Supercomputer Systems Research Articles

Пути организации параллельного доступа к структурированным данным

This paper explores ways to achieve the highest possible exchange performance with files containing structured data. The research was carried out on file systems with supercomputer systems parallel access designed to solve problems of physical and mathematical modeling of various processes and objects. For example, parallel access to raw data is considered using the Lustre file system. The article suggests a way to organize parallel access to structured data based on a specially developed PSIO storage format and the psio access library. A comparative analysis of the I/O performance of the developed data storage format and the HDF5 parallel version format is performed.

Влияние концевого вихря крыла-генератора на обтекание основного крыла под углом атаки

The effect of a tip vortex from an upstream generator wing on the flow around the main wing in a supersonic flow is considered. Numerical results were obtained on the multiprocessor hybrid supercomputer system K-60 at the Keldysh Institute of Applied Mathematics RAS.

Scalability and efficiency challenges for the exascale supercomputing system: practice of a parallel supporting environment on the Sunway exascale prototype system

Scalability and efficiency challenges for the exascale supercomputing system: practice of a parallel supporting environment on the Sunway exascale prototype system

On Bipartite Circulant Graph Decompositions Based on Cartesian and Tensor Products with Novel Topologies and Deadlock-Free Routing

Recent developments in commutative algebra, linear algebra, and graph theory allow us to approach various issues in several fields. Circulant graphs now have a wider range of practical uses, including as the foundation for optical networks, discrete cellular neural networks, small-world networks, models of chemical reactions, supercomputing and multiprocessor systems. Herein, we are concerned with the decompositions of the bipartite circulant graphs. We propose the Cartesian and tensor product approaches as helping tools for the decompositions. The proposed approaches enable us to decompose the bipartite circulant graphs into many categories of graphs. We consider the use cases of applying the described theory of bipartite circulant graph decomposition to the problems of finding new topologies and deadlock-free routing in them when building supercomputers and networks-on-chip.

An Analysis of the Status of China’s Digital Economy and Infrastructure

China’s digital economy, ranking second worldwide for many years, has become a major growth engine for the economy and promotes high-quality economic development. The increase of the digital economy size is driven by the industrial digitization and benefits from the construction of digital infrastructure. China leads the world in 5G, supercomputing, and navigation satellite system while also has the fastest-growing market of cloud computing service in the world. In the future, China should focus more on application of digital technology in production process instead of consumption process, narrowing the digital gap in different regions, and attaching importance to both the construction of hardware infrastructure and the development of underlying infrastructure technologies.

Enabling Large Scale Simulations for Particle Accelerators

International high-energy particle physics research centers, like CERN and Fermilab, require excessive studies and simulations to plan for the upcoming upgrades of the world's largest particle accelerators, and the design of future machines given the technological challenges and tight budgetary constraints. The Beam Longitudinal Dynamics (BLonD) simulator suite incorporates the most detailed and complex physics phenomena in the field of longitudinal beam dynamics, required for providing extremely accurate predictions. Modern challenges in beam dynamics dictate for longer, larger and numerous simulation studies to draw meaningful conclusions that will drive the baseline choices for the daily operation of current machines and the design choices of future projects. These studies are extremely time consuming, and would be impractical to perform without a High-Performance Computing oriented simulator framework. In this article, at first, we design and evaluate a highly-optimized distributed version of BLonD. We combine approximate computing techniques, and leverage a dynamic load-balancing scheme to relax synchronization and improve scalability. In addition, we employ GPUs to accelerate the distributed implementation. We evaluate the highly optimized distributed beam longitudinal dynamics simulator in a supercomputing system and demonstrate speedups of more than two orders of magnitude when run on 32 GPU platforms, w.r.t. the previous state-of-art. By driving a wide range of new studies, the proposed high performance beam longitudinal dynamics simulator forms an invaluable tool for accelerator physicists.

