Automatic Data Allocation and Buffer Management for Multi-GPU Machines

Multi-GPU machines are being increasingly used in high-performance computing. Each GPU in such a machine has its own memory and does not share the address space either with the host CPU or other GPUs. Hence, applications utilizing multiple GPUs have to manually allocate and manage data on each GPU. Existing works that propose to automate data allocations for GPUs have limitations and inefficiencies in terms of allocation sizes, exploiting reuse, transfer costs, and scalability. We propose a scalable and fully automatic data allocation and buffer management scheme for affine loop nests on multi-GPU machines. We call it the Bounding-Box-based Memory Manager (BBMM). BBMM can perform at runtime, during standard set operations like union, intersection, and difference, finding subset and superset relations on hyperrectangular regions of array data (bounding boxes). It uses these operations along with some compiler assistance to identify, allocate, and manage data required by applications in terms of disjoint bounding boxes. This allows it to (1) allocate exactly or nearly as much data as is required by computations running on each GPU, (2) efficiently track buffer allocations and hence maximize data reuse across tiles and minimize data transfer overhead, and (3) and as a result, maximize utilization of the combined memory on multi-GPU machines. BBMM can work with any choice of parallelizing transformations, computation placement, and scheduling schemes, whether static or dynamic. Experiments run on a four-GPU machine with various scientific programs showed that BBMM reduces data allocations on each GPU by up to 75% compared to current allocation schemes, yields performance of at least 88% of manually written code, and allows excellent weak scaling.

Minimum switching loss pulse width modulation for reduced power conversion loss in reactive power compensators

Advanced bus-clamping switching sequences, which employ an active vector twice in a subcycle, are used to reduce line current distortion and switching loss in a space vector modulated voltage source converter. This study evaluates minimum switching loss pulse width modulation (MSLPWM), which is a combination of such sequences, for static reactive power compensator (STATCOM) application. It is shown that MSLPWM results in a significant reduction in device loss over conventional space vector pulse width modulation. Experimental verification is presented at different power levels of up to 150 kVA.

Loss of frugivore seed dispersal services under climate change

The capacity of species to track shifting climates into the future will strongly influence outcomes for biodiversity under a rapidly changing climate. However, we know remarkably little about the dispersal abilities of most species and how these may be influenced by climate change. Here we show that climate change is projected to substantially reduce the seed dispersal services provided by frugivorous vertebrates in rainforests across the Australian Wet Tropics. Our model projections show reductions in both median and long-distance seed dispersal, which may markedly reduce the capacity of many rainforest plant species to track shifts in suitable habitat under climate change. However, our analyses suggest that active management to maintain the abundances of a small set of important frugivores under climate change could markedly reduce the projected loss of seed dispersal services and facilitate shifting distributions of rainforest plant species.

Improved tuning of the extended concentric tube resonator for wide-band transmission loss

Transmission loss (TL) of a simple expansion chamber (SEC) consists of periodic domes with sharp troughs. This limits practical application of the SEC in the variable-speed automobile exhaust systems. Three-fourths of the troughs of the SEC can be lifted by appropriate tuning of the extended inlet/outlet lengths. However, such mufflers suffer from high back pressure and generation of aerodynamic noise due to free shear layers at the area discontinuities. Therefore, a perforate bridge is made between the extended inlet and outlet. It is shown that the TL curve of a concentric tube resonator (CTR) can also be lifted in a similar way by proper tuning of the extended unperforated lengths. Differential lengths have to be used to correct the inlet/outlet lengths in order to account for the perforate inertance. The resonance peak frequencies calculated by means of the 1-D analysis are compared with those of the 3-D FEM, and appropriate differential lengths are calculated. It is shown how different geometric characteristics of the muffler and mean flow affect the differential lengths. A general correlation is obtained for the differential lengths by considering seven relevant geometric and environmental parameters in a comprehensive parametric study. The resulting expressions would help in design of extended-tube CTR for wide-band TL. (C) 2014 Institute of Noise Control Engineering.

A unique strategy towards high dielectric constant and low loss with multiwall carbon nanotubes anchored onto graphene oxide sheets

Multiwall carbon nanotubes (MWNTs) were anchored onto graphene oxide sheets (GOs) via diazonium and C-C coupling reactions and characterized by spectroscopic and electron microscopic techniques. The thus synthesized MWNT-GO hybrid was then melt mixed with 50/50 polyamide6-maleic anhydride-modified acrylonitrile-butadiene-styrene (PA6-mABS) blend to design materials with high dielectric constant (30) and low dielectric loss. The phase morphology was studied by SEM and it was observed that the MWNT-GO hybrid was selectively localized in the PA6 phase of the blend. The 30 scales with the concentration of MWNT-GO in the blends, which interestingly showed a very low dielectric loss (< 0.2) making them potential candidate for capacitors. In addition, the dynamic storage modulus scales with the fraction of MWNT-GO in the blends, demonstrating their reinforcing capability as well.

