从SSE到OpenCL：多核CPU上骨骼动画并行算法对比研究

系统仿真学报 ›› 2015, Vol. 27 ›› Issue (2): 336-343.

从SSE到OpenCL：多核CPU上骨骼动画并行算法对比研究

刘寿生, 陈戈, 马纯永, 韩勇

中国海洋大学信息科学与工程学院,青岛 266100

收稿日期:2014-01-22 修回日期:2014-04-30 发布日期:2020-09-02
作者简介:刘寿生(1985-), 男, 江西宁都人, 博士生, 研究方向为虚拟现实、并行计算、高性能;陈戈(1965-), 男, 浙江宁波人, 博士, 教授, 博导, 研究方向为海洋遥感、地理信息系统、虚拟现实;马纯永(1984-), 男, 山东潍坊人, 博士, 讲师, 研究方向为虚拟现实。
基金资助:
国家海洋局数字海洋科学技术重点实验室开放基金(KLDO201303)

From SSE to OpenCL: Comparison of Parallel Algorithms for Skeletal Animation on Multi-core CPUs

Liu Shousheng, Chen Ge, Ma Chunyong, Han Yong

College of Information Science and Engineering, Ocean University of China, Qingdao 266100, China

Received:2014-01-22 Revised:2014-04-30 Published:2020-09-02

摘要/Abstract

摘要： 拥有高精度蒙皮和复杂骨骼绑定关系的骨骼动画,渲染时存在很大的性能瓶颈。以往研究采用GPU加速动画,但高端GPU成本过高,而中低端GPU的通用计算性能有时不如高端CPU。为充分挖掘算法在多核CPU上的执行性能,弥补中低端GPU通用计算性能的不足,提出了基于OpenCL的针对指令和线程的新兴集成并行方案,并与基于SSE结合OpenMP针对指令和线程的传统独立并行方案展开对比。实验结果表明,在多数CPU和多种复杂度的数据上,基于OpenCL的新兴并行方案的性能明显高于基于SSE的传统并行方案,并且性能优势随着数据复杂度的增加而提升。

关键词: 骨骼动画, 并行计算, OpenCL, SSE

Abstract: While skeletal animations consist of high fidelity mesh and complex vertex binding, there is a huge performance bottleneck. Previous studies used GPUs to accelerate animation, but the general purpose computing performance of cheap GPUs was lower than high end CPUs. In order to explore the performance for multi-core CPUs to make up the shortfall of cheap GPUs for general purpose computing, the latest parallel scheme based on OpenCL which integrated the former two levels of parallelization was proposed, comparing to the traditional parallel scheme formed by SSE and OpenMP oriented to instruction parallelization and thread parallelization separately. The experimental results show that on most CPU with different complex data, the performance of the parallel scheme based on OpenCL is better than SSE. And the more complex the data is, the greater the performance advantage grows.

Key words: skeletal animation, parallel computing, OpenCL, SSE

中图分类号:

TP391.9

刘寿生, 陈戈, 马纯永, 韩勇. 从SSE到OpenCL：多核CPU上骨骼动画并行算法对比研究[J]. 系统仿真学报, 2015, 27(2): 336-343.

Liu Shousheng, Chen Ge, Ma Chunyong, Han Yong. From SSE to OpenCL: Comparison of Parallel Algorithms for Skeletal Animation on Multi-core CPUs[J]. Journal of System Simulation, 2015, 27(2): 336-343.

参考文献

[1] Kavan L.Real-time Skeletal Animation[D]. Prague, Czech Republic: Czech Technical University, 2007.
[2] 赵维, 谢晓方. 虚拟人技术发展现状及其在工程中的应用[J]. 系统仿真学报, 2009, 21(17): 5473-5476.
(ZHAO Wei, XIE Xiao-fang.Development of Virtual Human Technology and Its Engineering Application[J]. Journal of System Simulation (S1004-731X), 2009, 21(17): 5473-5476.)
[3] Burtnyk N, Wein M.Interactive Skeleton Techniques for Enhancing Motion Dynamics in Key Frame Animation[J]. Communications of the ACM (S0001-0782), 1976, 19(10): 564-569.
[4] Chafi H, Sujeeth A K, Brown K J, et al. A Domain-specific Approach to Heterogeneous Parallelism[C]// Proceedings of the 16th ACM symposium on Principles and practice of parallel programming. USA: ACM, 2011: 35-46.
[5] Brown K J, Sujeeth A K, Lee H J, et al. A Heterogeneous Parallel Framework for Domain-specific Languages[C]// Parallel Architectures and Compilation Techniques (PACT). USA: IEEE, 2011: 89-100.
[6] Ismail L, GuerchiD. Performance Evaluation of Convolution on the Cell Broadband Engine Processor[J]. IEEE Transactions on Parallel and Distributed Systems (S1045-9219), 2011, 22(2): 337-351.
[7] Lindholm E, Kilgard M J, Moreton H.A User-programmable Vertex Engine[C]// Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques. USA: ACM, 2001: 149-158.
[8] 季卓尔, 张景峤. 基于可编程 GPU 的骨骼动画[J]. 计算机工程与应用, 2008, 44(22): 77-80.
[9] 胡前亮, 陈炳发. 一种采用CUDA的骨骼动画阴影实时仿真方法[J]. 小型微型计算机系统, 2011, 32(1):165-168.
[10] 郝爱民, 赵永涛, 吴伟和, 等. 任意姿态虚拟人网格模型骨骼提取算法[J]. 中国图象图形学报, 2011, 16(6): 1008-1014.
[11] 刘登志. 人体角色的自动绑定与卡通运动[D]. 杭州:浙江大学, 2011.
[12] Yan H B, Hu S M, Martin R R, et al. Shape Deformation Using a Skeleton to Drive Simplex Transformations[J]. IEEE Transactions on Visualization and Computer Graphics (S1077-2626), 2008, 14(3): 693-706.
[13] Lee V W, Kim C, Chhugani J, et al. Debunking the 100X GPU vs. CPU Myth: An Evaluation of Throughput Computing on CPU and GPU[C]// ACM SIGARCH Computer Architecture News. USA: ACM, 2010: 451-460.
[14] Shi G, Li M, Lipasti M.Accelerating Search and Recognition Workloads with SSE 4.2 String and Text Processing Instructions[C]// Performance Analysis of Systems and Software (ISPASS), 2011 IEEE International Symposium on. USA: IEEE, 2011: 145-153.
[15] Shen J, Fang J, Sips H, et al. Performance Gaps between OpenMP and OpenCL for Multi-core CPUs[C]// Parallel Processing Workshops (ICPPW), 2012 41st International Conference on. Pittsburgh, PA, USA: IEEE, 2012: 116-125.
[16] Pennycook S J, Hammond S D, Wright S A, et al. An Investigation of the Performance Portability of OpenCL[J]. Journal of Parallel and Distributed Computing (S0743-7315), 2013, 73(11): 1439-1450.
[17] Khronos. The OpenCL Specification [EB/OL]. (2012) [2012-11-14]. http://www.khronos.org/r egistry/cl/specs/ opencl-1.2.pdf.
[18] Du P, Weber R, Luszczek P, et al. From CUDA to OpenCL: Towards a Performance-portable Solution for Multi-platform GPU Programming[J]. Parallel Computing (S0167-8191), 2012, 38(8): 391-407.
[19] Sun X H, Chen Y.Reevaluating Amdahl’s Law in the Multicore Era[J]. Journal of Parallel and Distributed Computing (S0743-7315), 2010, 70(2): 183-188.