Hierarchical Model of Architecture of Supercomputer Systems for Comparison and Ranking

The task of comparing the capabilities of computing systems with each other and forming various ratings has many possible goals. Here, there is the identification of trends, the promotion of proven general-purpose architectures, and the demonstration of superiority in a certain class of tasks, etc. It is, of course, not enough to describe the achieved performance for all these purposes, various rankings and comparisons use different levels of abstraction and generalization up to that level, which would allow to associate the identified performance indicators with certain features of the system. In practice, descriptions of the architectural peculiarities of systems in ratings are rather scarce, and the authors of the work solve the problem of development a formal description of computer systems of a relatively high level, which, at the same time, would allow to increase the required level of detail, corresponding to the goals of applied research. Such a hierarchical system description model has been proposed and tested on well-known systems from the Top50 and Top500 lists.

Molecular Dynamics Study of Adsorption Phenomenon of Aromatic Hydrocarbon Ionomer in Catalytic Layer of Polymer Electrolyte Fuel Cell

In response to the current global energy crisis and environmental problems, fuel cells are an optional solution. Among them, polymer electrolyte fuel cell (PEFC) has high expectations because of its low operating temperature and suitability for applications in households and mobility devices. Since PEFC requires the use of proton exchange membranes, the transmission efficiency and durability of the exchange membrane is a major concern. To improve the proton transport efficiency, which is generally required, a catalytic layer is integrated into the PEFC. The catalyst layer of PEFC generally contains platinum catalyst covered with the ionomer membrane. Proton conductivity increases with ionomer thickness, while oxygen permeability decreases with ionomer thickness. Therefore, in order to improve the power generation efficiency of PEFC, it is important to design the catalyst layer. In addition, the fluorinated ionomer membrane material (such as Nafion) used for PEFC has good proton conductivity and chemical stability, but the synthesis cost is relatively high. As a consequence, new materials such as composite materials and non-fluorine-based materials have received a lot of attention in recent years. In this study, we focus on a new ionomer material of aromatic hydrocarbon (named SPP-BP) and elucidate the adsorption state of the ionomer on the platinum surface on a nanoscale using molecular dynamics (MD) simulations. Then, the structural properties and water diffusion characteristics of the ionomers of new material (SPP-BP) and conventional Nafion were analyzed based on the MD simulation results.The degree of polymerization (m, p, s) of the monomers were given as 20, 5, and 23 (molecular weight 7.38 kDa), respectively, based on the reported experimental values of SPP-BP. For the polymer composition, molecular models of a random type (r-type), which shows a randomness of monomers, were used. A molecular model of Nafion was also created as the conventional fluorinated membrane materials for comparison with the SPP-BP. The platinum surface was built with a three layers of platinum atoms in a face-centered cubic lattice structure. Regarding the calculation potentials, the DREIDING forcefield was used for the interactions with the molecules. While for the intermolecular potential, the Spohr potential was used for Pt-H2O and Pt-H3O+ interactions, and the others were applied with the Lennard Jones (LJ). The parameters of those potentials were based on the previous studies.In the ionomer membrane, water content is an important indicator that affects the transfer of protons and substances. Based on the water absorption results obtained in the experiment, the water content of SPP-BP was calculated to be between 1.8 and 7.2. Therefore, the water content during the MD simulation was used 3, 5, and 7. The computational size was 55.5 Å × 52.9 Å in the transverse direction of the ionomer membrane (X and Y directions), and the thickness of the membrane adsorbed on the platinum surface was about 50-60 Å. A vacuum region of about 40 Å was set at the top of the system. To calculate the diffusion coefficient and the cluster character of water, multiple simulations with different initial structure were considered. Furthermore, in order to compare to the Nafion, the number of polymers and the computational size was adjusted for containing the same water content.Based on the simulation results, the density distribution of water, hydrophilic sulfo groups, and main body of polymer was analyzed. By the analysis of density distribution, the structure of the ionomer at the interface-vacuum, near the Pt surface and inside was studied. In addition, the orientation of the sulfo group of SPP-BP and Nafion near the Pt surface was analyzed. Besides, the water cluster characteristics and water diffusion properties of the ionomers of SPP-BP and Nafion were also investigated.Acknowledgements: This work was supported by the New Energy and Industrial Technology Development Organization (NEDO) of Japan through the fund for ECCEED'30 Projects (No. JPNP20003). Numerical simulations were performed on the Supercomputer system "AFI-NITY" at the Advanced Fluid Information Research Center, Institute of Fluid Science, Tohoku University.

