Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Kawamura, Takuma; Hasegawa, Yuta; Idomura, Yasuhiro
Journal of Visualization, 27(1), p.89 - 107, 2024/02
Times Cited Count：0 Percentile：0.01(Computer Science, Interdisciplinary Applications)Interactive in-situ steering is an effective tool for debugging, searching for optimal solutions, and analyzing inverse problems in fast and large-scale computational fluid dynamics (CFD) simulations. We propose an interactive in-situ steering framework for large-scale CFD simulations on GPU supercomputers. This framework employs in-situ particle-based volume rendering (PBVR), in-situ data sampling, and a file-based control that enables interactive communication of steering parameters, compressed particle data, and sampled monitoring data between supercomputers and user PCs. The parallelized PBVR is processed on the host CPU to avoid interference with CFD simulations on the GPU. We apply the proposed framework to a real-time plume dispersion analysis code CityLBM on GPU supercomputers. In the numerical experiment, we address an inverse problem to find a pollutant source from the monitoring data, and demonstrate the effectiveness of the human-in-the-loop approach.
Kawamura, Takuma
Dai-34-Kai Suchi Ryutai Rikigaku Shimpojiumu Koen Rombunshu (Internet), 3 Pages, 2020/12
The search for computational parameters in simulations is an important issue in optimizing design variables and increasing the accuracy of simulations. However, due to the recent improvement in the performance of computational units, the data I/O speed has become a bottleneck, making it difficult to store the calculation results consisting of huge parameters in storage. In this study, we focused on in-situ steering, in which computational parameters are explored simultaneously with the computation on the supercomputer. We have extended our previously developed particle-based interactive in-situ visualization framework to develop a technique for interactively steering the computational parameters of batch-processed simulations. We applied this technique to a real-time simulation under development in our mid-term plan, and showed that the user can optimize the parameter search by adjusting the computational parameters in real time with feedback from the visualization.
Kawamura, Takuma; Hasegawa, Yuta; Idomura, Yasuhiro
Proceedings of Joint International Conference on Supercomputing in Nuclear Applications + Monte Carlo 2020 (SNA + MC 2020), p.187 - 192, 2020/10
In order to realize the atmospheric dispersion prediction of pollutants, a fluid simulation by adaptive mesh refinement (AMR) optimized for GPU supercomputer has been developed, and interactive visualization and parameter steering of the simulation results are needed. In this study, we extend particle-based in-situ visualization method for structured grids into AMR, and enables in-situ steering of the simulation parameters by utilizing an in-situ control mechanism via files. By combining the developed method with plume dispersion simulation in urban areas running on a GPU platform, it was shown that human-in-the-loop pollution source search is possible without enormous parameter scanning.
Kawamura, Takuma; Idomura, Yasuhiro
Journal of Visualization, 23(4), p.695 - 706, 2020/08
Times Cited Count：1 Percentile：7.62(Computer Science, Interdisciplinary Applications)An in-situ visualization system based on the particle-based volume rendering offers a highly scalable and flexible visual analytics environment based on multivariate volume rendering. Although it showed excellent computational performance on the conventional CPU platforms, accelerated computation on the latest many core platforms revealed performance bottlenecks related to a function parser and particles I/O. In this paper, we develop a new SIMD-aware function parser and an asynchronous data I/O method based on task-based thread parallelization. Numerical experiments on the Oakforest-PACS, which consists of 8208 Intel Xeon Phi7250 (Knights Landing) processors, demonstrate an order of magnitude speedup with keeping improved strong scaling up to 
100 k cores.
Kawamura, Takuma; Noda, Tomoyuki; Idomura, Yasuhiro
Supercomputing Frontiers and Innovations, 4(3), p.43 - 54, 2017/07
We examine the performance of the in-situ data exploration framework based on the in-situ Particle Based Volume Rendering (In-Situ PBVR) on the latest many-core platform. In-Situ PBVR converts extreme scale volume data into small rendering primitive particle data via parallel Monte-Carlo sampling without costly visibility ordering. This feature avoids severe bottlenecks such as limited memory size per node and significant performance gap between computation and inter-node communication. In addition, remote in-situ data exploration is enabled by asynchronous file-based control sequences, which transfer the small particle data to client PCs, generate view-independent volume rendering images on client PCs, and change visualization parameters at runtime. In-Situ PBVR shows excellent strong scaling with low memory usage up to about 100k cores on the Oakforest-PACS, which consists of 8,208 Intel Xeon Phi7250 (Knights Landing) processors.
