Publisher：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

This paper addresses the implementation of a coupled numerical analysis of the Earth’s magnetospheric dynamics and spacecraft charging (SC) processes based on our in-house Code-To-Code Adapter (CoToCoA). The basic idea is that the magnetohydrodynamic (MHD) simulation reproduces the global dynamics of the magnetospheric plasma, and its pressure and density data at local spacecraft positions are provided and used for the SC calculations. This allows us to predict spacecraft charging that reflects the dynamic changes of the space environment. CoToCoA defines three types of independent programs: Requester, Worker, and Coupler, which are executed simultaneously in the analysis. Since the MHD side takes the role of invoking the SC analysis, Requester and Worker positions are assigned to the MHD and SC calculations, respectively. Coupler then supervises necessary coordination between them. Physical data exchange between the models is implemented using MPI remote memory access functions. The developed program has been tested to ensure that it works properly as a coupled physical model. The numerical experiments also confirmed that the addition of the SC calculations has a rather small impact on the MHD simulation performance with up to about 500-process executions.

DOI： 10.1007/978-3-031-36021-3_46

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

In this paper, we discuss the numerical reproducibility of the compressible fluid behavior around two tone holes of woodwind instruments by using compressible Large Eddy Simulation (LES). In particular, we focus on the situation that the tone holes are opened and closed with moving pads above the tone holes, which is regarded as a moving boundary problem with topology change, and reproduce the change of the pitch when opening and closing the tone holes. Our two-dimensional model of a "recorder" has two tone holes. To reproduce the opening and closing the tone holes, the pads are moved continuously. That is, the position of the pads were continuously changed in the order of "open - close". Our numerical results are consistent with the Keef's experimental results. We solved the moving boundary problem with topology change under the situation of acoustics of fluid-structure interaction, and reproduced the pitch change in the opening and closing the tone holes of the recorder like woodwind instrument model.

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

Two and three-dimensional models of a test-tube thermoacoustic engine were numerically analyzed using compressible Large Eddy Simulation (LES) to investigate initial transient behavior, i.e., generation mechanism of thermoacoustic waves in an initial state. In the model used in this study, a stack is placed near the bottom of the test-tube and a temperature gradient is applied between both ends of the stack. The model has an external region connecting the opening of the test-tube for radiations of sound waves and heat. As a result, a fluid flow was observed inside the stack, a strong pressure oscillation, i.e., acoustic resonance, was observed inside the test-tube, and sound radiation from the open end was also observed. Furthermore, the frequency of the sound vibration was almost the same as the theoretical estimation of the resonance frequency of the test-tube. Thus, we successfully reproduced the basic properties of the thermoacoustic engine in the initial state. However, noise components increased in time evolution and stationary oscillations were not attained yet. Thus, we need to improve our numerical method. We are also planning to analyze sound waves' generation mechanism taking into account aeroacoustic theory, e.g., Lighthill's acoustic analogy.

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

We numerically study port noise observed for a bass reflex speaker system with a compressible fluid solver, Large-Eddy Simulation(LES), to explore the noise generation mechanism from the viewpoint of aeroacoustics. The port noise is considered an aerodynamic sound generated by vortices, created by the interaction between the acoustic field and port opening. However, the detail of the sound generation mechanism is still an open problem. By using a 3D-model, the port-noise is well reproduced by compressible LES, when the speaker system is acoustically driven at its resonance frequency, the Helmholtz resonance frequency of the bass reflex speaker system. Vortices are created near the edges of the port and generate broadband noise. However, noises are enhanced due to resonance in some bands, which correspond to the acoustic resonance frequencies of the port and encloser themselves. We are also planning to investigate the noise generation mechanism by using Howe's energy corollary, which allows us to estimate the energy transfer between fluid dynamics and acoustics, namely we consider the problems of where the aeroacoustic noise is generated and how much energy is transferred from the vortex motions to the acoustic field.

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

A two-dimensional model of an oboe reed is studied numerically with a direct numerical simulation (DNS) of the compressible Navier-Stokes equations to investigate the sound generation mechanism from the viewpoint of aeroacoustics. The numerical tool is extremely accurate due to the smallest mesh size on the order of micrometers and successfully reproduces the details of fluid motion and acoustic vibrations inside and outside the reed. Particular attention is paid to the effect of reed vibration on the sound generation mechanism. When the reeds are fixed and a periodically varying flow is injected through the fixed reed slit, an aerodynamics sound created inside the reeds is an almost monotone including a few overtones. On the other hand, when a flow is injected through periodically vibrating reeds from an oral cavity, more overtone components are observed and the pressure waveforms are similar to those observed in the experiment. This indicates that the richness of the overtones of the double-reed instrument is mainly attributed to the aerodynamic sound created by the flow injected through vibrating reeds and the bore, a linear resonator, just enhances characteristics of the instrument, e.g., formant.