Molecular Dynamics Simulations of Cerium Ion Transport Phenomena in Polymer Electrolyte Membranes of Polymer Electrolyte Fuel Cells

As energy demand increases, global warming progresses, and energy resources are scarce in the future, expectations for fuel cells, which generate electricity through a chemical reaction between hydrogen and oxygen, are rising. There are various types of fuel cells, which are classified according to the electrolyte. This study deals with polymer electrolyte fuel cells (PEFCs), which are used in automobiles and household fuel cells. Currently, there are two main challenges for the practical application of PEFCs: durability and cost reduction. The target value for durability is required to be 40,000 h or more. However, the current durability is about 10,000 h, so an investigation into the causes of deterioration and countermeasures are urgently needed. One of the causes of degradation is the chemical deterioration of polymeric membranes. When the by-product hydrogen peroxide comes into contact with impurities such as iron ions, hydroxyl radicals (・OH) are generated, which attack and decompose the polymer membrane. In order to suppress this degradation, research have been conducted to add a substance that renders hydroxyl radicals inactive (radical scavenger) to polymer membranes, and this has been put to practical use. One of the most useful radical scavengers is cerium ions. However, it has been reported that cerium ions migrate in the electrolyte membrane, resulting in non-uniform distribution in the proton exchange membranes (PEMs), and in places where the concentration is low, and degradation progresses. The addition of too much radical scavenger hinders proton transport in the electrolyte membrane, reducing the output capacity of the PEFC. Therefore, understanding the cerium ions transport mechanism in the PEM is important to control cerium ions migration and improve the durability of the electrolyte membrane to maintain its performance as a fuel cell. However, the cost of operating experiments over 40,000 h is enormous, and it is difficult to analyze the phenomena occurring inside nanostructured electrolyte membranes by experiment, so simulation analysis is required.In this study, molecular dynamics simulations have been performed for analysis of structure and cerium ions transport properties in PEM, consisting of polymer chains, cerium ions, water molecules, and hydronium ions. In all simulations, Nafion chain, which has the chemical structure with EW=1114, has been employed. Water content λ, which indicates the ratio of solvent molecules to sulfonic groups in Nafion, was changed to 4, 9, 14.Fig. 1 shows the self-diffusion coefficient of cerium ions at a cerium ions content of 1.1 wt% (20 pcs added) as a function of temperature for each water content, indicating that the coefficient increases with increasing temperature for all water contents. In addition, the increase in the self-diffusion coefficient of cerium ions with temperature is small at low water content (λ = 4), and the rate of increase in the self-diffusion coefficient of cerium ions with temperature becomes larger as the water content increases. From the radial distribution functions (RDFs) of cerium ions around the sulfone group, cerium ions are distributed at about 4.15 Å around the sulfone group at low water content, while at high water content they are distributed at about 6.0 Å. From the RDF of water molecules around the sulfone group, water molecules are located at about 4 Å around the sulfone group. The results indicate that cerium ions coordinate directly to the sulfone groups at low water content because fewer water molecules collect around the sulfone groups. Conversely, at high water content, water molecules cover the sulfone groups, and cerium ions coordinate outside the sulfone groups, which lead to the increase of the diffusion coefficient.Acknowledgments This work was supported by the New Energy and Industrial Technology Development Organization (NEDO) of Japan, Grant number JPNP20003. It was performed on the Supercomputer system “AFI-NITY” at the Advanced Fluid Information Research Center, Institute of Fluid Science, Tohoku University. Figure 1

Molecular Dynamics Simulation of Scattering and Surface Diffusion of Oxygen Molecules on Ionomers in Catalyst Layers of PEFCs

