Kiyoshi Kumahata

High-performance conjugate gradient performance improvement on the K computer

The high-performance conjugate gradient (HPCG) is new benchmark software for supercomputers that provides a more realistic performance metric than existing benchmarks, such as the LINPACK benchmark. The HPCG measures the speed of solving symmetric sparse linear system equations using the conjugate gradient method preconditioned by a multigrid symmetric Gauss–Seidel smoother. The combination of a sparse linear system and a preconditioned conjugate gradient method is widely used in many scientific and engineering computer applications. This study introduces a tuning method for the K computer. According to weak-scaling measurements on the K computer, it has good parallel scalability. Therefore, our tuning strategy focuses on single CPU performance rather than parallel performance. Single CPU performance strongly depends on memory throughput and multicore utilization. Therefore, we attempt to improve memory/cache access performance and multithreading efficiency. As a result, a HPCG score obtained with the K computer achieved second place at SC’14.

Impaired Air Conditioning within the Nasal Cavity in Flat-Faced Homo.

We are flat-faced hominins with an external nose that protrudes from the face. This feature was derived in the genus Homo, along with facial flattening and reorientation to form a high nasal cavity. The nasal passage conditions the inhaled air in terms of temperature and humidity to match the conditions required in the lung, and its anatomical variation is believed to be evolutionarily sensitive to the ambient atmospheric conditions of a given habitat. In this study, we used computational fluid dynamics (CFD) with three-dimensional topology models of the nasal passage under the same simulation conditions, to investigate air-conditioning performance in humans, chimpanzees, and macaques. The CFD simulation showed a horizontal straight flow of inhaled air in chimpanzees and macaques, contrasting with the upward and curved flow in humans. The inhaled air is conditioned poorly in humans compared with nonhuman primates. Virtual modifications to the human external nose topology, in which the nasal vestibule and valve are modified to resemble those of chimpanzees, change the airflow to be horizontal, but have little influence on the air-conditioning performance in humans. These findings suggest that morphological variation of the nasal passage topology was only weakly sensitive to the ambient atmosphere conditions; rather, the high nasal cavity in humans was formed simply by evolutionary facial reorganization in the divergence of Homo from the other hominin lineages, impairing the air-conditioning performance. Even though the inhaled air is not adjusted well within the nasal cavity in humans, it can be fully conditioned subsequently in the pharyngeal cavity, which is lengthened in the flat-faced Homo. Thus, the air-conditioning faculty in the nasal passages was probably impaired in early Homo members, although they have survived successfully under the fluctuating climate of the Plio-Pleistocene, and then they moved “Out of Africa” to explore the more severe climates of Eurasia.

Minor contributions of the maxillary sinus to the air-conditioning performance in macaque monkeys

ABSTRACT The nasal passages mainly adjust the temperature and humidity of inhaled air to reach the alveolar condition required in the lungs. By contrast to most other non-human primates, macaque monkeys are distributed widely among tropical, temperate and subarctic regions, and thus some species need to condition the inhaled air in cool and dry ambient atmospheric areas. The internal nasal anatomy is believed to have undergone adaptive modifications to improve the air-conditioning performance. Furthermore, the maxillary sinus (MS), an accessory hollow communicating with the nasal cavity, is found in macaques, whereas it is absent in most other extant Old World monkeys, including savanna monkeys. In this study, we used computational fluid dynamics simulations to simulate the airflow and heat and water exchange over the mucosal surface in the nasal passage. Using the topology models of the nasal cavity with and without the MS, we demonstrated that the MS makes little contribution to the airflow pattern and the air-conditioning performance within the nasal cavity in macaques. Instead, the inhaled air is conditioned well in the anterior portion of the nasal cavity before reaching the MS in both macaques and savanna monkeys. These findings suggest that the evolutionary modifications and coetaneous variations in the nasal anatomy are rather independent of transitions and variations in the climate and atmospheric environment found in the habitats of macaques. Summary: Computational fluid dynamics simulations show that, in macaques, morphological variation in the nasal region, including the maxillary sinus, is not explained by the differences and transitions of habitat environment, including temperature and humidity.

