Language：Japanese   Publisher：Academic eXcange for Information Environment and Strategy  

DOI： 10.24669/axies.2024.0_143

CiNii Research

Language：Japanese   Publisher：Academic eXchange for Information Environment and Strategy  

DOI： 10.24669/jacn.28.1_114

CiNii Research

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

ABSTRACT

Mixed‐precision computation is a promising method for substantially improving high‐performance computing applications. However, using mixed‐precision data is a double‐edged sword. While it can improve computational performance, the reduction in precision introduces more uncertainties and errors. As a result, precision tuning is necessary to determine the optimal mixed‐precision configurations. Much effort is therefore spent on selecting appropriate variables while balancing execution time and numerical accuracy. Auto‐tuning (AT) is one of the technologies that can assist in alleviating this intensive task. In recent years, ppOpen‐AT, an AT language, introduced a directive for mixed‐precision tuning called “Blocks.” In this study, we investigated an AT strategy for the “Blocks” directive for multi‐region tuning of a program. The non‐hydrostatic icosahedral atmospheric model (NICAM), a global cloud‐resolving model, was used as a benchmark program to evaluate the effectiveness of the AT strategy. Experimental results indicated that when a single region of the program performed well in mixed‐precision computation, combining these regions resulted in better performance. When tested on the supercomputer “Flow” Type I (Fujitsu PRIMEHPC FX1000) and Type II (Fujitsu PRIMEHPC CX1000) subsystems, the mixed‐precision NICAM benchmark program tuned by the AT strategy achieved a speedup of nearly 1.31× on the Type I subsystem compared to the original double‐precision program, and a 1.12× speedup on the Type II subsystem.

DOI： 10.1002/cpe.8326

Web of Science

Scopus

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

Concerns have arisen regarding the unregulated utilization of artificial
intelligence (AI) outputs, potentially leading to various societal issues.
While humans routinely validate information, manually inspecting the vast
volumes of AI-generated results is impractical. Therefore, automation and
visualization are imperative. In this context, Explainable AI (XAI) technology
is gaining prominence, aiming to streamline AI model development and alleviate
the burden of explaining AI outputs to users. Simultaneously, software
auto-tuning (AT) technology has emerged, aiming to reduce the man-hours
required for performance tuning in numerical calculations. AT is a potent tool
for cost reduction during parameter optimization and high-performance
programming for numerical computing. The synergy between AT mechanisms and AI
technology is noteworthy, with AI finding extensive applications in AT.
However, applying AI to AT mechanisms introduces challenges in AI model
explainability. This research focuses on XAI for AI models when integrated into
two different processes for practical numerical computations: performance
parameter tuning of accuracy-guaranteed numerical calculations and sparse
iterative algorithm.

DOI： 10.1109/MCSoC64144.2024.00095

Scopus

arXiv

Other Link： http://arxiv.org/pdf/2405.10973v1

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

In this paper, support vector machine (SVM) performance was assessed
utilizing a quantum-inspired complementary metal-oxide semiconductor (CMOS)
annealer. The primary focus during performance evaluation was the accuracy rate
in binary classification problems. A comparative analysis was conducted between
SVM running on a CPU (classical computation) and executed on a quantum-inspired
annealer. The performance outcome was evaluated using a CMOS annealing machine,
thereby obtaining an accuracy rate of 93.7% for linearly separable problems,
92.7% for non-linearly separable problem 1, and 97.6% for non-linearly
separable problem 2. These results reveal that a CMOS annealing machine can
achieve an accuracy rate that closely rivals that of classical computation.