Polymer electrolyte fuel cells (PEFCs) are used in fuel cell vehicles and households stationary power sources because of their low operating temperatures, small size, and light weight. For further development of PEFCs, it is necessary to reduce the amount of platinum used in the catalyst layers. If the amount of platinum used in the catalyst layers is reduced, the current density per mass of platinum increases. The effect of diffusion polarization then becomes dominant in the overall voltage drop. One factor that causes diffusion polarization is the transport resistance of oxygen molecules in the catalyst layers. Therefore, an understanding of the transport of oxygen molecules in the catalyst layers is necessary to improve the performance of PEFCs under high current density conditions. The catalyst layers comprise porous structures by aggregation of ionomer-coated platinum-supported carbon. The Knudsen number is important for gas diffusion phenomena in porous structures. The Knudsen number is the ratio of the representative length of the system to the mean free path of gas molecules, and when this value is sufficiently small, the gas diffusion phenomenon can be regarded as molecular diffusion. When the Knudsen number is sufficiently large, the gas diffusion phenomenon becomes Knudsen diffusion, in which collisions with solid walls are more dominant than collisions between gas molecules. The Knudsen number is sufficiently large in the porous structure of the catalyst layer. Therefore, the effect of collision with solid surfaces is important in the oxygen transport in the catalyst layer. In previous studies, it has been clarified that the behavior of oxygen molecules incident on an ionomer thin film includes surface scattering, in which diffuse reflection occurs on the surface, and surface diffusion, in which oxygen molecules trapped on the surface move across the surface in a series of multiple collisions. However, the effect of surface diffusion on oxygen transport across the catalyst layer is not clear. The purpose of this study is to analyze the effect of oxygen molecules and ionomer conditions on surface diffusion using the MD simulation. We analyze the behavior of oxygen molecules impinging on ionomers in a computational system that reproduces the pores of the catalyst layer. The simulation system consists of polymer chains, solvent molecules (water molecules and hydronium ions), and oxygen molecules. Periodic boundary conditions were applied in the x-, y-, and z-directions. The behavior of oxygen molecules after they impact the ionomer wall such as their residence time during surface diffusion and the surface diffusion coefficient will be analyzed.AcknowledgmentsThis work was supported by the New Energy and Industrial Technology Development Organization (NEDO) of Japan, Grant number JPNP20003. It was performed on the Supercomputer system “AFI-NITY” at the Advanced Fluid Information Research Center, Institute of Fluid Science, Tohoku University. Figure 1

(Digital Presentation) A Coarse-Grained Molecular Dynamics Study on the Aggregation and Adsorption Behavior of Ionomer from Solution Onto Pt/C Substrate

