Software Pipelined Execution of Stream Programs on GPUs

The StreamIt programming model has been proposed to exploit parallelism in streaming applications on general purpose multi-core architectures. This model allows programmers to specify the structure of a program as a set of filters that act upon data, and a set of communication channels between them. The StreamIt graphs describe task, data and pipeline parallelism which can be exploited on modern Graphics Processing Units (GPUs), as they support abundant parallelism in hardware. In this paper, we describe the challenges in mapping StreamIt to GPUs and propose an efficient technique to software pipeline the execution of stream programs on GPUs. We formulate this problem - both scheduling and assignment of filters to processors - as an efficient Integer Linear Program (ILP), which is then solved using ILP solvers. We also describe a novel buffer layout technique for GPUs which facilitates exploiting the high memory bandwidth available in GPUs. The proposed scheduling utilizes both the scalar units in GPU, to exploit data parallelism, and multiprocessors, to exploit task and pipelin parallelism. Further it takes into consideration the synchronization and bandwidth limitations of GPUs, and yields speedups between 1.87X and 36.83X over a single threaded CPU.

Compiler Assisted Leakage Energy Optimization for Clustered VLIW Architectures

Miniaturization of devices and the ensuing decrease in the threshold voltage has led to a substantial increase in the leakage component of the total processor energy consumption. Relatively simpler issue logic and the presence of a large number of function units in the VLIW and the clustered VLIW architectures attribute a large fraction of this leakage energy consumption in the functional units. However, functional units are not fully utilized in the VLIW architectures because of the inherent variations in the ILP of the programs. This underutilization is even more pronounced in the context of clustered VLIW architectures because of the contentions for the limited number of slow intercluster communication channels which lead to many short idle cycles.In the past, some architectural schemes have been proposed to obtain leakage energy bene .ts by aggressively exploiting the idleness of functional units. However, presence of many short idle cycles cause frequent transitions from the active mode to the sleep mode and vice-versa and adversely a ffects the energy benefits of a purely hardware based scheme. In this paper, we propose and evaluate a compiler instruction scheduling algorithm that assist such a hardware based scheme in the context of VLIW and clustered VLIW architectures. The proposed scheme exploits the scheduling slacks of instructions to orchestrate the functional unit mapping with the objective of reducing the number of transitions in functional units thereby keeping them off for a longer duration. The proposed compiler-assisted scheme obtains a further 12% reduction of energy consumption of functional units with negligible performance degradation over a hardware-only scheme for a VLIW architecture. The benefits are 15% and 17% in the context of a 2-clustered and a 4-clustered VLIW architecture respectively. Our test bed uses the Trimaran compiler infrastructure.

Strength of Gamma Rhythm Depends on Normalization

Neuronal assemblies often exhibit stimulus-induced rhythmic activity in the gamma range (30-80 Hz), whose magnitude depends on the attentional load. This has led to the suggestion that gamma rhythms form dynamic communication channels across cortical areas processing the features of behaviorally relevant stimuli. Recently, attention has been linked to a normalization mechanism, in which the response of a neuron is suppressed (normalized) by the overall activity of a large pool of neighboring neurons. In this model, attention increases the excitatory drive received by the neuron, which in turn also increases the strength of normalization, thereby changing the balance of excitation and inhibition. Recent studies have shown that gamma power also depends on such excitatory-inhibitory interactions. Could modulation in gamma power during an attention task be a reflection of the changes in the underlying excitation-inhibition interactions? By manipulating the normalization strength independent of attentional load in macaque monkeys, we show that gamma power increases with increasing normalization, even when the attentional load is fixed. Further, manipulations of attention that increase normalization increase gamma power, even when they decrease the firing rate. Thus, gamma rhythms could be a reflection of changes in the relative strengths of excitation and normalization rather than playing a functional role in communication or control.