DOI： 10.1109/MCSoC64144.2024.00094

Scopus

arXiv

Other Link： http://arxiv.org/pdf/2404.15752v2

CiNii Research

CiNii Research

Publishing type：Research paper (international conference proceedings)   Publisher：ACM  

DOI： 10.1145/3636480.3637221

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/hpcasia/hpcasia2024w.html#KawaiIHH24

Language：Japanese   Publisher：Society of Computer Chemistry, Japan  

<p>In August 2023, we released the latest version of our ABINIT-MP program, Open Version 2 Revision 8. In this version, the most commonly used FMO-MP2 calculations are even faster than in the previous Revision 4. It is now also possible to calculate excitation and ionization energies for regions of interest. Improved interaction analysis is also available. In addition, we have started GPU-oriented modifications. In this preliminary report, we present the current status of ABINIT-MP.</p>

DOI： 10.2477/jccj.2024-0001

CiNii Research

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

DOI： 10.1109/CLUSTERWorkshops61563.2024.00033

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/cluster/clusterw2024.html#RenKTNKKHN24

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2404.15752

Language：Japanese   Publisher：Academic eXcange for Information Environment and Strategy  

DOI： 10.24669/axies.2023.0_67

CiNii Research

Language：Japanese   Publisher：Academic eXcange for Information Environment and Strategy  

DOI： 10.24669/axies.2023.0_81

CiNii Research

Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

DOI： 10.1109/candar60563.2023.00031

Scopus

Language：Japanese   Publisher：The Japanese Society of Medical Imaging Technology  

<p>With the development of medical imaging technology, various techniques have been developed and used to visually understand the inside of the living body. However, these technologies can only directly obtain images and videos, and diagnosis is still performed by human hands, such as physicians. There are great expectations for software that can reduce such labor, and an increasing number of technologies are already being used in the field of medicine, but the target is limited because they require knowledge and skills in both medicine (medical imaging) and computing technology. Therefore, in this study, researchers in the medical imaging and high-performance computing fields are collaborating to accelerate and scale up the image reconstruction of PET. This paper describes the details of this effort and the results obtained so far.</p>

DOI： 10.11409/mit.41.150

CiNii Research

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

DOI： 10.1109/IPDPSW59300.2023.00115

Web of Science

Scopus

Other Link： https://dblp.uni-trier.de/db/conf/ipps/ipdps2023w.html#HashinokiOKNH23

Language：English   Publishing type：Research paper (scientific journal)   Publisher：Oxford University Press (OUP)  

Summary

Large-scale earthquake sequence simulations using the boundary element method (BEM) incur extreme computational costs through multiplying a dense matrix with a slip rate vector. Hierarchical matrices (H-matrices) have often been used to accelerate this multiplication. However, the complexity of the structures of the H-matrices and the communication costs between processors limit their scalability, and they therefore cannot be used efficiently in distributed memory computer systems. Lattice H-matrices have recently been proposed as a tool to improve the parallel scalability of H-matrices. In this study, we developed a method for earthquake sequence simulations applicable to 3D nonplanar faults with lattice H-matrices. We present a simulation example and verify the mesh convergence of our method for a 3D nonplanar thrust fault using rectangular and triangular discretizations. We also performed performance and scalability analyses of our code. Our simulations, using over ${10^5}$ degrees of freedom, demonstrated a parallel acceleration beyond ${10^4}$ MPI processors and a &amp;gt; 10-fold acceleration over the best performance when the normal H-matrices are used. Using this code, we can perform unprecedented large-scale earthquake sequence simulations on geometrically complex faults with supercomputers. The software is made an open-source and freely available.

DOI： 10.1093/gji/ggac386

Web of Science

arXiv

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

DOI： 10.1109/CLUSTER51413.2022.00056

Web of Science

Other Link： https://dblp.uni-trier.de/db/conf/cluster/cluster2022.html#HoshinoIH22

Open Access

CiNii Research

Open Access

CiNii Research

CiNii Research

Open Access

CiNii Research

Publishing type：Research paper (international conference proceedings)  

DOI： 10.1007/s11548-018-1766-y

Language：Japanese  

Open Access

J-GLOBAL

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE Computer Society  

DOI： 10.1109/CLUSTER.2018.00016

Web of Science

Language：English   Publishing type：Research paper (scientific journal)   Publisher：Springer Verlag  