Kawamura, Takuma; Noda, Tomoyuki; Idomura, Yasuhiro
no journal, ,
Expanding scales of nuclear simulations make In-Situ visualization more important. In-Situ visualization generates visualization images simultaneously as simulations on the same computing environment. However, in the conventional In-Situ visualization, visualization failure often occurs because of visualization parameters such as a view point, color, and opacity, which are prescribed before batch simulations. To resolve this issue, we developed In-Situ visualization framework, which enables interactive change of visualization parameters using particle data instead of images. Massively parallel particle generation processes are computed without changing domain decomposition models in simulations, and the size of the particle data is compressed small enough than that of the original data. A daemon program transfers the compressed particle data to a client PC, and it also set visualization parameters received from a client PC. We applied the developed tool to simulations of molten debris relocation in reactor pressure vessels using the multi-phase CFD code JUPITER, and the interactive visualization and analysis were enabled without degradation of the simulation performance.
Kawamura, Takuma; Noda, Tomoyuki; Idomura, Yasuhiro
no journal, ,
We examine the performance portability of the In-Situ visualization system based on the Particle Based Volume Rendering (In-Situ PBVR). In this system, parallelized In-Situ processing converts extreme scale volume data into small rendering primitive data given by particles without costly visibility ordering, and the small particles can be rendered as view-independent volume rendering image on client user PC. These features enable us to avoid severe bottlenecks on latest many core platforms such as limited memory size per node and significant performance gap between computation and inter-node communication. The system shows excellent strong scaling up to about 50k threads on the Oakforest-PACS, which consists of 8,208 Intel Xeon Phi7250 (Knights Landing) processors. Asynchronous file based control sequences are designed to enable interactive and remote in-situ data exploration without interfering this strong scaling.
Kawamura, Takuma
no journal, ,
I present the research result of "Remote Interactive In-Situ Visualization using Particle Data for Volume Visualization" supported by the Joint Usage/Research Center for Interdisciplinary Large-scale Information Infrastructures (JHPCN), 2017. With the increase of simulation scale, the conventional visualization process involving data transfer to the pre-post node requires a huge processing cost. To enable real-time visualization and analysis of large-scale simulations, an interactive In-Situ visualization system using particle data for visualization is optimized for the latest SIMD platforms such as XeonPhi, and FX100. Using this system, interactive visualization at runtime is enabled for batch processing simulations.
Kawamura, Takuma
no journal, ,
We discuss the development and application of an in-situ visualization system using particle based visualization data. As the simulation becomes larger, the conventional visualization method involving data transfer to the pre-post node requires a huge amount of processing time. In this study, in order to enable real-time visualization and analysis for large-scale simulations, an interactive in-situ visualization system using particle data was developed and optimized for the latest many-core environment. Using the developed system, it was confirmed that interactive visualization is possible when executing batch processing of simulations. A visualization example of AMR based simulations on the latest GPGPU machine is also reported.
Kawamura, Takuma; Hasegawa, Yuta
no journal, ,
In a large-scale simulation that uses a huge amount of computational resources and runs for a long time on a supercomputer, it is difficult to repeatedly perform recalculation due to adjustment of computational parameters. We have developed a technique for interactive manipulation of simulation parameters by file-based control on an in-situ visualization framework using particle-based volume rendering. We have confirmed that it can be applied to fluid simulation on GPU clusters for interactive visualization and parameter manipulation.
Kawamura, Takuma
no journal, ,
In-Situ PBVR, an in-situ visualization framework based on particle-based volume rendering, provides an advanced visual analysis environment based on multivariate volume rendering. In-Situ PBVR provides a fast and interactive visualization of simulations on a supercomputer. In addition, an interactive parameter steering mechanism is developed for the CityLBM simulation on a GPU supercomputer, and have shown that it is possible to search for contaminants.