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

A two-dimensional model of a French Horn mouthpiece is numerically studied with a 2D direct numerical simulation (DNS) of the compressible Navier-Stokes equations to investigate the sound generation mechanism from the viewpoint of aeroacoustics. That is, we consider the sounding mechanism of buzzing, when the mouthpiece without a bore is played. Our numerical tool is highly accurate due to the minimum mesh size of the order of the micro-meter, and details of fluid motion and acoustic oscillation inside and near the mouthpiece are successfully reproduced. In particular, we focus on the roles of vibrating lips and an oral cavity in the sound generation mechanism. When the mouthpiece without lips and an oral cavity is driven by a periodic flow with a certain frequency, a single tone without overtones is observed. On the other hand, when the mouthpiece is driven by vibrating lips with an oral cavity, a generating sound includes rich overtones and its waveform is similar to that observed experimentally. Since the bore is a linear element and cannot generates overtones from a single tone by itself, the sound of a horn including rich overtones is generated by a mouthpiece with the vibrating lips and oral cavity.

Scopus

Publisher：Proceedings of the International Congress on Acoustics  

We numerically investigate fundamental problems of the edge tone with compressible Large Eddy Simulation (LES) together with acoustic solver FDTD and incompressible LES. Jet oscillation and edge tone in the first mode are successfully reproduced by a 3D model with compressible LES. Namely, the acoustic intensity changes with the jet velocity well following the sixth power law, which is an evidence of the reliability of our numerical method. Next, we estimate the intensity of acoustic feedback in the following way. According to Kaykayoglu and Rockwell, effective pressure sources are considered to be located slightly downstream of the edge tip on both sides of the edge. Indeed, such a pair of positive and negative pressure spots periodically appear in our numerical calculation. Then, we set the pressure spots on both sides of the edge and reproduce acoustic waves radiated from them by FDTD. The acoustic particle velocity of the reproduced acoustic field at the nozzle outlet is regarded as acoustic feedback. Even though such acoustic feedback may make a contribution to driving the jet, we can consider that the fluid feedback is still dominant in a low-Reynolds number regime as pointed out by Paál et al. from the results of incompressible LES.

Scopus

Language：English   Publishing type：Research paper (scientific journal)  

Language：English   Publishing type：Research paper (scientific journal)  

In previous studies, we determined that workloads often contain many input-output (IO) concentrations. Such concentrations are aggregations of IO accesses. They appear in narrow regions of a storage volume and continue for durations of up to about an hour. These narrow regions occupy a small percentage of the logical unit number capacity, include most IO accesses, and appear at unpredictable logical block addresses. We investigated these workloads by focusing on page-level regularity and found that they often include few regularities. This means that simple caching may not reduce the response time for these workloads sufficiently because the cache migration algorithm uses page-level regularity. We previously developed an on-the-fly automated storage tiering (OTFAST) system consisting of an SSD and an HDD. The migration algorithm identifies IO concentrations with moderately long durations and migrates them from the HDD to the SSD. This means that there is little or no reduction in the response time when the workload includes few such concentrations. We have now developed a hybrid storage system consisting of a cache drive with an SSD and HDD and a multi-tier SSD that uses OTFAST, called "OTF-AST with caching." The OTF-AST scheme handles the IO accesses that produce moderately long duration IO concentrations while the caching scheme handles the remaining IO accesses. Experiments showed that the average response time for our system was 45% that of Facebook FlashCache on a Microsoft Research Cambridge workload.