Parallel Boundary Element Method (BEM) analyses are typically conducted using a purpose-built software framework called BEM-BB. This framework requires a user-defined function program that calculates the i-th row and the j-th column of the coefficient matrix arising from the convolution integral term in the fundamental BEM equation. Owing to this feature, the framework can encapsulate MPI and OpenMP hybrid parallelization with H-matrix approximation. Therefore, users can focus on implementing a fundamental solution or a Green’s function, which is the most important element in BEM and depends on the targeted physical phenomenon, as a user-defined function. However, the framework does not consider single instruction multiple data (SIMD) vectorization, which is important for high-performance computing and is supported by the majority of existing processors. Performing SIMD vectorization of a user-defined function is difficult because SIMD exploits instruction-level parallelization and is closely associated with the user-defined function. In this paper, a conceptual framework for enhancing SIMD vectorization is proposed. The proposed framework is evaluated using two BEM problems, namely, static electric field analysis with a perfect conductor and static electric field analysis with a dielectric, on Intel Broadwell (BDW) processor and Intel Xeon Phi Knights Landing (KNL) processor. It offers good vectorization performance with limited SIMD knowledge, as can be verified from the numerical results obtained herein. Specifically, in perfect conductor analyses conducted using the H-matrix, the framework achieved performance improvements of 2.22x and 4.34x compared to the original BEM-BB framework for the BDW processor and KNL, respectively.

DOI： 10.1007/978-3-319-93698-7_46

Web of Science

Scopus

Language：Japanese  

J-GLOBAL

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

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

Open Access

J-GLOBAL

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

Directive-based programming interfaces such as OpenACC and OpenMP are becoming more prevalent in application development targeting accelerators, in particular when porting existing CPU-only code. Unlike vendor-specific alternatives such as CUDA, they are designed to be portable across different accelerators, and therefore once necessary directives are added to an existing CPU-only code, it can be executed on different accelerator architectures depending on the availability of supporting compilers. However, it does not automatically mean that such code runs efficiently on different architectures, and in fact, architecture-specific coding such as choosing optimal data layouts is almost mandatory for optimal performance, imposing a significant burden if implemented manually. Towards realizing performance portability in accelerator programming, we propose a set of extended directives that allow the programmer to optimize data layouts for a given accelerator without modifying original program code. Unlike the manual approach, the code change is confined in the directives with the original code kept as it is. This paper evaluates the effectiveness of our proposed extensions in the OpenACC standard by extending UPACS and CCS-QCD OpenACC applications. A prototype source-to-source translator for the extensions achieves 123% and 120% of the baseline performance, respectively, which are comparable to manually tuned versions.

DOI： 10.1109/HPCC-SmartCity-DSS.2016.34

Web of Science

Language：English   Publishing type：Research paper (international conference proceedings)   Publisher：Institute of Electrical and Electronics Engineers Inc.  

OpenACC is gaining momentum as an implicit and portable interface in porting legacy CPU-based applications to heterogeneous, highly parallel computational environment involving many-core accelerators such as GPUs and Intel Xeon Phi. OpenACC provides a set of loop directives similar to OpenMP for the parallelization and also to manage data movement, attaining functional portability across different heterogeneous devices
however, the performance portability of OpenACC is said to be insufficient due to the characteristics of different target devices, especially those regarding memory layouts, as automated attempts by the compilers to adapt is currently difficult. We are currently working to propose a set of directives to allow compilers to have better semantic information for adaptation
here, we particularly focus on data layout such as Structure of Arrays, advantageous data structure for GPUs, as opposed to Array of Structures, which exhibits good performance on CPUs. We propose a directive extension to OpenACC that allows the users to flexibility specify optimal layouts, even if the data structures are nested. Performance results show that we gain as much as 96 % in performance for CPUs and 165% for GPUs compared to programs without such directives, essentially attaining both functional and performance portability in OpenACC.

