Improving Memory Traffic by Assembly-Level Exploitation of Reuses for Vector Registers

In this paper, we propose a compilation scheme to analyze and exploit the implicit reuses of vector register data. According to the reuse analysis, we present a translation strategy that translates the vectorized loops into assembly vector codes with exploitation of vector reuses. Experimental results show that our compilation technique can improve the execution time and traffic between shared memory and vector registers. Techniques discussed here are simple, systematic, and easy to be implemented in the conventional vector compilers or translators to enhance the data locality of vector registers.     

Efficient Broadcast in Heterogeneous Networks of Workstations Using Two Sub-Networks

This paper presents efficient algorithms for broadcasting on heterogeneous switch-based networks of workstations (HNOW) by two partitioned sub-networks. In an HNOW, many multiple speed types of workstations have different send and receive overheads. Previous research has found that routing by two sub-networks in a NOW can significantly increase system’s performance (Proc. 10th International Conference on Computer Communications and Networks, pp. 68–73, 2001). Similarly, EBS and VBBS (Proc. 8th IEEE International Symposium on Computer and Communication, pp. 1277–1284, (2003)), designed by applying the concept of fastest nodes first, can be executed in O(nlog(n)) time, where n is the number of workstations. This paper proposes two schemes TWO-EBS and TWO-VBBS for broadcasting in an HNOW. These two schemes divide an HNOW into two sub-networks that are routed concurrently and combine EBS and VBBS to broadcast in an HNOW. Based on simulation results, TWO-VBBS outperforms EBS, VBBS, VBBSWF (Proc. 8th IEEE International Symposium on Computer and Communication, pp. 1277–1284, (2003)), the postorder recursive doubling (Proc. Merged IPPS/SPDP Conference, pp. 358–364, (1998)), and the optimal scheduling tree (Proc. Parallel and Distributed Processing Symposium, Proc. 15th International (2001)) generated by dynamic programming in an HNOW.     

A new evaluation to friction analysis for the ring test

A non-constant shear friction approach is proposed to evaluate the friction in ring test. The distribution of shear friction is derived and found to be dependent on the ratio (called F-coefficient) of the second derivative and first derivative of velocity at the z-direction. An upper bound method combined with the F-coefficient is used to simulate the processes of ring test. The calibration curves are reconstructed according to the F-coefficient and show good agreement compared with the literature. A relation between the F-coefficient, neutral line and constant friction factor m are also shown for discussion.     

A Priority MAC Protocol to Support Real-Time Traffic in Ad Hoc Networks

Carrier sense multiple access and its variants have been widely used in mobile ad hoc networks. However, most existing access mechanisms cannot guarantee quality for real-time traffic. This paper presents a distributed medium access control protocol that provides multiple priority levels for stations to compete for the wireless channel. One common channel is assumed to be shared by all stations. Stations are assumed to be able to hear each other (i.e., the network is fully connected). The channel is accessed by stations according to their priorities, and for stations with the same priority, they send frames in a round robin manner. The channel access procedure is divided into three stages: priorities classification period, ID initialization period, and transmission period. Simulation results indicate that our protocol provides high channel utilization and bounded delays for real-time frames.     

An adaptive quorum-based energy conserving protocol for IEEE 802.11 ad hoc networks

The lifetime of a mobile ad hoc network (MANET) depends on the durability of the mobile hosts' battery resources. In the IEEE 802.11 Power Saving Mode, a host must wake up at every beacon interval, to check if it should remain awake. Such a scheme fails to adjust a host's sleep duration according to its traffic, thereby reducing its power efficiency. This paper presents new MAC protocols for power saving in a single hop MANET. The essence of these protocols is a quorum-based sleep/wake-up mechanism, which conserves energy by allowing the host to sleep for more than one beacon interval, if few transmissions are involved. The proposed protocols are simple and energy-efficiency. Simulation results showed that our protocols conserved more energy and extended the lifetime of a MANET.     

Circuit-Switched Broadcasting in Multi-Port Multi-Dimensional Torus Networks

The one-to-all broadcast is the most primary collective communication pattern in a multicomputer network. This paper studies this problem in a circuit-switched torus with -port capability, where a node can simultaneously send and receive messages at one time. This is a generalization of the one-port and all-port models. We show how to efficiently perform broadcast in tori of any dimension, any size, square or nonsquare, using near optimal numbers of steps. The main techniques used are: (i) a span-by-dimension approach, which makes our solution scalable to torus dimensions, and (ii) a squeeze-then-expand approach, which makes possible solving the difficult cases where tori are non-square. Existing results, as compared to ours, can only solve very restricted sizes or dimensions of tori, or use more numbers of steps.     

Fault-tolerant ring embedding in a star graph with both link andnode failures

The star graph interconnection network has been recognized as an attractive alternative to the hypercube network. Previously, the star graph has been shown to contain a Hamiltonian cycle. In this paper, we consider an injured star graph with some faulty links and nodes. We show that even with f_e⩽n-3 faulty links, a Hamiltonian cycle still can be found in an n-star, and that with f_v⩽n-3 faulty nodes, a ring containing at most 4f_v nodes less than that in a Hamiltonian cycle can be found (i.e. the ring contains at least n!-4f_v nodes). In general, in an n-star with f_e faulty links and f_v faulty nodes, where f_e+f_v⩽n-3, our embedding is able to establish a ring containing at least n!-4f_v nodes     

Data broadcasting and seamless channel transition for highlydemanded videos

One way to broadcast a popular video is to use a number of dedicated channels, each responsible for broadcasting some portion of the video periodically in a predefined way. The stress on the channels can be alleviated, and new viewers do not have to wait long to start their playback. Many approaches falling in this category have been proposed. One such scheme that interests us is the fast broadcasting (FB) scheme, which can broadcast a video using k channels by incurring at most O(D/2^k) waiting time on new-coming viewers, where D is the length of the video. We consider a set of videos, each being broadcast by the FB scheme. Since the demand levels on these videos may change with time, it is sometimes inevitable to change the numbers of channels assigned to some videos. We propose a novel seamless channel transition enhancement on top of the FB scheme to dynamically change the number of channels assigned to a video on-the-fly. Clients currently viewing this video will not experience any disruption because of the transition. A channel allocation scheme is also proposed based on the arrival rates of videos to minimize the average waiting experienced by all viewers. From the system manager's point of view, the enhancement will make the FB scheme more attractive     

Artificial neural network approach to authentication of coins by vision-based minimization

A new inspection system, consisting of two procedures for the authentication of coins, is proposed in this paper. In the first procedure, optimum image-matching positions are found by minimizing the matching error of the test coins with their prototype coins. The second procedure is the decision-making process that inspects the coins as genuine or spurious by the Back-Propagation Neural Network combined with the concept of eigen-section. Unlike the traditional approach based on gray-level values, the quantity (8 bits) of the color’s scale has been adopted. The discrimination results are presented and discussed in this study.