DOI： 10.1587/transinf.2018EDP7253

Language：English   Publishing type：Research paper (scientific journal)  

The response times of solid state drives (SSDs) have decreased dramatically due to the growing use of non-volatile memory express (NVMe) devices. Such devices have response times of less than 100 micro seconds on average. The response times of all-flash-array systems have also decreased dramatically through the use of NVMe SSDs. However, there are applications, particularly virtual desktop infrastructure and in-memory database systems, that require storage systems with even shorter response times. Their workloads tend to contain many input-output (IO) concentrations, which are aggregations of IO accesses. They target narrow regions of the storage volume and can continue for up to an hour. These narrow regions occupy a few percent of the logical unit number capacity, are the target of most IO accesses, and appear at unpredictable logical block addresses. To drastically reduce the response times for such workloads, we developed an automated tiered storage system called “automated tiered storage with fast memory and slow flash storage” (ATSMF) in which the data in targeted regions are migrated between storage devices depending on the predicted remaining duration of the concentration. The assumed environment is a server with non-volatile memory and directly attached SSDs, with the user applications executed on the server as this reduces the average response time. Our system predicts the effect of migration by using the previously monitored values of the increase in response time during migration and the change in response time after migration. These values are consistent for each type of workload if the system is built using both non-volatile memory and SSDs. In particular, the system predicts the remaining duration of an IO concentration, calculates the expected response-time increase during migration and the expected response-time decrease after migration, and migrates the data in the targeted regions if the sum of response-time decrease after migration exceeds the sum of response-time increase during migration. Experimental results indicate that ATSMF is at least 20% faster than flash storage only and that its memory access ratio is more than 50%.

DOI： 10.1587/transinf.2018PAP0005

Language：English  

This article summarizes the works established in Advanced Communication for Exa (ACE) project. The most important motivation of this project was the severe demands for scalable communication toward Exa-scale computations. Therefore, in the project, we have built a PGAS-based communication library, Advanced Communication Primitives (ACP). Its fundamental communication model is onesided, based on PGAS model, so that it can consume internal memory footprint as small as possible. Based on this model, several applications including simulations of magnetohydrodynamic, molecular orbitals, and particles were tuned to achieve higher scalability. In addition to that, some communication optimization techniques have been investigated. Especially, tuning methods of collective communications, such as message ordering, algorithm selection, and overlapping, are studied. Also, in this project, a network simulator NSIM-ACE is developed. It simulates behavior of packets for one-sided communications to study the effects of congestions on interconnects.

DOI： 10.1007/978-981-13-1924-2_7

Language：English   Publishing type：Research paper (other academic)  

Design of a Flexible In Situ Framework with a Temporal Buffer for Data Processing and Visualization of Time-Varying Datasets.

DOI： 10.1007/978-3-030-02465-9_17

Language：English  

The magnetohydrodynamic (MHD) simulation is often applied to study the global dynamics and configuration of a planetary magnetosphere for the space weather. In this paper, the computational performance of MHD code is evaluated with 128 nodes Xeon Phi KNL of Cray XC40. As the results, the 2D and 3D domain decompositions of SoA (structure of array) make the effective performances and AoS (array of structure) and hybrid parallel computation become low performances. Adding the performance optimizations for Xeon Phi to our MHD simulation code, then we have obtained 2.4 % increase of execution efficiency in total and we achieved 3 TFlops performance gain using 128 nodes.

DOI： 10.3233/978-1-61499-843-3-178

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：English   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

A Neighboring Communication Algorithm Using Effective Multiple Communication Devices on Direct Connection Network

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：English  

DOI： 10.3233/978-1-61499-381-0-387

Language：English  

DOI： 10.1007/978-3-642-45037-2_61

Language：Others  

Implementation of Neighbor Communication Algorithm Using Multi-NICs Effectively by Extended RDMA Interface

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Task Allocation Technique for Avoiding Contentions on Multi-dimensional Mesh/Torus

Language：Others  

A Neighbor Communication Algorithm with Making an Effective Use of NICs on Multidimensional-Mesh/torus

Language：English   Publishing type：Research paper (scientific journal)  

DOI： 10.1109/MAP.2012.6348145

Language：Japanese   Publishing type：Research paper (scientific journal)  

Language：Others  

Task Allocation Optimization for Neighboring Communication on Fat Tree

DOI： 10.1109/HPCC.2012.179

Language：Others  

A Method for Predicting a Penalty of Contentions by Considering Priorities of Routing among Packets on Direct Interconnection Network

DOI： 10.1109/CSO.2011.35

Language：Others  

Task Allocation Method for Avoiding Contentions by the Information of Concurrent Communication

DOI： 10.2316/P.2011.719-025

Language：Japanese   Publishing type：Research paper (scientific journal)  