DOI： 10.1109/WACCPD.2014.12

Scopus

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

OpenACC is a new accelerator programming interface that provides a set of OpenMP-like loop directives for the programming of accelerators in an implicit and portable way. It allows the programmer to express the offloading of data and computations to accelerators, such that the porting process for legacy CPU-based applications can be significantly simplified. This paper focuses on the performance aspects of OpenACC using two microbenchmarks and one real-world computational fluid dynamics application. Both evaluations show that in general OpenACC performance is approximately 50% lower than CUDA. However, for some applications it can reach up to 98% with careful manual optimizations. The results also indicate several limitations of the OpenACC specification that hamper full use of the GPU hardware resources, resulting in a significant performance gap when compared to a highly tuned CUDA code. The lack of a programming interface for the shared memory in particular results in as much as three times lower performance.

DOI： 10.1109/CCGrid.2013.12

Web of Science

Generative AI technology based on Large Language Models (LLM) has been
developed and applied to assist or automatically generate program codes. In
this paper, we evaluate the capability of existing general LLMs for Basic
Linear Algebra Subprograms (BLAS) code generation for CPUs. We use two LLMs
provided by OpenAI: GPT-4.1, a Generative Pre-trained Transformer (GPT) model,
and o4-mini, one of the o-series of Reasoning models. Both have been released
in April 2025. For the routines from level-1 to 3 BLAS, we tried to generate
(1) C code without optimization from routine name only, (2) C code with basic
performance optimizations (thread parallelization, SIMD vectorization, and
cache blocking) from routine name only, and (3) C code with basic performance
optimizations based on Fortran reference code. As a result, we found that
correct code can be generated in many cases even when only routine name are
given. We also confirmed that thread parallelization with OpenMP, SIMD
vectorization, and cache blocking can be implemented to some extent, and that
the code is faster than the reference code.

DOI： 10.48550/arXiv.2507.04697

arXiv

Other Link： http://arxiv.org/pdf/2507.04697v1

Language：Japanese   Publishing type：Article, review, commentary, editorial, etc. (scientific journal)   Publisher：Society of Computer Chemistry, Japan  

<p>The fragment molecular orbital (FMO) program ABINIT-MP has a quarter-century history, and related research and development of the Open Version 2 series is currently underway. This paper first summarizes the current status of the latest Revision 8 (released on August 2023). It then describes future improvements and enhancements, including GPU support. The connection with coarse-grained simulation (dissipative particle dynamics) and the possibility of cooperation with quantum computation are also touched upon.</p>

DOI： 10.2477/jccj.2024-0022

Web of Science

CiNii Research

Language：Japanese   Publisher：Academic eXchange for Information Environment and Strategy  

DOI： 10.24669/jacn.28.1_114

CiNii Research

Language：Japanese   Publisher：東京 : 日本計算工学会  

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I033560385

Language：Japanese   Publishing type：Rapid communication, short report, research note, etc. (scientific journal)   Publisher：Society of Computer Chemistry, Japan  

In August 2023, we released the latest version of our ABINIT-MP program, Open Version 2 Revision 8. In this version, the most commonly used FMO-MP2 calculations are even faster than in the previous Revision 4. It is now also possible to calculate excitation and ionization energies for regions of interest. Improved interaction analysis is also available. In addition, we have started GPU-oriented modifications. In this preliminary report, we present the current status of ABINIT-MP.

DOI： 10.2477/jccj.2024-0001

CiNii Research

Language：Japanese   Publisher：情報処理学会  

量子アニーリングが注目されており，その中でも疑似量子アニーラのサービスが展開されている．一方で、サポートベクターマシン（SVM）は広くクラス分類に使われているが，疑似量子アニーラでの評価は十分でない．そこで本研究では，疑似量子アニーラの一つであるCMOSアニーリングマシンでSVMにより誤差を有するクラス分類を行い，その性能を評価する．性能評価では2値分類問題における正解率を評価対象とし，訓練データの誤差の割合ごとに，従来のCPUにおけるSVMとCMOSアニーリングマシンにおけるSVMを比較する．