Optimal power allocation for protective jamming in wireless networks: A flow based model

We address the problem of passive eavesdroppers in multi-hop wireless networks using the technique of friendly jamming. The network is assumed to employ Decode and Forward (DF) relaying. Assuming the availability of perfect channel state information (CSI) of legitimate nodes and eavesdroppers, we consider a scheduling and power allocation (PA) problem for a multiple-source multiple-sink scenario so that eavesdroppers are jammed, and source-destination throughput targets are met while minimizing the overall transmitted power. We propose activation sets (AS-es) for scheduling, and formulate an optimization problem for PA. Several methods for finding AS-es are discussed and compared. We present an approximate linear program for the original nonlinear, non-convex PA optimization problem, and argue that under certain conditions, both the formulations produce identical results. In the absence of eavesdroppers' CSI, we utilize the notion of Vulnerability Region (VR), and formulate an optimization problem with the objective of minimizing the VR. Our results show that the proposed solution can achieve power-efficient operation while defeating eavesdroppers and achieving desired source-destination throughputs simultaneously. (C) 2015 Elsevier B.V. All rights reserved.

Optimal Power Allocation Policies for Low Powered Wireless Communication Systems

We consider optimal average power allocation policies in a wireless channel in the presence of individual delay constraints on the transmitted packets. Power is consumed in transmission of data only. We consider the case when the power used in transmission is a linear function of the data transmitted. The transmission channel may experience multipath fading. We have developed a computationally efficient online algorithm, when there is same hard delay constraint for all packets. Later on, we generalize it to the case when there are multiple real time streams with different hard deadline constraints. Our algorithm uses linear programming and has very low complexity.

Performance of New Dynamic Benefit-Weighted Scheduling Scheme in DiffServ Networks

In this paper, we design a new dynamic packet scheduling scheme suitable for differentiated service (DiffServ) network. Designed dynamic benefit weighted scheduling (DBWS) uses a dynamic weighted computation scheme loosely based on weighted round robin (WRR) policy. It predicts the weight required by expedited forwarding (EF) service for the current time slot (t) based on two criteria; (i) previous weight allocated to it at time (t-1), and (ii) the average increase in the queue length of EF buffer. This prediction provides smooth bandwidth allocation to all the services by avoiding overbooking of resources for EF service and still providing guaranteed services for it. The performance is analyzed for various scenarios at high, medium and low traffic conditions. The results show that packet loss is minimized, end to end delay is minimized and jitter is reduced and therefore meet quality of service (QoS) requirement of a network.

Probabilistic Prediction based Scheduling for Delay Sensitive Traffic in Internet of Things

This paper proposes a probabilistic prediction based approach for providing Quality of Service (QoS) to delay sensitive traffic for Internet of Things (IoT). A joint packet scheduling and dynamic bandwidth allocation scheme is proposed to provide service differentiation and preferential treatment to delay sensitive traffic. The scheduler focuses on reducing the waiting time of high priority delay sensitive services in the queue and simultaneously keeping the waiting time of other services within tolerable limits. The scheme uses the difference in probability of average queue length of high priority packets at previous cycle and current cycle to determine the probability of average weight required in the current cycle. This offers optimized bandwidth allocation to all the services by avoiding distribution of excess resources for high priority services and yet guaranteeing the services for it. The performance of the algorithm is investigated using MPEG-4 traffic traces under different system loading. The results show the improved performance with respect to waiting time for scheduling high priority packets and simultaneously keeping tolerable limits for waiting time and packet loss for other services. Crown Copyright (C) 2015 Published by Elsevier B.V.

QEP Based Solution for Power Allocation in Amplify and Forward Networks

The problem of cooperative beamforming for maximizing the achievable data rate of an energy constrained two-hop amplify-and-forward (AF) network is considered. Assuming perfect channel state information (CSI) of all the nodes, we evaluate the optimal scaling factor for the relay nodes. Along with individual power constraint on each of the relay nodes, we consider a weighted sum power constraint. The proposed iterative algorithm initially solves a set of relaxed problems with weighted sum power constraint and then updates the solution to accommodate individual constraints. These relaxed problems in turn are solved using a sequence of Quadratic Eigenvalue Problems (QEP). The key contribution of this letter is the generalization of cooperative beamforming to incorporate both the individual and weighted sum constraint. Furthermore, we have proposed a novel algorithm based on Quadratic Eigenvalue Problem (QEP) and discussed its convergence.

Efficient Low Complexity Power Allocation Policies for Wireless Communication Systems Guaranteeing QoS

We consider near-optimal policies for a single user transmitting on a wireless channel which minimize average queue length under average power constraint. The power is consumed in transmission of data only. We consider the case when the power used in transmission is a linear function of the data transmitted. The transmission channel may experience multipath fading. Later, we also extend these results to the multiuser case. We show that our policies can be used in a system with energy harvesting sources at the transmitter. Next we consider data users which require minimum rate guarantees. Finally we consider the system which has both data and real time users. Our policies have low computational complexity, closed form expression for mean delays and require only the mean arrival rate with no queue length information.

Pricing and Power allocation in Femtocells using Stackelberg Game Theory

We consider a system with multiple Femtocells operating in a Macrocell. The transmissions in one Femtocell interfere with its neighboring Femtocells as well as with the Macrocell Base Station. We model Femtocells as selfish nodes and the Macrocell Base Station protects itself by pricing subchannels for each usage. We use Stackelberg game model to study this scenario and obtain equilibrium policies that satisfy certain quality of service.