Dynamic Optimization of Load Balance in MPI Broadcast

Language：English   Publishing type：Research paper (scientific journal)  

Language：Others  

Language：Japanese   Publishing type：Research paper (scientific journal)  

Optimization of MPI Rank Allocation Considering Communication Timing for Reducing Contention

Event date： 2023.6

Language：English   Presentation type：Oral presentation (general)  

Venue：Prague   Country：Czech Republic  

Event date： 2022.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：仙台国際センター   Country：Japan  

Event date： 2021.12

Language：English   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2020.10

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：オンライン   Country：Japan  

Event date： 2020.1

Language：English  

Venue：Fukuoka   Country：Japan  

Event date： 2019.11

Language：English  

Venue：Nagasaki   Country：Japan  

In this study, we have introduced the Code-to-Code Adapter (CoToCoA) library to couple the magnetohydrodynamic (MHD) simulation and the Electron Hybrid (EH) simulation of planetary magnetospheres. CoToCoA has been developed newly to connect the different codes easily. The concept of CoToCoA is that we do not add modifications to each code as possible without data transfer functions, and we do not need to know the referred code without data format. With CoToCoA, we have been developing the cross-reference simulation of macro (MHD) and micro (EH) scales in the magnetosphere. Then, we have evaluated the performance of cross-reference simulation using CoToCoA on the massively parallel computer system.

Event date： 2019.11

Language：English  

Venue：Nagasaki   Country：Japan  

Hybrid storage techniques are useful methods to improve the cost performance for input-output (IO) intensive workloads. These techniques choose areas of concentrated IO accesses and migrate them to an upper tier to extract as much performance as possible through greater use of upper tier areas. Automated tiered storage with fast memory and slow flash storage (ATSMF) is a hybrid storage system situated between non-volatile memories (NVMs) and solid-state drives (SSDs). ATSMF aims to reduce the average response time for IO accesses by migrating areas of concentrated IO access from an SSD to an NVM. When a concentrated IO access finishes, the system migrates these areas from the NVM back to the SSD. Unfortunately, the published ATSMF implementation temporarily consumes much NVM capacity upon migrating concentrated IO access areas to NVM, because its algorithm executes NVM migration with high priority. As a result, it often delays evicting areas in which IO concentrations have ended to the SSD. Therefore, to reduce the consumption of NVM while maintaining the average response time, we developed new techniques for making ATSMF more practical. The first is a queue handling technique based on the number of IO accesses for NVM migration and eviction. The second is an eviction method that selects only write-accessed partial regions in finished areas. The third is a technique for variable eviction timing to balance the NVM consumption and average response time. Experimental results indicate that the average response times of the proposed ATSMF are almost the same as those of the published ATSMF, while the NVM consumption is drastically lower.

Event date： 2019.8

Language：English  

Venue：Fukuoka   Country：Japan  

This paper proposes a novel file transfer tool to improve file transfer performance over Japan high performance computing and networking (JHPCN). We first develop a high-performance and flexible protocol (HpFP) for inter-datacenter transport network. An original HpFP is designed first for specified networks and puts more emphasis on latency and packet loss tolerances than fairness and friendliness, while an enhanced HpFP is more suitable for real network environments. Then, based on the enhanced HpFP, we implement a file transfer tool, called high-performance copy (HCP). The performance of our file transfer tool is evaluated between datacenters of JHPCN using real datasets collected from supercomputer resources. The results show that the HCP achieves higher throughput than traditional tool for file transfer over JHPCN.

Event date： 2018.11

Language：English  

Venue：Takayama   Country：Japan  

Automated tiered storage with fast memory and slow flash storage (ATSMF) is a hybrid storage system located between non-volatile memories (NVMs) and solid state drives (SSDs). ATSMF aims to reduce average response time for inputoutput (IO) accesses by migrating concentrated IO access areas from SSD to NVM. However, the current ATSMF implementation cannot reduce average response time sufficiently because of the bottleneck caused by the Linux brd driver, which is used for the NVM access driver. The response time of the brd driver is more than ten times larger than memory access speed. To reduce the average response time sufficiently, we developed a block-level driver for NVM called a 'two-mode (2M) memory driver.' The 2M memory driver has both the. map IO access mode and direct IO access mode to reduce the response time while maintaining compatibility with the Linux device-mapper framework. The direct IO access mode has a drastically lower response time than the Linux brd driver because the ATSMF driver can execute the IO access function of 2M memory driver directly. Experimental results also indicate that ATSMF using the 2M memory driver reduces the IO access response time to less than that of ATSMF using the Linux brd driver in most cases.