The catalyst layer (CL) was an important component in polymer electrolyte fuel cells (PEFCs). In the CL, Nafion ionomers coated the Pt particles on the carbon surface for the proton transfer assistance, and therefore the structure optimization of ionomer films is very important for the improvement of catalytic utilization and proton transport performance. The microscopic structure of ionomer covered carbon and Pt particles in the solution was related to the final structural characteristics of the ionomer film determining the performance of CL. In order to elucidate the structural evolution and drying process of the catalyst ink, in the present study, we carried out coarse-grained molecular dynamics (CGMD) simulations to investigate the adsorption behavior of ionomers and morphology evolution of Nafion ionomer from the aqueous solutions to the Pt/C substrate surface under various solution conditions including ionomer concentration and I/C ratio, and substrate conditions including the size and distribution of Pt particles, and Pt/C ratio. We found that the ionomers formed a cylindrical aggregate in the solution and are easily attracted by Pt particles due to their hydrophilic ionic shells. Although ionomer aggregates are easily to adsorb onto Pt particles, it tended to unfold and spread onto the carbon substrate finally, which is resulted from the hydrophobic attraction of the exposed ionomer backbone to the carbon surface and the hydrophilic attraction of ionomer sulfonate groups to the aqueous solution. The ionomer adsorption displayed a strong correlation with the ionomer concentration in the dilute solution. However, the I/C ratio had little effect on ionomer adsorption. This adsorption phenomenon of ionomers in the dilution solution was found to fit the Langmuir adsorption model very well but failed when the concentration of ionomer became higher. A wettability switching behavior of ionomers is observed because the morphology of ionomers changed from the solution to the surface. Two types of the adsorption behavior of ionomer aggregates in the aqueous solution were observed. The first adsorption behavior was the direct adsorption of ionomer aggregate to the carbon substrate without any interaction with Pt particles. For the ionomer aggregate, exposed apolar backbones of ionomers that were close to the carbon could help the ionomer aggregate adsorb on the hydrophobic carbon surface. During the adsorption process, cylindrical-like shape ionomer aggregates gradually spread onto the carbon surface. Finally, the ionomers were entirely distributed on the carbon surface without aggregating together. The second adsorption behavior was indirect adsorption of ionomer aggregate, where the ionomer aggregates firstly adsorbed onto the Pt particles from solution, and finally to the carbon substrate. In the solution, the apolar backbones of ionomer aggregates located at the core of the aggregate were difficult to contact with the substance outside such as carbon surface, while the anionic groups located at the aggregate-solvent interface were easy to contact with the substance outside. When such Nafion aggregates were close to the Pt particles, the anionic shells of the cylindrical aggregates were easily to adsorb onto hydrophilic Pt particles. When the Nafion aggregate adsorbed onto Pt particles, Nafion still presents a cylinder-like aggregate state until it started to approach the carbon surface. When the Nafion aggregate begin to move onto the carbon substrate, it gradually unfolded due to the strong hydrophobic interaction between the ionomer backbone and the carbon surface. During this process, Nafion ionomers gradually relieved their internal hydrophobic associations between their backbones because the unfolded ionomers created more surface contacts and better anchored themselves to the surface. Therefore, when the exposed carbon surface was large enough, increasing the number and sizes of Pt particles could increase the surface area of Pt particles, which favors the ionomer adsorption. When the space between Pt particles was too small to create enough carbon surface contacts anchoring the ionomer backbone, the coverage of the ionomer dropped rapidly. The present study provided insight into ionomer adsorption from solutions to Pt/C surface, which could contribute to the catalyst ink design to improve the structure of catalyst layers.AcknowledgmentsThis work was supported by the New Energy and Industrial Technology Development Organization (NEDO) of Japan, Grant number JPNP20003. It was performed on the Supercomputer system “AFI-NITY” at the Advanced Fluid Information Research Center, Institute of Fluid Science, Tohoku University. We also thank W. Shirakawa for her assistance with the simulations.

Parallel fully coupled methods for bound-preserving solution of subsurface flow and transport in porous media

As more powerful supercomputer systems with lots of memory and a large number of computing cores become available, the family of fully coupled algorithms is drawing more attention in scientific and engineering applications, due to its impressive robustness and scalability in extreme-scale simulations. In this work, we introduce and study some parallel domain decomposition preconditioned generalized Newton algorithms for solving the fully coupled and bound-preserving formulation of the subsurface flow and transport problem in porous media. In the approach, we present the active–set reduced–space (ASRS) method to guarantee the nonlinear consistency of the fully coupled system in a monolithic way, and meanwhile ensure the boundedness requirement of the solution. Furthermore, we focus on the application of the overlapping additive Schwarz preconditioning technique to accelerate the linear convergence of Newton iterations and be beneficial to the scalability of the inner linear solver. We present some numerical experiments to demonstrate the parallel scalability of the proposed algorithm on the Shaheen-II supercomputer with thousands of processors. In particular, the numerical results also show that the fully coupled framework has the nature of bound preservation and is efficient for the proposed reservoir flow problems.

