We consider reduction of cache miss which frequently occurs in the operations of Matrix-vector multiplication in case of nonzero entries stored using CCS format of edge-type of matrix. Edge-type of matrix often appears in the filed of superposition of FEM (Finite Element Method) analysis. The efficiency of iterative method as the Conjugate Gradient method depends on both mathematical convergence rate and implementation of Matrix-vector multiplication. In this paper, we propose zigzag CCS format for the store of nonzero entries of edge-type of matrix. Through numerical experiments, we examine that the proposed zigzag CCS format works well as compared with the conventional CCS format in case of edge-type of matrix.

抄録全体を表示

  Epidemic transmission can distribute information over the service area by direct wireless communication between mobile nodes and mobility of the mobile nodes. By using epidemic transmission repeatedly, we can send information to mobile nodes entering a specific area. This is called Information Floating (IF). In this paper, we consider IF to distribute an advertisement to guide customers to a target place (shopping place). The purpose of advertisement delivery is to deliver an advertisement to mobile nodes before they arrive at the target place and to guide mobile nodes to the target place. We divide a service area into some small areas, and consider how to decide the areas where the advertisement is virtually floated. To do this, we use probe data such as traffic volume entering each small area. We consider three methods to decide the areas for IF, and evaluate these methods by computer simulation. Two of the three methods use probe data. In computer simulation, we consider uniform and non-uniform traffic distributions, and small and large traffic volumes to assume various situations. We show effective methods in the various situations.

抄録全体を表示

  We consider to enhance performance of Crout version of incomplete LU decomposition (ILU) for Krylov subspace iterative method using inverse of triangular matrix by means of diagonal compensation. Moreover, we consider also to improve stability of the iterative method. Through numerical experiments, we verify that the proposed Crout ILU decomposition with diagonal compensation works well as compared with the original ILUC decomposition and inverse-based ILUC decomposition as preconditioning.

抄録全体を表示

  GPBiCG (Generalized Product of BiCG) method is often used for the solution of realistic problems on parallel computers. However, the number of synchronizations points for inner products needs three times per iteration, and these synchronization points cause great increase of communication time. Then, we devised a strategy to reduce the number of synchronization points. We applied this strategy to the original BiCGSafe method and succeeded in fair reduction of their communication time. In this paper, we apply this strategy to the original GPBiCG method. Through many numerical experiments, we will make clear that the proposed parallel version of GPBiCG method with double synchronization points outperforms the conventional methods in view of parallel computation.

抄録全体を表示

  This paper treats with an implementation of directive of “reduction” for calculation of inner product in parallel computing. The accuracy of calculation of inner product depends delicately on the order of data because of rounding errors occured by the finite arithmetics. Therefore, we propose an implementation for inner product without directive of “reduction”. we verify effectiveness of the proposed implementation of calculation of inner products in parallel computing.

抄録全体を表示

  Ad-hoc network is expected as a network in place of the cellular network, has been researched widely. The development of routing protocol is one of the problems in an ad-hoc network. Since the flooding packets cause decrease of performance of the network, wasteful communications should refrain. FORP (Flow Oriented Routing Protocol) is one of routing protocols in Ad-hoc network, which uses the mobility information (e.g. location, speed and direction), and chooses the longest expiration route for source and destination pairs. In the flooding of the route search, the packets explode, as the number of nodes increases in the service area. However, FORP doesn't take explosion-control. In this paper, we discuss the effect to the destruction of extra control packets in the route search using the mobility information.

抄録全体を表示

  Recent years have seen a great deal of research into Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communications, which will contribute to advances in driving support systems. In contrast to V2I communications, V2V communications offer the advantage of not requiring infrastructure, but introduction of such systems is not expected to be effective unless there is sufficient penetration by On-Board Units (OBUs). For that reason, it is essential to determine what extent of penetration should be encouraged in the initial stage in order to attain effectiveness. Therefore, we focus on the fact that even though Large Vehicles (LVs) such as large commercial freight carriers are large and tall, and are obstacles blocking V2V communications from the viewpoint of Small Vehicles (SVs), antennas on LVs could also be used to relay communications, thereby facilitating the success of such systems.
  This paper presents experiment s simulating traffic environment s consisting of mixed LVs and SVs, and examined how data V2V communication distribution rates varied with changes of the percentage of vehicles equipped with OBUs. Through the simulation experiments, by causing actively penetrate LVs, it revealed that the reception rate of SVs can be improved, and clarified some guidelines for policies that could encourage wider usage.

抄録全体を表示