a pressed MoS2+Pb mixture to highly compressed shock-heated argon gas, with estimated temperatures exceeding 9900 K. In the second step, these molybdenum suboxide nanowhiskers served as templates for the sulfidization of the oxide into MoS2 nanotubes (by using H2S gas in a reducing atmosphere at 820 degrees C). Unlike the case of WS2 nanotubes, the synthesis of a pure phase of MoS2 nanotubes from molybdenum oxide has proven challenging, due mostly to the volatile nature of the latter at the high requisite reaction temperatures (>800 degrees C). In contrast, the nature and apparent reaction mechanism of the method reported herein are amenable to future scale-up. The high-temperature shockwave system should also facilitate the synthesis of new nanostructures from other layered materials.

First-Principles Study of Structure, Vibrational, and Elastic Properties of Stoichiometric and Calcium-Deficient Hydroxyapatite

Hydroxyapatite (HAp), a primary constituent of human bone, is usually nonstoichiometric with varying Ca/P molar ratios, with the well-known fact that Ca deficiency can cause marked reductions in its mechanical properties. To gain insights into the mechanism of this degradation, we employ first-principles calculations based on density functional theory and determine the effects of Ca deficiency on structure, vibrational, and elastic properties of HAp. Our simulation results confirm a considerable reduction in the elastic constants of HAp due to Ca deficiency, which was experimentally reported earlier. Stress-induced transformation of the Ca-deficient defected structure into a metastable state upon the application of stress could be a reason for this. Local structural stability of HAp and Ca-deficient HAp structures is assessed with full phonon dispersion studies. Further, specific signatures in the computed vibrational spectra for Ca deficiency in HAp can be utilized in experimental characterization of different types of defected HAp.