基于Optix库的GPU端流场数据本地可视化研究

工业软件是工业发展走向数字化时代、转向智能制造的基石，是工业走向现代化的灵魂，是工业变革的核心动力。可视化作为工业软件的重要组成部分，发挥着重要的作用。

  可视化可以分为离线可视化（Offline Visualization）和原位可视化（In-situ Visualization）两种。前者是常见的可视化方式，先将数据存储在磁盘上，再在计算时从磁盘读取数据进行可视化渲染；后者直接在芯片上进行处理，省去了存储的开销，可以更快获得可视化结果，并且in-situ可视化可以突破存储限制，完成高时空分辨率的可视化。随着计算机的不断发展，越来越多的高性能计算依赖GPU（Graphics Processing Unit）。把GPU数据传输到CPU（Central Processing Unit）再保存到磁盘上做传统的离线可视化的方法已经极大拖慢GPU高性能计算速度，离线可视化已经不能满足使用者的需求。同时GPU本就是图形处理器，擅长处理可视化渲染计算，在可视化渲染方面的计算效率远比CPU的可视化渲染计算速率更快。因此对数据在GPU上进行in-situ可视化的需求愈发强烈。

  本文利用NVIDIA的函数库Optix，在GPU上利用CUDA（Compute Unified Device Architecture）和C++编程语言混合编程，设计并开发可以在GPU上对流场数据进行in-situ可视化体渲染的程序模块。利用计算流场时的网格作为体素的划分，将网格上的标量数据作为体素的信息。上述做法既节省将数据划分为体素的时间，也节省数据空间位置转换的开销。从视角位置射出光线，光线依次穿过流体网格，根据流场数据与体素信息之间的关系，计算网格处的体素信息，并对光线所有穿过网格处的体素信息进行积分计算。为了更好的显示数据的关系，将网格上的标量数据进行归一化，再将归一化后的数据映射到色带和透明度函数上，通过对色带和透明度函数的调整可视化渲染不同的重点。该程序模块可以对GPU上得到的流场数据直接在GPU上进行可视化渲染，充分利用GPU图形渲染方面的优势，而且由于程序不用经过CPU存储到磁盘，节省数据传输的时间，极大加快可视化速率，更适合在高性能计算机集群上对大规模流场数据使用，缓解大量流场数据可视化耗时的问题。in-situ可视化渲染得出的结果与成熟离线可视化软件渲染的结果效果相当。测试不同用例时，in-situ可视化渲染花费时间远小于离线可视化所需时间，满足大规模流场的可视化需求。

Industrial software is the cornerstone of industrial development in the digital age and intelligent manufacturing, the soul of industrial modernization and the core driving force of industrial transformation. As an important part of industrial software, visualization plays an important role.

  Visualization of flow field can be divided into offline visualization and in-situ visualization. The former is a common visualization method, which stores the data on the disk first, and then reads the data from the disk for visual rendering during calculation; the latter is processed directly on the chip, which saves the storage cost and can obtain the visualization result faster. In-situ visualization can break through the storage limitations and complete the visualization at high spatio-temporal resolution. With the continuous development of computers, more and more high-performance computing relies on GPU. The traditional method of transferring GPU data to CPU and then saving it to disk for offline visualization has greatly slowed down the high-performance computing speed of GPUs, and offline visualization can no longer meet the needs of its users. At the same time, the GPU is a graphics processor, which is good at processing visual rendering calculation, and its calculation efficiency in visual rendering is much faster than that of the CPU. Therefore, there is a growing demand for in-situ visualization of data on GPUs.

  In this paper, NVIDIA's function library Optix is used, and CUDA and C++ programming languages are mixed on the GPU to design and develop a program module that can render in-situ visualization of flow field data on the GPU. The grid when calculating the flow field is used as the division of voxels, and the scalar data on the grid is used as the information of voxels. The above method not only saves the time of dividing data into voxels, but also saves the overhead of data space position transformation. Light rays are emitted from the visual angle and pass through the fluid grid in turn. According to the relationship between flow field data and voxel information, the voxel information at the grid is calculated, and the voxel information at all the places where the light rays pass through the grid is integrated. In order to better show the relationship between data, the scalar data on the grid is normalized, and then the normalized data is mapped to the color band and transparency function, and different key points are visually rendered by adjusting the color band and transparency function. This program module can directly visualize the flow field data obtained from GPU, making full use of the advantages of GPU graphics rendering. Moreover, because the program does not need to be stored on disk through CPU, it saves the time of data transmission and greatly speeds up the visualization rate. It is more suitable for the use of large-scale flow field data in high-performance computer clusters, and relieves the time-consuming problem of visualization of a large number of flow field data. In-situ visual rendering results are equivalent to those rendered by mature offline visualization software. When testing different use cases, in-situ visualization rendering takes much less time than offline visualization, which meets the visualization requirements of large-scale flow fields.