CiNii Research

Language：Japanese   Publisher：情報処理学会  

AIを使用することで複雑な計算に対して入力データから処理結果や実行時間などを予測することが研究されているが、その予測の妥当性を検証することは効率的で誤りのない実行のために必須である。そこで本研究では不完全前処理CG法(ICTCG法)の実行時間の予測を行列画像及びICTCG法のパラメータを入力として行う機械学習モデルを生成、説明可能なAI(XAI)ツールとしてSHAPを用いて説明することによってAIの回答の妥当性を検証することを目的とする。ここでICTCG法とは、共役勾配法(CG法)などの反復法の前処理に連立一次方程式の解法として用いられる不完全コレスキー分解(IC分解)前処理に閾値の値を設定した数値計算アルゴリズムである。

CiNii Research

Language：Japanese   Publisher：情報処理学会  

科学技術計算において数値計算ソフトウェアのテスト効率化は非常に重要である。本研究では数値解析ライブラリであるLAPACKを用いて、固有値計算におけるテストシーケンスの最適化を目的とする。ここでテストシーケンスとは固有値計算の理論解がわかっている問題を与え、計算結果がその範囲内に入っているかをチェックする系列である。ソフトウェア工学の観点から数値計算ソフトウェアにおけるテストシーケンスの最適化について評価した例はあまりない。本研究ではまず、LAPACKのライブラリに意図的にバグを発生させ、テストシーケンスを入れ替える最適化によってテスト時間が短縮できるかを検討する。

CiNii Research

Language：Japanese   Publisher：情報処理学会  

境界要素法（BEM）を用いた地震シミュレーションにおいて、大規模な密行列・ベクトル積が現れるため、シミュレーションを高速化するには、この演算を高速で行うことが必要である。従来のアプリケーションでは、密行列を格子H行列に近似した上で、OpenMPとMPIによるCPUを用いた並列化により、密行列・ベクトル積の高速化が行われていた。しかし、並列計算に適したGPUを用いた並列計算は実装されておらず、格子H行列・ベクトル積の高速化の余地が存在している。そこで、本研究では、対象のアプリケーションに対し、OpenACCを用いたGPU並列化による格子H行列・ベクトル積を実装し、その性能の評価を行った。

CiNii Research

Language：Japanese  

CiNii Research

J-GLOBAL

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I033543008

J-GLOBAL

J-GLOBAL

J-GLOBAL

Language：Japanese   Publisher：Academic eXcange for Information Environment and Strategy  

DOI： 10.24669/axies.2023.0_67

CiNii Research

Language：Japanese   Publisher：Academic eXcange for Information Environment and Strategy  

DOI： 10.24669/axies.2023.0_81

CiNii Research

Language：Japanese   Publisher：The Japanese Society of Medical Imaging Technology  

With the development of medical imaging technology, various techniques have been developed and used to visually understand the inside of the living body. However, these technologies can only directly obtain images and videos, and diagnosis is still performed by human hands, such as physicians. There are great expectations for software that can reduce such labor, and an increasing number of technologies are already being used in the field of medicine, but the target is limited because they require knowledge and skills in both medicine (medical imaging) and computing technology. Therefore, in this study, researchers in the medical imaging and high-performance computing fields are collaborating to accelerate and scale up the image reconstruction of PET. This paper describes the details of this effort and the results obtained so far.