A refined parallel stacking InSAR workflow for large-scale deformation fast extraction—a case study of Tibet

The Tibet Autonomous Region has a complex topography, and there are many difficulties in extracting its large-scale surface deformation. Traditional time-series Synthetic Aperture Radar Interferometry (InSAR) approaches are time-consuming and difficult to handle the massive growth of data in a timely manner. In this paper, a refined parallel stacking InSAR workflow for large-scale deformation fast extraction is proposed. The burst images of Sentinel-1 data were first extracted, and a series of procedures were introduced to improve accuracy, including using a short spatial baseline to mitigate topographic error, using a polynomial to fit and remove the small-scale atmospheric error, and using the joint polarization stacking method to reduce random noise. The workflow was deployed to a supercomputing system for parallel processing to improve efficiency. Using 1638 Sentinel-1 acquisitions, the surface deformation rate of the entire Tibet Autonomous Region was obtained and in good agreement with the Global Positioning System (GPS) data, with a Root Mean Squared Error (RMSE) of 3.27 mm/year, indicating high accuracy.

Possible existence of chiral and multiple chiral nuclei in thallium isotopes * *This work is partly support by the Major Program of Natural Science Foundation of Shandong Province (ZR2020ZD30), the Outstanding Youth Fund of Natural Science Foundation of Shandong Province (ZR2020YQ07), the National Natural Science Foundation of China (11675094, 12075138), and the Young Scholars Program of Shandong University, Weihai. The

The chirality in thallium isotopes is investigated using the adiabatic and configuration-fixed constrained triaxial relativistic mean field theory. Several minima with prominent triaxial deformation and proper configuration, where the chiral doublet bands may appear, are obtained in odd-odd nuclei Tl and odd-mass nuclei Tl. Furthermore, the possible existence of multiple chiral doublet bands (MχD) is demonstrated in Tl. As the chiral doublet bands in Tl and MχD in 195Tl have been observed experimentally, further experimental exploration for the chirality in Tl and MχD in thallium isotopes is expected to verify the predictions.

Fast and accurate variable batch size convolution neural network training on large scale distributed systems

AbstractLarge‐scale distributed convolution neural network (CNN) training brings two performance challenges: model performance and system performance. Large batch size usually leads to model test accuracy loss, which counteracts the benefits of parallel SGD. The existing solutions require massive hyperparameter hand‐tuning. To overcome this difficult, we analyze the training process and find that earlier training stages are more sensitive to batch size. Accordingly, we assert that different stages should use different batch size, and propose a variable batch size strategy. In order to remain high test accuracy under larger batch size cases, we design an auto‐tuning engine for automatic parameter tuning in the proposed variable batch size strategy. Furthermore, we develop a dataflow implementation approach to achieve the high‐throughput CNN training on supercomputer system. Our approach has achieved high generalization performance on SOAT CNN networks. For the ShuffleNet, ResNet‐50, and ResNet‐101 training with ImageNet‐1K dataset, we scale the batch size to 120 K without accuracy loss and to 128 K with only a slight loss. And the dataflow implementation approach achieves 93.5% scaling efficiency on 1024 GPUs compared with the state‐of‐the‐art.

Heterogeneous Parallel Implementation of Large-Scale Numerical Simulation of Saint-Venant Equations

Large-scale floods are one of the major events that impact the national economy and people’s livelihood every year during the flood season. Predicting the factors of flood evolution is a worldwide problem. We use the two-dimensional Saint-Venant equations as an example and for high-performance computing in modelling the flood behavior. Discretization of the two-dimensional Saint-Venant equations with initial and boundary conditions with the finite difference method in the explicit leapfrog scheme is carried out. Afterwards, we employed a large-scale heterogeneous parallel solution on the “SunRising-1” supercomputer system using MPI, OpenMP, Pthread, and OpenCL runtime libraries. On this basis, we applied communication/calculation overlapping and the local memory acceleration to optimize the performance. Finally, various performance tests of the parallel scheme are carried out from different perspectives. We have found this method is efficient and recommend this approach be used in solving systems of partial differential equations similar to the Saint-Venant equations.