Fluid-Structure Interaction Simulation for Blood Vessel Using 3D Voxel Data Derived From Medical Image

We have used the voxel-based method to examine the Fluid-Structure Interaction (FSI) of blood flow and blood vessel wall. This method uses the voxel data that are Cartesian structured grid from medical image. We have confirmed the accuracy and reliability of this method for blood vessel. In this paper, we introduce backgrounds, kinetic models of blood vessel, a numerical method, and a result of experiment.Copyright

Performance improvement of the general-purpose CFD code FrontFlow/blue on the K computer

The general-purpose fluid simulation software FrontFlow/blue (FFB) is based on the finite element method (FEM). It was designed to accept extremely large-scale simulations and is an important application in the manufacturing field in Japan. Moreover, since this application is significant in both the manufacturing field and the development of the post-K supercomputer, it is employed as an important application for the new post-K supercomputer that is under development. The K computer is still the important infrastructure in Japan. And there are some supercomputers having the same architecture to the K computer. Therefore we continue to improve the performance of the FFB on the K computer. On significant subroutines, several improvement techniques, store order based loop modification decreasing total load and store operations, unrolled loop rerolling to employ SIMD load instruction, adjusting number of arrays in loop, using sector cache function, and so on, were employed. As a result, an improvement of 160% was obtained on a single CPU performance. This paper shows and discusses the detail of these improvements.

Examination of Extraction with Vortex Regions in Paranasal Sinus of Human Nose

The nasal cavity has functions which is breathing, smelling, humidification, warming and cleaning of the inhaled air. It is important for human life-sustaining. And nasal cavity has complex anatomy. In this paper, we have examined the flow/vortex in paranasal sinus. Paranasal sinus divides into four parts depend on location, 1) Maxillary Sinuses, 2) Frontal Sinuses, 3) Ethmoidal Sinuses and 4) Spenoid Sinuses. We focus especially Maxillary Sinuses (MS) in this paper. Maxillary Sinuses is an area connected with both sides of both nasal cavities one by one. The examination of the simulation of the nasal cavity that contains the MS is not enough. And, we thought that MS has some influences for flow of nasal cavity. Therefore, we examine flow in nasal cavity that contains MS. In the general, CFD (Computational Fluid Dynamics) simulation results are visualized by general visualization method such as Vector, Stream lines, Particle flow and LIC (Line Integral Convolution). By these techniques, the flow can be abstractly understood. However, the vortex area cannot be clearly visualized. Especially, shape of vortex region is necessary to understand the phenomenon in MS and connected point between nasal cavity and MS. In this paper, we have examined extraction method for airflow as cavity flow. There are some extraction methods of vortex regions, for example, Lambda2, Q-criterion and vorticity magnitude, etc [3]. And we have extracted vortex regions from results of simulation in nasal cavity. In the result, we can see the two vortexes in MS, and we confirmed it was right and left of MS and the result was different. The shape of MS is right and left and slightly different. However, the pattern of the flow became a different clearly.

Development of blood flow and blood wall interaction simulation system using voxel data from medical images

The ability to accurately simulate human tissues will be a valuable tool to help in medical diagnosis and treatment. Generally, complex unstructured grids are used because of human tissue simulations must deal with kinetic interactions between fluids and structures. Additionally, a lot of advanced knowledge and much time are necessary to make computational grids because human tissues shapes are complex and widely varied. As tool capable of reducing the time and the trouble required for making a grid, we have developed blood flow and blood vessel interaction simulation system, which directly uses images from medical diagnosis instruments for constructing a computational grid. We developed a technique to compute wall shear stress, which is important in blood vessel diseases, and we confirmed that our system reliably computes the wall shear stress. In this paper, we present an outline of the system, some ideas, and give some examples of computing the wall shear stress and the behavior of blood flow and blood vessel interaction