GPU Computing Advances In 3D Electromagnetic Simulation

2y ago
30 Views
2 Downloads
1.71 MB
23 Pages
Last View : 1d ago
Last Download : 2m ago
Upload by : Jewel Payne
Transcription

Session S0069:GPU Computing Advances in 3DElectromagnetic SimulationAndreas Buhr, Alexander Langwost, Fabrizio ZanellaCST (Computer Simulation Technology)CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

AbstractComputer Simulation Technology (CST) has been providing GPU acceleration for its3D Full Wave electromagnetic field simulation tools for several years. The latestversion of CST Studio Suite supports the full range of Tesla products on bothWindows and Linux operating systems.Using GPU, multi-GPU and MPI-GPU Computing drastically reduces the simulationtimes for CST customers. We will provide a status of current and future GPUdevelopments at CST and share detailed simulation results.CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

CST Milestones1992Foundation of CSTCommercialization of MAFIA (FIT)1998CST MICROWAVE STUDIO — PERFECTBOUNDARY APPROXIMATION (PBA) 2005Complete Technology for 3D EM2011System Assembly and Modeling (SAM)CSTCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12Market (excluding CST)

CST WorldwideCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

CST CustomersCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

CST GPU Computing Update In 2007 began support of nVidia GPU acceleration in our mainproduct, the MWS Time Domain Solver (FIT) Between 2008-2010 added multiple GPU and MPI GPU support forthe FIT solver In 2012, we added GPU acceleration support for three other solvers: Particle in Cell Solver Integral Equation Direct Solver TLM SolverCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Particle in Cell Solver: Introduction The PIC solver calculates the electromagnetic field by integratingMaxwell’s equationsAt the same time, it calculates trajectories of particles through thecalculated field.CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Magnetron BenchmarkNumber of MeshcellsAv. Particle Number1,610,2801.39e4Time CPU36h 40m 07sTime GPU9h 05m 48sTotal Speed UpCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-124.03

Particle in Cell: Core BenchmarksPIC Solver Loop SpeedupPIC Total illions00Number of Meshcells10203040MillionsNumber of MeshcellsComparison: nVidia Tesla C2050 vs. 2x Intel XEON E5620 @ 2.4 GHz (8 cores total)CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Integral Equation Solver: Introduction Boundary Element Solver- works on surface mesh Frequency Domain Targeted at electrically large & midsize structures ( 10 λ) Generates a complex dense matrix tobe factorizedCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Dielectric Lens in Ku/K-Band 14 - 28 GHz, Lens diameter 60mmLens: eps 3, thickness 12.2mmSurface cells: 16kSAsw 500Simulation time w/ Tesla C2075: 69minMemory used : 11 GBCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12Diam. 60mm

Integral Equation Solver: Core BenchmarksI-Solver Core SpeedupI-Solver Total Speedup164.51443.5310SpeedupSpeedup128642.521x C20501.52x C20504x C2050120.500010203040506070Thousands010Number of Unknowns203040506070ThousandsNumber of UnknownsComparison: nVidia Tesla C2050 vs. 2x Intel XEON E5620 @ 2.4 GHz (8 cores total)CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

TLM Solver: Introduction Broadband Transient SolverSpecial aperture and wire modeling (EMC/EMI)Octree localized meshCompact models: vents, seams, slots, shielded cablesSCNCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12Octree mesh

TLM GPU Speedup Rectangular waveguide horn antenna with parabolic reflector 2m diameter dish @ 3.5GHz Approximately 23 wavelengthsTimeSpeedup20122012 GPU20:094:354.4CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Satellite system benchmarkShielded TWPcableCable connectsto 3D structureOverlapping jointcompact seamwith 24 segmentsGPU Tesla 2050 solve time 45min8-core Westmere-EX CPU 180min: 4X speedupCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Time Domain (FIT) Solver Transient, low memory, broadband solution Conformal meshing (PBA, TST) Support for multi-GPU, MPI GPUPBATSTCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

PCB with USB connector/cable20M mesh cells12 Westmere-EX core Solver Loop 7840s2xM2070 Solver Loop 1596sSolver Loop Speedup 5xCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

SwappingModel too largeSpeedup of the solverloopSmall ModelGPU Computing – Typical PerformanceThe features which need thelargest amount of memory onthe GPU are: dispersive materials lossy metal open boundariesCPU PerformanceModel Sizes (Number of Mesh Cells)CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