Traffic engineered NoC for streaming applications

Streaming applications demand hard bandwidth and throughput guarantees in a multiprocessor environment amidst resource competing processes. We present a Label Switching based Network-on-Chip (LS-NoC) motivated by throughput guarantees offered by bandwidth reservation. Label switching is a packet relaying technique in which individual packets carry route information in the form of labels. A centralized LS-NoC Management framework engineers traffic into Quality of Service (QoS) guaranteed routes. LS-NoC caters to the requirements of streaming applications where communication channels are fixed over the lifetime of the application. The proposed NoC framework inherently supports heterogeneous and ad hoc system-on-chips. The LS-NoC can be used in conjunction with conventional best effort NoC as a QoS guaranteed communication network or as a replacement to the conventional NoC. A multicast, broadcast capable label switched router for the LS-NoC has been designed. A 5 port, 256 bit data bus, 4 bit label router occupies 0.431 mm(2) in 130 nm and delivers peak bandwidth of 80 Gbits/s per link at 312.5 MHz. Bandwidth and latency guarantees of LS-NoC have been demonstrated on traffic from example streaming applications and on constant and variable bit rate traffic patterns. LS-NoC was found to have a competitive AreaxPower/Throughput figure of merit with state-of-the-art NoCs providing QoS. Circuit switching with link sharing abilities and support for asynchronous operation make LS-NoC a desirable choice for QoS servicing in chip multiprocessors. (C) 2013 Elsevier B.V. All rights reserved.

Do gamma oscillations play a role in cerebral cortex?

Gamma rhythm (which has a center frequency between 30 and 80 Hz) is modulated by cognitive mechanisms such as attention and memory, and has been hypothesized to play a role in mediating these processes by supporting communication channels between cortical areas or encoding information in its phase. We highlight several issues related to gamma rhythms, such as low and inconsistent power, its dependence on low-level stimulus features, problems due to conduction delays, and contamination due to spike-related activity that makes accurate estimation of gamma phase difficult. Gamma rhythm could be a potentially useful signature of excitation-inhibition interactions in the brain, but whether it also provides a mechanism for information processing or coding remains an open question.

Stability of scheduled multi-access communication over quasi-static flat fading channels with random coding and independent decoding

The stability of scheduled multiaccess communication with random coding and independent decoding of messages is investigated. The number of messages that may be scheduled for simultaneous transmission is limited to a given maximum value, and the channels from transmitters to receiver are quasistatic, flat, and have independent fades. Requests for message transmissions are assumed to arrive according to an i.i.d. arrival process. Then, we show the following: (1) in the limit of large message alphabet size, the stability region has an interference limited information-theoretic capacity interpretation, (2) state-independent scheduling policies achieve this asymptotic stability region, and (3) in the asymptotic limit corresponding to immediate access, the stability region for non-idling scheduling policies is shown to be identical irrespective of received signal powers.

Stability of scheduled message communication over degraded broadcast channels

We develop a multi-class discrete-time processor-sharing queueing model for scheduled message communication over a discrete memoryless degraded broadcast channel. The framework we consider here models both the random message arrivals and the subsequent reliable communication by suitably combining techniques from queueing theory and information theory. Requests for message transmissions are assumed to arrive according to i.i.d. arrival processes. Then, (i) we derive an outer bound to the stability region of message arrival rate vectors achievable by the class of stationary scheduling policies, (ii) we show for any message arrival rate vector that satisfies the outer bound, that there exists a stationary "state-independent" policy that results in a stable system for the corresponding message arrival processes, and (iii) under an asymptotic regime, we show that the stability region of information arrival rate vectors is the information-theoretic capacity region of a degraded broadcast channel.

Information-theoretic and communication-theoretic optimal power allocation for fading channels

In this paper we review the most peculiar and interesting information-theoretic and communications features of fading channels. We first describe the statistical models of fading channels which are frequently used in the analysis and design of communication systems. Next, we focus on the information theory of fading channels, by emphasizing capacity as the most important performance measure. Both single-user and multiuser transmission are examined. Further, we describe how the structure of fading channels impacts code design, and finally overview equalization of fading multipath channels.

Scheduling for Stable and Reliable Communication Over Multiaccess Channels and Degraded Broadcast Channels

We model communication of bursty sources: 1) over multiaccess channels, with either independent decoding or joint decoding and 2) over degraded broadcast channels, by a discrete-time multiclass processor sharing queue. We utilize error exponents to give a characterization of the processor sharing queue. We analyze the processor sharing queue model for the stable region of message arrival rates, and show the existence of scheduling policies for which the stability region converges to the information-theoretic capacity region in an appropriate limiting sense.

Correlated Gaussian Sources over Orthogonal Gaussian Channels

We consider the transmission of correlated Gaussian sources over orthogonal Gaussian channels. It is shown that the Amplify and Forward (AF) scheme which simplifies the design of encoders and the decoder, performs close to the optimal scheme even at high SNR. Also, it outperforms a recently proposed scalar quantizer scheme both in performance and complexity. We also study AF when there is side information at the encoders and decoder.