Event date： 2018.6

Language：English  

Venue：Frankfurt   Country：Germany  

This paper presents an in situ framework focused on time-varying simulations, and uses a novel temporal buffer for storing simulation results sampled at user-defined intervals. This framework has been designed to provide flexible data processing and visualization capabilities in modern HPC operational environments composed of powerful front-end systems, for pre-and post-processing purposes, along with traditional back-end HPC systems. The temporal buffer is implemented using the functionalities provided by Open Address Space (OpAS) library, which enables asynchronous one-sided communication from outside processes to any exposed memory region on the simulator side. This buffer can store time-varying simulation results, and can be processed via in situ approaches with different proximities. We present a prototype of our framework, and code integration process with a target simulation code. The proposed in situ framework utilizes separate files to describe the initialization and execution codes, which are in the form of Python scripts. This framework also enables the runtime modification of these Python-based files, thus providing greater flexibility to the users, not only for data processing, such as visualization and analysis, but also for the simulation steering.

Event date： 2018.6

Language：English  

Venue：Frankfurt   Country：Germany  

Event date： 2018.1

Language：English   Presentation type：Oral presentation (invited, special)  

Venue：Tokyo   Country：Japan  

Event date： 2017.11

Language：English  

Venue：Aomori   Country：Japan  

The response time of solid state drives (SSDs) has dramatically reduced according to the spread of non-volatile memory express (NVMe) devices. These devices have response times of less than 100 micro seconds on average. The response time of all-flash-array systems has also drastically reduced through the use of NVMe SSDs. However, there are applications, particularly, virtual desktop infrastructure and in-memory database systems, that require storage systems with even shorter response time. Their workloads were found to contain many input-output (IO) concentrations. We define IO concentration by using a declarative style. Input-output (IO) concentrations are aggregations of IO accesses. They appear in narrow regions of the storage volume and continue for periods of up to about an hour. These narrow regions occupy a few percent of the logical unit number capacity, include most IO accesses, and appear at unpredictable logical block addresses. To drastically reduce the response time of these workloads, we developed automated tiered storage system called 'automated tiered storage with fast memory and slow flash storage' (ATSMF). The memory component of ATSMF is a memory with a non-volatile feature. The system predicts the remaining duration of IO concentration, calculates the response-time increase during migration and response-time decrease after migration, and migrates the IO concentrations if the response-time decrease after migration surpasses the response-time increase during migration. Experimental results indicate that ATSMF is at least 20% faster than flash storage only and its memory access ratio is more than 50%.

Event date： 2017.7

Language：English  

Venue：Vancouver   Country：Canada  

The investigation of related research is very important for research activities. However, it is not easy to choose an appropriate and important academic paper from among the huge number of possible papers. The researcher searches by combining keywords and then selects an paper to be checked because it uses an index that can be evaluated. The citation count is commonly used as this index, but information about recently published papers cannot be obtained. This research attempted to identify good papers using only the words included in the abstract. We constructed a classifier by machine learning and evaluated it using cross validation. As a result, it was found that a certain degree of discrimination is possible.

Event date： 2017.3

Language：English   Presentation type：Oral presentation (general)  

Country：Russian Federation  

Event date： 2016.11

Language：English  

Venue：Hiroshima   Country：Japan  

Various vendors develop a byte accessible Nonvolatile Dual-Inline Memory Module (NVDIMM). The performance of the NVDIMM drastically surpasses that of the Solid State Drive (SSD), which is connected by PCI express. However, the cost of the NVDIMM is much higher than that of the SSD. Therefore, a hybrid storage system between the NVDIMM and SSD is an effective technique for improving cost-performance. If a system uses the NVDIMM less while maintaining performance, its cost-performance should be improved. Our previous work involves on-the-fly automated storage tiering (OTF-AST). OTF-AST is a hybrid storage system consisting of an SSD and HDD. It aims to reduce the average response time of IO accesses by migrating only the IO concentration area to the SSD when IO concentration happens. Therefore, we construct OTF-AST with both the DIMM and SSD and evaluate it in order to understand how to build a cost-effective hybrid storage system with these devices. We use a DIMM instead of a byte accessible NVDIMM, which is difficult to obtain. As a result, we found that the original OTF-AST is suitable for a hybrid storage system consisting of the DIMM and SSD. Moreover, we can improve the performance of OTF-AST if replace its migration algorithm with a more positive migration algorithm. This is because the IO access response time barely increases when the data migration between the DIMM and SSD is done. We will build a more positive migration algorithm in the near future.