GPU Computing – Typical PerformanceMean Speedup of Solver Loop(compared to dual Intel Xeon X5550, fastest memory configuration)25Speedup201510501 GPU (Tesla 10) 2 GPU (Tesla 10) 4 GPU (Tesla 10) 8 GPU (Tesla 10) 1 GPU (Tesla 20) 4 GPU (Tesla 20)ConfigurationCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Time Domain Core BenchmarksT Solver Core SpeedupT Solver Total Speedup20161814161212SpeedupSpeedup1410861081x C205062x C2050444x C20502200020406080100Millions020Number of Meshcells406080100MillionsNumber of MeshcellsComparison: nVidia Tesla C2050 vs. 2x Intel XEON E5620 @ 2.4 GHz (8 cores total)CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Complex Package Benchmark (SI)Very high complexity50 million mesh cellsCPU (2x Quad Core IntelXeon E5530, 2.4 GHz)Solver Loop Time/sSolver Loop Speedup4x Tesla 10 GPUs4x Tesla 20 GPUs394671662990123.739.8CST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Passenger Aircraft Benchmark: MPI GPU146M mesh cells, (4) MPI GPU nodes, (2) Tesla C1060 per nodeTotal simulation time 8 hoursCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Conclusion Significant performance improvement using GPUComputing CST GPU solvers (PIC, TLM, IE, TD) available forvarious applications Support of up to 8 GPUs/host for TD, IE Solvers;Cluster MPI GPU for TD SolverCST – COMPUTER SIMULATION TECHNOLOGY www.cst.com May-12

Latest developments in GPU acceleration for 3D Full Wave Electromagnetic simulation. Current and future GPU developments at CST; detailed simulation results. Keywords: gpu acceleration; 3d full wave electromagnetic simulation, cst studio suite, mpi-gpu, gpu technology confere

Related Documents:

the gpu computing era gpu computing is at a tipping point, becoming more widely used in demanding consumer applications and high-performance computing.this article describes the rapid evolution of gpu architectures—from graphics processors to massively parallel many-core multiprocessors, recent developments in gpu computing architectures, and how the enthusiastic

GPU Tutorial 1: Introduction to GPU Computing Summary This tutorial introduces the concept of GPU computation. CUDA is employed as a framework for this, but the principles map to any vendor’s hardware. We provide an overview of GPU computation, its origins and development, before presenting both the CUDA hardware and software APIs. New Concepts

OpenCV GPU header file Upload image from CPU to GPU memory Allocate a temp output image on the GPU Process images on the GPU Process images on the GPU Download image from GPU to CPU mem OpenCV CUDA example #include opencv2/opencv.hpp #include <

Introduction to GPU computing Felipe A. Cruz Nagasaki Advanced Computing Center Nagasaki University, Japan. Felipe A. Cruz Nagasaki University The GPU evolution The Graphic Processing Unit (GPU) is a processor that was specialized for processing graphics. The GPU has recently evolved towards a more flexible architecture.

GPU Computing in Matlab u Included in the Parallel Computing Toolbox. u Extremely easy to use. To create a variable that can be processed using the GPU, use the gpuArray function. u This function transfers the storage location of the argument to the GPU. Any functions which use this argument will then be computed by the GPU.

limitation, GPU implementers made the pixel processor in the GPU programmable (via small programs called shaders). Over time, to handle increasing shader complexity, the GPU processing elements were redesigned to support more generalized mathematical, logic and flow control operations. Enabling GPU Computing: Introduction to OpenCL

Will Landau (Iowa State University) Introduction to GPU computing for statisticicans September 16, 2013 20 / 32. Introduction to GPU computing for statisticicans Will Landau GPUs, parallelism, and why we care CUDA and our CUDA systems GPU computing with R CUDA and our CUDA systems Logging in

Software Development , Scrum [11] [12], Scrumban [Ladas 2009 and several va-riant methods of agile]. The agile methodology is based on the “iterative enhancement” [13] technique [14]. As a iteration based methodology, each iteration in the agile methodology represents a small scale and selfcontained Software Development Life Cycle - (SDLC) by itself . Unlike the Spiral model [1] , agile .