[1] Altair Engineering Inc. CFD 仿真解决方案 | Altair[EB/OL]. 2023
[2023-12-06]. https://ww w.altair.com.cn/altair-cfd/.
[2] TECHNIA. PowerFLOW|TECHNIA[EB/OL]. 2023
[2023-12-06]. https://www.technia.com/ software/powerflow/.
[3] ARTHUR A. Some Techniques for Shading Machine Renderings of Solids[C]//Proceedings of the April 30–May 2, 1968, Spring Joint Computer Conference. New York, NY, USA: Association for Computing Machinery, 1968: 37–45.
[4] WHITTED T. An Improved Illumination Model for Shaded Display[J]. Commun. ACM, 1980, 23(6): 343–349.
[5] KAJIYA J T. The Rendering Equation[J]. SIGGRAPH Comput. Graph., 1986, 20(4): 143–150.
[6] VEACH E. Robust monte carlo methods for light transport simulation[D]. Stanford, CA, USA: Stanford University, 1998.
[7] FEDKIW R, STAM J, JENSEN H W. Visual Simulation of Smoke[C]//Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques. New York, NY, USA: Association for Computing Machinery, 2001: 15–22.
[8] JENSEN H W, MARSCHNER S R, LEVOY M, et al. A Practical Model for Subsurface Light Transport[C]//Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques. New York, NY, USA: Association for Computing Machinery, 2001: 511– 518.
[9] JAROSZ W. Efficient Monte Carlo Methods for Light Transport in Scattering Media[D]. USA: University of California at San Diego, 2008.
[10] KETTUNEN M, MANZI M, AITTALA M, et al. Gradient-Domain Path Tracing[J]. ACMTrans. Graph., 2015, 34(4).
[11] 郭海波. 实时渲染技术研究综述[J]. 电脑编程技巧与维护, 2023(1): 137-139.
[12] 闫润, 黄立波, 郭辉, 等. 实时光线追踪相关研究综述[J]. 计算机科学与探索, 2023, 17(263-278).
[13] PAN X, MAI J, JIANG X, et al. Predicting Loose-Fitting Garment Deformations Using Bone-Driven Motion Networks[C]//ACM SIGGRAPH 2022 Conference Proceedings. New York, NY, USA: Association for Computing Machinery, 2022.
[14] XIONG S, WANG Z, WANG M, et al. A Clebsch Method for Free-Surface Vortical Flow Simulation[J]. ACM Trans. Graph., 2022, 41.
[15] 陈川, 任强, 邓金华. 基于光线追踪的可视性分析研究与模块开发[J]. 信息技术与网络安全, 2021, 40(26-30).
[16] 彭博, 汪强, 青芮冰, 等. 基于光线追踪的实时超声模拟与虚拟现实的集成[J]. 系统仿真学报, 2022, 34(11): 2425-2436.
[17] 周升腾. 小规模流体动画仿真的实时渲染研究[D]. 燕山大学, 2019.
[18] LI W, CHEN Y, DESBRUN M, et al. Fast and Scalable Turbulent Flow Simulation with TwoWay Coupling[J]. ACM Trans. Graph., 2020, 39(4).
[19] LI W, LIU D, DESBRUN M, et al. Kinetic-Based Multiphase Flow Simulation[J]. IEEE Transactions on Visualization and Computer Graphics, 2021, 27(7): 3318-3334.
[20] BAI K, WANG C, DESBRUN M, et al. Predicting High-Resolution Turbulence Details in Space and Time[J]. ACM Trans. Graph., 2021, 40(6).
[21] LYU C, LI W, DESBRUN M, et al. Fast and Versatile Fluid-Solid Coupling for Turbulent Flow Simulation[J]. ACM Trans. Graph., 2021, 40(6).
[22] 顾恩祖. 虚拟手术系统中血流场景渲染方法研究与程序实现[D]. 哈尔滨理工大学, 2022.
[23] MARCHESIN S, CHEN C K, HO C, et al. View-Dependent Streamlines for 3D Vector Fields [J]. IEEE Transactions on Visualization and Computer Graphics, 2010, 16(6): 1578-1586.
[24] LIU D, XIONG C, LIU X. Vectorizing Quantum Turbulence Vortex-Core Lines for Real-Time Visualization[J]. IEEE Transactions on Visualization and Computer Graphics, 2021, 27(9):3794-3807.
[25] 肖祥云, 杨旭波. 基于物理及数据驱动的流体动画研究[J]. 软件学报, 2020, 31(3251-3265).
[26] ZHANG Y, BAN X, DU F, et al. FluidsNet: End-to-end learning for Lagrangian fluid simulation [J]. Expert Systems with Applications, 2020, 152: 113410.
[27] BAI K, LI W, DESBRUN M, et al. Dynamic Upsampling of Smoke through Dictionary-Based Learning[J]. ACM Trans. Graph., 2020, 40(1).
[28] kitware, inc. ParaView - Open-source, multi-platform data analysis and visualization application[EB/OL]. 2023