Event date： 2016.10

Language：English   Presentation type：Oral presentation (general)  

Venue：Kyoto   Country：Japan  

Event date： 2016.10

Language：English   Presentation type：Oral presentation (general)  

Venue：Kyoto   Country：Japan  

Event date： 2016.10

Language：English   Presentation type：Oral presentation (general)  

Venue：Kyoto   Country：Japan  

Event date： 2016.10

Language：English   Presentation type：Oral presentation (general)  

Venue：Kyoto   Country：Japan  

Event date： 2016.9

Language：English  

Venue：Edinburgh   Country：United Kingdom  

The issue of halo communication is the decrease of parallel scalability. To overcome the issues, we have introduced "Halo thread" to our simulation code. However, we have not solved the issue basically in the strong scaling. In this study, we have developed the Halo functions which perform the halo communication effectively. Then we can perform the calculation and communication in a pipeline and obtained good performance.

Event date： 2016.9

Language：English   Presentation type：Oral presentation (general)  

Country：Portugal  

Event date： 2016.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Birmingham   Country：United Kingdom  

Event date： 2016.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Edinburgh   Country：United Kingdom  

Event date： 2016.8

Language：English   Presentation type：Oral presentation (general)  

Venue：Columbus, Ohio   Country：United States  

Event date： 2016.7

Language：English   Presentation type：Oral presentation (general)  

Country：Portugal  

Event date： 2016.6

Language：English  

Venue：Porto   Country：Portugal  

Current MPI (Message Passing Interface) communication libraries require larger memories in proportion of the number of processes, and can not be used for exa-scale systems. This paper proposes a global memory based communication design to reduce memory usage for exa-scale communication. To realize exa-scale communication, we propose true global memory based communication primitives called Advanced Communication Primitives (ACPs). ACPs provide global address, which is able to use remote atomic memory operations on the global memory, RDMA (Remote Direct Memory Access) based remote memory copy operation, global heap allocator and global data libraries. ACPs are different from the other communication libraries because ACPs are global memory based so that house keeping memories can be distributed to other processes and programmers explicitly consider memory usage by using ACPs. The preliminary result of memory usage by ACPs is 70 MB on one million processes.

Event date： 2016.6

Language：English  

Venue：Nara   Country：Japan  

Previously, communications in parallel programs forHigh Performance Computing (HPC) and Distributed Computing(DC) are mostly written with two-sided communicationinterfaces that are based on a pair of operations, Send andReceive. Since such interface requires explicit synchronizationbetween both sides of the communication, techniquesfor communication optimization such as overlapping are notefficiently described in many cases. On the other hand, onesidedcommunication interface is becoming important as amethod to describe asynchronous communications to enablehighly overlapped communication with computation. As oneof such interface, in this demonstration, Advanced CommunicationPrimitives (ACP) is introduced. ACP is a portableinterface that supports UDP, IBverbs of InfiniBand and Tofulibrary of K Computer. In addition to that, it is designed tobe memory efficient. For example, with 10 thousand processes, the memory consumption of ACP over UDP is estimated to beless than 1MB. Since the number of computational elements isincreasing more rapidly than the amount of available memory, this memory efficiency is becoming one of the keys for parallelprograms in HPC and DC. To show this characteristics, we runACP library on Raspberry Pi 2, and examine its performanceand memory consumption.

Event date： 2016.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Fukuoka   Country：Japan  

Event date： 2016.1

Language：English   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2015.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Chonqing   Country：China  

Event date： 2015.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Sapporo   Country：Japan  

Event date： 2015.5

Language：Japanese   Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2015.4

Language：English   Presentation type：Oral presentation (general)  

Venue：Buenos Aires   Country：Argentina  

Event date： 2015.3

Language：English   Presentation type：Oral presentation (general)  

Venue：Turku   Country：Finland  

Event date： 2015.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Tsukuba   Country：Japan  

Event date： 2014.2

Language：English   Presentation type：Oral presentation (general)  

Venue：Turin   Country：Italy  

Event date： 2013.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：長崎市   Country：Japan  