DOI： 10.11409/mit.41.150

CiNii Research

Language：Japanese   Publisher：小宮山印刷工業  

CiNii Research

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I032988217

Language：Japanese   Publisher：日本計算工学会  

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I032887313

Language：Japanese  

CiNii Research

J-GLOBAL

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I033209226

Language：Japanese  

CiNii Research

J-GLOBAL

Other Link： https://ndlsearch.ndl.go.jp/books/R000000004-I033353067

J-GLOBAL

J-GLOBAL

Authorship：Lead author  

J-GLOBAL

J-GLOBAL

J-GLOBAL

Language：Japanese  

CiNii Research

J-GLOBAL

Language：English  

CiNii Research

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese  

J-GLOBAL

Language：Japanese   Publisher：一般社団法人情報処理学会  

近年増加傾向にある GPU 等のアクセラレータを搭載した計算環境への既存プログラムの移植方法として，CUDA・OpenCL に代表されるローレベルなプログラミングモデルを用いる方法に対し，ディレクティブベースの OpenACC のようなハイレベルなプログラミングモデルを用いる方法が注目されている．このようなディレクティブベースのプログラミングモデルの利点として，元のプログラムを維持したまま移植を行えるために，デバイス間の機能的な可搬性が高いことがあげられる．しかし現状の OpenACC などの High-level なプログラミングモデルは，スカラプロセッサとメニーコアアクセラレータの得意とするデータレイアウトの相違に対応することが出来ず，異なる性質を持ったデバイス間の性能可搬性に問題がある．そこで本研究では，データレイアウトを抽象化し，異なるデバイス間での性能可搬性を向上させるための OpenACC の拡張ディレクティブを試作し，姫野ベンチマークのデータレイアウトをトランスレーターにより変更し，マルチコア CPU，Intex Xeon Phi，K20X GPU のそれぞれで評価を行った．その結果，オリジナルと同一のデータレイアウトと比較して，Intel Xeon Phi では 27%，K20X GPU では 24%の性能向上が得られることを確認した．

CiNii Research

Language：Japanese   Publisher：一般社団法人情報処理学会  

近年増加傾向にある GPU 等のアクセラレータを搭載した計算環境への既存プログラムの移植方法として，CUDA・OpenCL に代表される Low-level なプログラミングモデルを用いる方法に対し，ディレクティブベースの OpenACC のような High-level なプログラミングモデルを用いる方法が考えられる．このようなディレクティブベースのプログラミングモデルの利点として，元のプログラムを壊さずに移植を行えるために，デバイス間の可搬性が高いことがあげられる．しかし現状の OpenACC などのプログラミングモデルは，スカラプロセッサとメニーコアアクセラレータの得意とするデータレイアウトの相違等に対応することが出来ず，異なる性質を持ったデバイス間の性能可搬性に問題がある．そこで本研究では，データレイアウトを抽象化し，異なるデバイス間での性能可搬性を向上させるための OpenACC の拡張ディレクティブを試作し，評価を行った．

CiNii Research

Language：Japanese   Publisher：情報処理学会  

CiNii Research

Language：Japanese  

地震や気象予測，航空機や高層ビル設計といったシミュレーションに利用される数値流体力学アプリケーションは，近年一般的になりつつある GPU を用いたスーパーコンピュータにおいて，目覚ましい成果を上げている．しかし，GPU を用いたプログラミングは，高い性能を得ること難しいと言われており，レガシープログラムの GPU 環境への移植が問題となっている．本稿では，実際に利用されている大規模流体アプリケーションである UPACS を手動により CUDA 化し，性能と移植コストの面から評価を行った．また，プログラムの移植性を解決すると期待されている，OpenACC の予備評価を行った．これら評価の結果を示し，今後解決すべき課題について述べる．Computational fluid dynamics (CFD) applications used for an earthquake and meteorological simulation are one of the most important application executed with high-speed supercomputers. Especially, GPU-based supercomputers have been showing remarkable performance of CFD applications. However, GPU-programing is still difficult to obtain high performance, which prevents legacy applications from being ported to GPU environment. We apply classical optimizations to a real-world CFD application UPACS and evaluate it&#039;s performance and porting costs, and we also evaluate OpenACC expected to provide portability across CPUs and GPUs. We demonstrate these results of evaluation and mention performance problems should be resolved in the future.

CiNii Research

Language：Japanese   Publisher：情報処理学会  

CiNii Research

Language：English   Publisher：一般社団法人映像情報メディア学会  

DOI： 10.3169/itej.66.817

Scopus

CiNii Research