Design and Reliability Evaluation of Switching Element for Improving the QoS of Multistage Interconnection Network

Reliable and cost effective multistage interconnection networks (MINs) are the backbone of efficient distributed computing in supercomputer systems. In this paper new 3×3, 4×2, and 4×4 switching element (SE) structures are introduced which can be employed in various existing MINs to provide enhanced reliability at much reduced cost. The reliability function of proposed as well as the existing SEs have been formulated on basis of which their reliability values have been computed. The computed reliability values of all SEs (existing and proposed ones) have been compared and analyzed. The achieved results clearly depicts that the proposed SEs possess higher reliability and can replace the existing SEs of similar size without changing the existing design of MIN to improve reliability.

TianheGraph: Customizing Graph Search for Graph500 on Tianhe Supercomputer

As the era of exascale supercomputing is coming, it is vital for next-generation supercomputers to find appropriate applications with high social and economic benefit. In recent years, it has been widely accepted that extremely-large graph computation is a promising killer application for supercomputing. Although Tianhe series supercomputers are leading in the world-wide competition of supercomputing (ranked No. 1 in the Top500 list for six times), previously they had been inefficient in graph computation according to the Graph500 list. This is mainly because the previous graph processing system cannot leverage the advanced hardware features of Tianhe supercomputers. To address the problem, in this paper we present our integrated optimizations for improving the graph computation performance on our next-generation Tianhe supercomputing system, mainly including sorting with buffering for heavy vertices, vectorized searching with SVE (Scalable Vector Extension) on matrix2000+ CPUs, and group communication on the proprietary interconnection network. Performance evaluation on a subset of the Tianhe supercomputer (with 512 nodes and 196,608 cores) shows that our customized graph processing system effectively improves the graph search performance and achieves the BFS performance of 2131.98 GTEPS.

Optimized realization of Quantum Fourier Transform for domestic DCU accelerator

Abstract Quantum Fourier Transform (QFT) is a key part of quantum algorithms such as Shor algorithm, Grover algorithm, and quantum phase estimation algorithm. The “Songshan” supercomputer system adopts the heterogeneous architecture of domestic CPU+DCU. The simulation implementation of the quantum Fourier transform on the “Songshan” supercomputer can promote the research and optimization of related quantum algorithms. Based on the HIP heterogeneous programming model, this paper first maps the phase shift transformation R gate operation of the QFT transformation to the DCU accelerator, and initially constructs the heterogeneous implementation of QFT on the “Songshan” supercomputer; then, the Hadamard gate transformation operation originally implemented on the host side was implemented on the DCU, which greatly reduces the communication overhead caused by the quantum information data transmission between the CPU and the DCU device; finally, the quantum logic gate is analyzed, and based on the operation of the phase shift transformation R gate on the probability amplitude information in the simulation algorithm, an optimization scheme to improve the activity of the DCU thread is proposed. During the experimental verification process, the realized 28-qubit quantum Fourier transform optimization has achieved 13.440 speedup compared with the serially realized QFT simulation on the CPU. The basic quantum gate optimization method based on DCU acceleration equipment in this paper can provide a reference for the efficient simulation of other quantum algorithms on domestic heterogeneous platforms.

MPI-GWAS: a supercomputing-aided permutation approach for genomewide association studies.

Permutation testing is a robust and popular approach for significance testing in genomic research that has the advantage of reducing inflated type 1 error rates; however, its computational cost is notorious in genome-wide association studies (GWAS). Here, we developed a supercomputing-aided approach to accelerate the permutation testing for GWAS, based on the message-passing interface (MPI) on parallel computing architecture. Our application, called MPI-GWAS, conducts MPI-based permutation testing using a parallel computing approach with our supercomputing system, Nurion (8,305 compute nodes, and 563,740 central processing units [CPUs]). For 107 permutations of one locus in MPI-GWAS, it was calculated in 600 s using 2,720 CPU cores. For 107 permutations of ~30,000–50,000 loci in over 7,000 subjects, the total elapsed time was ~4 days in the Nurion supercomputer. Thus, MPI-GWAS enables us to feasibly compute the permutation-based GWAS within a reason-able time by harnessing the power of parallel computing resources.