Event date： 2013.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：長崎市   Country：Japan  

Event date： 2013.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：長崎市   Country：Japan  

Event date： 2013.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Matsuyama   Country：Japan  

Event date： 2013.12

Language：English   Presentation type：Oral presentation (general)  

Venue：Matsuyama   Country：Japan  

Event date： 2013.10

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：那覇市   Country：Japan  

Event date： 2013.9

Language：English   Presentation type：Oral presentation (general)  

Venue：Tokyo   Country：Japan  

Event date： 2013.8

Language：English   Presentation type：Oral presentation (general)  

Venue：Columbus   Country：United States  

Event date： 2013.2 - 2013.3

Language：English   Presentation type：Oral presentation (general)  

Venue：Boston   Country：United States  

Event date： 2013.2

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：あわら市   Country：Japan  

Event date： 2013.1

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：東京   Country：Japan  

Event date： 2013.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Beppu   Country：Japan  

Event date： 2013.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Beppu   Country：Japan  

Event date： 2013.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Beppu   Country：Japan  

Event date： 2013.1

Language：English   Presentation type：Oral presentation (general)  

Venue：Beppu   Country：Japan  

Event date： 2012.12

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：札幌市   Country：Japan  

Event date： 2012.10 - 2012.11

Language：English   Presentation type：Oral presentation (general)  

Venue：Amsterdam   Country：Netherlands  

Event date： 2012.10

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：那覇市   Country：Japan  

スーパーコンピュータの大規模化に伴って，ノード間インターコネクトネットワークとして，コストの低い多次元メッシュ/トーラストポロジを採用したものを用いる事例が増えている．多次元メッシュ/トーラスは，使用するノード数が同じでも，プロセスが配置されるノード群の形状によって性能が大きく変動する．本研究では，京コンピュータや，その互換機である Fujitsu PRIMEHPC FX10で用いられている Tofuインターコネクトネットワークを対象として，プロセス配置の形状による集団通信アルゴリズムの性能への影響を計測した．得られた性能を，Tofuインターコネクトの性能解析ツールを用いて取得した通信衝突による転送待ち時間と比較したところ，プロセス配置形状による変動がどちらもほぼ同じ傾向を示すことを明らかにした．これらの結果から，集団通信アルゴリズムの選択において，プロセス配置の形状を考慮した性能見積もりが重要であることを示した．

Event date： 2012.10

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：那覇市   Country：Japan  

Event date： 2012.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Chicago   Country：United States  

Event date： 2012.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Las Vegas   Country：United States  

Event date： 2012.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Kobe   Country：Japan  

Event date： 2012.7

Language：English   Presentation type：Oral presentation (general)  

Venue：Nice   Country：France  

Event date： 2012.6

Language：English   Presentation type：Oral presentation (general)  

Venue：Liverpool   Country：United Kingdom  

Event date： 2012.6

Language：English   Presentation type：Oral presentation (general)  

Venue：Liverpool   Country：United Kingdom  

Event date： 2012.5

Presentation type：Oral presentation (general)  

Venue：神戸   Country：Japan  

Event date： 2012.3

Presentation type：Oral presentation (general)  

Venue：神戸   Country：Japan  

Event date： 2012.3

Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2012.2

Language：Japanese   Presentation type：Oral presentation (general)  

Venue：東京   Country：Japan  

Event date： 2011.12

Presentation type：Oral presentation (general)  

Venue：東京大学   Country：Japan  

Event date： 2011.11

Presentation type：Oral presentation (general)  

Country：Japan  

Event date： 2011.7

Presentation type：Oral presentation (general)  

Venue：Istanbul   Country：Turkey  

Event date： 2011.4

Presentation type：Oral presentation (general)  

Venue：Kunming   Country：China  

Event date： 2011.2

Presentation type：Oral presentation (general)  

Venue：Innsbruck   Country：Austria  

Event date： 2010.8

Presentation type：Oral presentation (general)  

Venue：金沢   Country：Japan  

Event date： 2010.3

Presentation type：Oral presentation (general)  

Venue：Kona   Country：United States  

Event date： 2009.5

Presentation type：Oral presentation (general)  

Venue：Rome   Country：Italy  

Event date： 2009.5

Presentation type：Oral presentation (general)  

Venue：東京   Country：Japan  

Implementation and Performance Evaluation of Parallelized CG method with PAGME - Preconditioning Method on Hierarchical Parallel Computers