Three-User Cognitive Channels with Cumulative Message Sharing: An Achievable Rate Region

In this paper, an achievable rate region for the three-user discrete memoryless interference channel with asymmetric transmitter cooperation is derived. The three-user channel facilitates different ways of message sharing between the transmitters. We introduce a manner of noncausal (genie aided) unidirectional message-sharing, which we term cumulative message sharing. We consider receivers with predetermined decoding capabilities, and define a cognitive interference channel. We then derive an achievable rate region for this channel by employing a coding scheme which is a combination of superposition and Gel'fand-Pinsker coding techniques.

Detection of SFBC-OFDM signals in frequency- and time-selective MIMO channels

Use of space-frequency block coded (SFBC) OFDM signals is advantageous in high-mobility broadband wireless access, where the channel is highly time- as well as frequency-selective because of which the receiver experiences both inter-symbol interference (ISI) as well as inter-carrier interference (10). ISI occurs due to the violation of the 'quasi-static' fading assumption caused due to frequency- and/or time-selectivity of the channel. In addition, ICI occurs due to time-selectivity of the channel which results in loss of orthogonality among the subcarriers. In this paper, we are concerned with the detection of SFBC-OFDM signals on time- and frequency-selective MIMO channels. Specifically, we propose and evaluate the performance of an interference cancelling receiver for SFBC-OFDM which alleviates the effects of ISI and ICI in highly time- and frequency-selective channels.

Performance of Quantized Equal Gain Transmission With Noisy Feedback Channels

In this paper, new results and insights are derived for the performance of multiple-input, single-output systems with beamforming at the transmitter, when the channel state information is quantized and sent to the transmitter over a noisy feedback channel. It is assumed that there exists a per-antenna power constraint at the transmitter, hence, the equal gain transmission (EGT) beamforming vector is quantized and sent from the receiver to the transmitter. The loss in received signal-to-noise ratio (SNR) relative to perfect beamforming is analytically characterized, and it is shown that at high rates, the overall distortion can be expressed as the sum of the quantization-induced distortion and the channel error-induced distortion, and that the asymptotic performance depends on the error-rate behavior of the noisy feedback channel as the number of codepoints gets large. The optimum density of codepoints (also known as the point density) that minimizes the overall distortion subject to a boundedness constraint is shown to be the same as the point density for a noiseless feedback channel, i.e., the uniform density. The binary symmetric channel with random index assignment is a special case of the analysis, and it is shown that as the number of quantized bits gets large the distortion approaches the same as that obtained with random beamforming. The accuracy of the theoretical expressions obtained are verified through Monte Carlo simulations.

Constellation Constrained Capacity of Two-user Broadcast Channels

Capacity region for two-user Gaussian Broadcast Channels (GBC) is well known with the optimal input being Gaussian. In this paper we explore the capacity region for GBC when the users' symbols are taken from finite complex alphabets (like M-QAM, M-PSK). When the alphabets for both the users are the same we show that rotation of one of the alphabets enlarges the capacity region. We arrive at an optimal angle of rotation by simulation. The effect of rotation on the capacity region at different SNRs is also studied using simulation results. Using the setup of Fading Broadcast Channel (FBC) given by [Li and Goldsmith, 2001], we study the ergodic capacity region with inputs from finite complex alphabets. It is seen that, using the procedure for optimum power allocation obtained in [Li and Goldsmith, 2001] for Gaussian inputs, to allocate power to symbols from finite complex alphabets, relative rotation between the alphabets does not improve the capacity region. Simulation results for a modified heuristic power allocation procedure for finite-constellation case, show that Constellation Constrained capacity region enlarges with rotation.

Information Theoretic Results for Three-User Cognitive Channels

In this paper, we introduce the three-user cognitive radio channels with asymmetric transmitter cooperation, and derive achievable rate regions under several scenarios depending on the type of cooperation and decoding capability at the receivers. Two of the most natural cooperation mechanisms for the three-user channel are considered here: cumulative message sharing (CMS) and primary-only message sharing (PMS). In addition to the message sharing mechanism, the achievable rate region is critically dependent on the decoding capability at the receivers. Here, we consider two scenarios for the decoding capability, and derive an achievable rate region for each one of them by employing a combination of superposition and Gel'fand-Pinsker coding techniques. Finally, to provide a numerical example, we consider the Gaussian channel model to plot the rate regions. In terms of achievable rates, CMS turns out to be a better scheme than PMS. However, the practical aspects of implementing such message-sharing schemes remain to be investigated.