[2023-12-13]. https://www.paraview.org/.
[29] SideFX. Houdini - 3D modeling, animation, VFX, look development, lighting and rendering [EB/OL]. 2023
[2023-12-13]. https://www.sidefx.com/.
[30] Tecplot, Inc. Tecplot Visualization and Analysis Tools for CFD Post-processing[EB/OL]. 2023
[2023-12-13]. https://tecplot.com/.
[31] Unity Technologies. Unity 实时开发平台 13D、2D、VR 和 AR 引擎[EB/OL]. 2023
[2023-12-15]. https://unity.com/.
[32] Epic Games, Inc. 虚幻引擎 | 最强大的实时 3D 创作平台- Unreal Engine[EB/OL]. 2023
[2023-12-15]. https://www.unrealengine.com/.
[33] 2010 Wenzel Jakob. Mitsuba - physically based renderer[EB/OL]. 2023
[2023-12-15]. https: //www.mitsuba-renderer.org/.
[34] BNà S, COLOMBO A, CRIVELLINI A, et al. In situ visualization for high-fidelity CFD— Case studies[J]. Computers & Fluids, 2023, 267: 106066.
[35] LEHMANN M. Computational study of microplastic transport at the water-air interface with a memory-optimized lattice Boltzmann method[D]. Bayreuth, 2023.
[36] LEHMANN M. Esoteric Pull and Esoteric Push: Two Simple In-Place Streaming Schemes for the Lattice Boltzmann Method on GPUs[J]. Computation, 2022, 10(6).
[37] LEHMANN M, KRAUSE M, AMATI G, et al. Accuracy and performance of the lattice Boltzmann method with 64-bit, 32-bit, and customized 16-bit number formats[J]. Physical Review E, 2022, 106.
[38] LEHMANN M. Combined scientific CFD simulation and interactive raytracing with OpenCL [C]//IWOCL ’22: Proceedings of the 10th International Workshop on OpenCL. New York, NY, USA: Association for Computing Machinery, 2022.
[39] LEHMANN M. High Performance Free Surface LBM on GPUs[Z]. 2021.
[40] LEHMANN M, GEKLE S. Analytic Solution to the Piecewise Linear Interface Construction Problem and Its Application in Curvature Calculation for Volume-of-Fluid Simulation Codes [J]. Computation, 2022, 10(2).
[41] F.HUGHES J. 计算机图形学原理及实践[M]. 机械工业出版社, 2022.
[42] NICODEMUS F E. Directional Reflectance and Emissivity of an Opaque Surface[J]. Appl. Opt., 1965, 4(7): 767-775.
[43] BARTELL F O, DERENIAK E L, WOLFE W L. The theory and measurement of bidirectional reflectance distribution function (BRDF) and bidirectional transmittance distribution function (BTDF)[C]//Radiation scattering in optical systems: Vol. 257. SPIE, 1981: 154-160.
[44] IMMEL D S, COHEN M F, GREENBERG D P. A Radiosity Method for Non-Diffuse Environments[J]. SIGGRAPH Comput. Graph., 1986, 20(4): 133–142.
[45] HECKBERT P S. Adaptive Radiosity Textures for Bidirectional Ray Tracing[J]. SIGGRAPH Comput. Graph., 1990, 24(4): 145–154.
[46] PHONG B T. Illumination for computer generated pictures[J]. Commun. ACM, 1975, 18(6): 311–317.
[47] BLINN J F. Models of light reflection for computer synthesized pictures[J]. SIGGRAPH Comput. Graph., 1977, 11(2): 192–198.
[48] LAFORTUNE E P, FOO S C, TORRANCE K E, et al. Non-linear approximation of reflectance functions[C]//Proceedings of the 24th annual conference on Computer graphics and interactive techniques. 1997: 117-126.
[49] TORRANCE K E, SPARROW E M. Theory for off-specular reflection from roughened surfaces [J]. Josa, 1967, 57(9): 1105-1114.
[50] COOK R L, TORRANCE K E. A Reflectance Model for Computer Graphics[J]. ACM Trans.Graph., 1982, 1(1): 7–24.
[51] OREN M, NAYAR S K. Generalization of Lambert’s reflectance model[C]//SIGGRAPH ’94: Proceedings of the 21st Annual Conference on Computer Graphics and Interactive Techniques.New York, NY, USA: Association for Computing Machinery, 1994: 239–246.