Event date： 2009.4

Presentation type：Oral presentation (general)  

Venue：Sanya, Hainan   Country：China  

Event date： 2009.3

Presentation type：Oral presentation (general)  

Venue：Kaohsiung   Country：Taiwan, Province of China  

Event date： 2009.3

Presentation type：Oral presentation (general)  

Venue：Kaohsiung   Country：Taiwan, Province of China  

Event date： 2008.12

Presentation type：Oral presentation (general)  

Venue：福岡   Country：Japan  

Performance analysis of the CG method with parallelized PAGME

Event date： 2008.12

Presentation type：Oral presentation (general)  

Venue：福岡   Country：Japan  

Evaluation of Dynamic Algorithm Selection with Performance Prediction Models on Alltoall Operation

Event date： 2008.5

Presentation type：Oral presentation (general)  

Venue：東京都   Country：Japan  

Dependence on combination with number of processes and threads for com-
putation times of hybrid-parallel version of IDR(s) Method

Presentation type：Oral presentation (general)  

Venue：Hiro   Country：United States  

Presentation type：Oral presentation (general)  

Venue：Gramad   Country：Brazil  

Venue：Paris   Country：France  

Presentation type：Oral presentation (general)  

Venue：Corfu   Country：Greece  

Presentation type：Oral presentation (general)  

Venue：Corfu   Country：Greece  

Presentation type：Oral presentation (general)  

Venue：Corfu   Country：Greece  

Presentation type：Oral presentation (general)  

Venue：Seoul   Country：Korea, Republic of  

Presentation type：Oral presentation (general)  

Venue：旭川市   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：東京   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：札幌市   Country：Japan  

Presentation type：Oral presentation (general)  

Country：India  

Presentation type：Oral presentation (general)  

Country：Greece  

Presentation type：Oral presentation (general)  

Venue：沖縄県   Country：Japan  

Presentation type：Oral presentation (general)  

Venue：ハワイ   Country：United States  

Venue：Anaheim   Country：United States  

Venue：Barcelona   Country：Spain  

Venue：Rhodes   Country：Greece  

Venue：Las Vegas   Country：United States  

Language：Japanese  

Country：Other  

Language：Japanese  

Country：Other  

Language：English  

Country：Other  

Investigating the relevant literature is very important for research activities. However, it is difficult to select the most appropriate and important academic papers from the enormous number of papers published annually. Researchers search paper databases by combining keywords, and then select papers to read using some evaluation measure—often, citation count. However, the citation count of recently published papers tends to be very small because citation count measures accumulated importance. This paper focuses on the possibility of classifying high-quality papers superficially using attributes such as publication year, publisher, and words in the abstract. To examine this idea, we construct classifiers by applying machine-learning algorithms and evaluate these classifiers using cross-validation. The results show that our approach effectively finds high-quality papers.

Language：Japanese  

ACPライブラリの通信性能およびメモリ使用量の評価

Language：Japanese  

ACPライブラリによるMPI_Comm_spawnの置き換えとOpenFMOへの適用

Language：Japanese  

Development of Effective Halo Communication and Calculation Model on Stencil Computation

Language：Japanese  

ACP通信ライブラリを用いたOpenFMOプログラムの実装

Language：Japanese  

Effects of Halo Thread to the Magnetohydrodynamic Simulation toward Exascale Computing

Language：Japanese  

Interconnection Network Simulation of Synchronization Communication in RDMA

Language：Japanese  

Implementation and Evaluation of ACP Basic layer

Language：Japanese  

ACPライブラリの集団通信インターフェース

Language：Japanese  

NSIM-ACE: A Simulator for Evaluating RDMA on Large-Scale Interconnection Networks

Language：Japanese  

Proposal of a Method for Selecting Algorithm of Collective Communications on Multi-Dimensional Mesh/Torus

Language：Japanese  

通信衝突を削減するタスク配置最適化における通信タイミングの予測方式の影響

Language：Japanese  

Evaluation of optimization of task allocation for reducing contentions

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Language：Japanese  

Optimization of rank allocation considerin communication timing

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)  

Type：Competition, symposium, etc. 

Number of participants：110

Type：Competition, symposium, etc. 

Number of participants：50

Type：Competition, symposium, etc. 

Type：Competition, symposium, etc. 

Number of participants：200

Type：Competition, symposium, etc. 

Number of participants：200