[52] BECKMANN P. Progress in Optics: Vol. 6 II Scattering of Light by Rough Surfaces[Z]. 1967: 53-69.
[53] LAFORTUNE E, WILLEMS Y. Bi-Directional Path Tracing[J]. Proceedings of Third International Conference on Computational Graphics and Visualization Techniques (Compugraphics’, 1998, 93.
[54] JENSEN H W. Global Illumination using Photon Maps[C]//Rendering Techniques. Vienna: Springer Vienna, 1996: 21-30.
[55] HE S, WONG H C, WONG U H. An Efficient Adaptive Vortex Particle Method for RealTime Smoke Simulation[C]//2011 12th International Conference on Computer-Aided Design and Computer Graphics. 2011: 317-324.
[56] RUDOLF M, RACZKOWSKI J. Modeling the Motion of Dense Smoke in the Wind Field[J].Comput. Graph. Forum, 2000, 19.
[57] DREBIN R A, CARPENTER L, HANRAHAN P. Volume rendering[J]. SIGGRAPH Comput.Graph., 1988, 22(4): 65–74.
[58] LEVOY M. Display of surfaces from volume data[J]. IEEE Computer Graphics and Applications, 1988, 8(3): 29-37.
[59] CHARLTON T. LEWIS C S. Lewis & Short Latin Dictionary[EB/OL]. 2023
[2023-12-06]. ht tps://www.perseus.tufts.edu/hopper/text?doc=Perseus%3Atext%3A1999.04.0059%3Aentry% 3Dsitus2.
[60] MUNRO J I, RAMAN V, SALOWE J S. Stable in situ sorting and minimum data movement[J]. BIT, 1990, 30(2): 220-234.
[61] ZHOU T, FANG L, YAN T, et al. In situ optical backpropagation training of diffractive optical neural networks[J]. Photon. Res., 2020, 8(6): 940-953.
[62] 弯港. 基于格子 Boltzmann 方法的流动控制机理数值研究[D]. 南京理工大学, 2013.
[63] 张辛健. 基于 LBM 算法耦合流场的枝晶相场模型[D]. 兰州理工大学, 2014.
[64] 郭照立, 郑楚光. 格子 Boltzmann 方法的原理及应用[M]. 科学出版社, 2009.
[65] 贾思昊. 基于 LBM 的多孔介质中 CO2/咸水两相运移模拟研究[D]. 东北石油大学, 2023.
[66] WOLF-GLADROW D. Lattice-Gas Cellular Automata and Lattice Boltzmann Models - An Introduction[J]. Lattice-Gas Cellular Automata and Lattice Boltzmann Models, 2000, 1725.
[67] QIAN Y H, D’HUMIèRES D, LALLEMAND P. Lattice BGK Models for Navier-Stokes Equation[J]. Europhysics Letters, 1992, 17(6): 479.
[68] GEIER M, SCHöNHERR M, PASQUALI A, et al. The cumulant lattice Boltzmann equation in three dimensions: Theory and validation[J]. Computers & Mathematics with Applications, 2015, 70(4): 507-547.
[69] SEEGER S, HOFFMANN H. The cumulant method for computational kinetic theory[J]. Continuum Mechanics and Thermodynamics, 2000, 12: 403-421.
[70] 李旭晖, 单肖文, 段文洋. 格子玻尔兹曼正则化碰撞模型的理论进展[J]. 空气动力学学报,2022, 40: 46-64.
[71] BANARI A, GEHRKE M, JANßEN C F, et al. Numerical simulation of nonlinear interactions in a naturally transitional flat plate boundary layer[J]. Computers & Fluids, 2020, 203: 104502.
[72] NISHIMURA S, HAYASHI K, NAKAYE S, et al. Implicit Large-Eddy Simulation of rotating and non-rotating machinery with Cumulant Lattice Boltzmann method aiming for industrial applications[J]. AIAA Aviation 2019 Forum, 2019.
[73] SITOMPUL Y P, AOKI T. A filtered cumulant lattice Boltzmann method for violent two-phase flows[J]. Journal of Computational Physics, 2019, 390: 93-120.
[74] LENZ S, SCHöNHERR M, GEIER M, et al. Towards real-time simulation of turbulent air flow over a resolved urban canopy using the cumulant lattice Boltzmann method on a GPGPU[J].Journal of Wind Engineering and Industrial Aerodynamics, 2019, 189: 151-162.
[75] PERLIN K, HOFFERT E M. Hypertexture[J]. SIGGRAPH Comput. Graph., 1989, 23(3): 253–262.
[76] PARKER S G, BIGLER J, DIETRICH A, et al. OptiX: a general purpose ray tracing engine [C]//SIGGRAPH ’10: ACM SIGGRAPH 2010 Papers. New York, NY, USA: Association for Computing Machinery, 2010.
[77] CORPORATION N. OptiX Programming Guide[M]. NVIDIA Corporation, 2023.