使用“基于分析的代码转换方法”来提升GPU特定的OpenCL kernel在多核/众核CPU上的性能移植性

WEN Mei; HUANG Da-fei; XUN Chang-qing; CHEN Dong

doi:10.1631/FITEE.1500032

Your Location：

Home >

Browse articles >

使用“基于分析的代码转换方法”来提升GPU特定的OpenCL kernel在多核/众核CPU上的性能移植性

计算机与自动化 | Updated：2022-06-06

- 使用“基于分析的代码转换方法”来提升GPU特定的OpenCL kernel在多核/众核CPU上的性能移植性
- Improving performance portability for GPU-specific OpenCL kernels on multi-core/many-core CPUs by analysis-based transformations
- 信息与电子工程前沿（英文） 2015年16卷第11期页码：899-916
- Affiliations：
  
  1.School of Computer, National University of Defense Technology, Changsha 410073, China
  2.National Key Laboratory of Parallel and Distributed Processing, Changsha 410073, China
- Author bio：
  
  [ "Mei WEN, E-mail: meiwen@nudt.edu.cn" ]
  Da-fei HUANG, huangdafei1012@163.com
  [ "Chang-qing XUN, E-mail: xunchangqing@nudt.edu.cn" ]
  [ "Dong CHEN: chendong@nudt.edu.cn" ]
- Funds：
  
  Project supported by the National Natural Science Foundation of China (No. 61272145) and the National High-Tech R & D Program (863) of China (No. 2012AA012706)
- DOI：10.1631/FITEE.1500032
  中图分类号： TP312
- 收稿：2015-01-30，
  
  修回：2015-10-19，
  
  纸质出版：2015-11
- Accepted：
Scan QR Code
使用“基于分析的代码转换方法”来提升GPU特定的OpenCL kernel在多核/众核CPU上的性能移植性[J]. 信息与电子工程前沿（英文）, 2015,16(11):899-916.

WEN Mei, HUANG Da-fei, XUN Chang-qing, et al. Improving performance portability for GPU-specific OpenCL kernels on multi-core/many-core CPUs by analysis-based transformations[J]. Frontiers of Information Technology & Electronic Engineering, 2015, 16(11): 899-916.
使用“基于分析的代码转换方法”来提升GPU特定的OpenCL kernel在多核/众核CPU上的性能移植性[J]. 信息与电子工程前沿（英文）, 2015,16(11):899-916. DOI： 10.1631/FITEE.1500032.

WEN Mei, HUANG Da-fei, XUN Chang-qing, et al. Improving performance portability for GPU-specific OpenCL kernels on multi-core/many-core CPUs by analysis-based transformations[J]. Frontiers of Information Technology & Electronic Engineering, 2015, 16(11): 899-916. DOI： 10.1631/FITEE.1500032.

摘要

目的

针对面向GPU设计的OpenCL kernel程序在CPU上性能移植性欠佳这一问题，设计一种基于访存特征分析的代码转换方法，提升性能移植性。

创新点

通过分析OpenCL kernel中的访存模式，去除不必要的局部存储数组及其带来的同步语句，并使用向量化和局域性重开发进一步优化代码，最终取得显著的性能提升。

方法

首先，针对OpenCL kernel代码中的数组访问，设计一种精确的线性化访问描述子（图2）。然后，利用该描述子，分两步对GPU特定的OpenCL kernel代码进行转换，以提高其在CPU上的性能（图7）。第一步为基于分析的work-item折叠，即通过分析访问描述子，找出并去除不必要的局部存储数组及其带来的同步语句，然后完成work-item折叠。第二步为适应架构的代码优化，即针对CPU架构的特点，使用向量化和局域性重开发进一步优化折叠后的代码。最后，上述代码转换过程被整合为一个工具链，连同一个调度程序，嵌入到一个开源的OpenCL运行时系统中（图11）。实验结果表明，这种转换方法可以显著提升GPU特定的OpenCL kernel在Intel Sandy Bridge架构CPU和Intel Knights Corner架构协处理器上的性能。

结论

准确分析OpenCL kernel代码中的访存模式，不仅利于判断局部存储数组是否适合于CPU架构，还能用于指导之后的代码优化过程，因此是提高性能移植性的重要步骤。

Abstract

OpenCL is an open heterogeneous programming framework. Although OpenCL programs are functionally portable

they do not provide performance portability

so code transformation often plays an irreplaceable role. When adapting GPU-specific OpenCL kernels to run on multi-core/many-core CPUs

coarsening the thread granularity is necessary and thus has been extensively used. However

locality concerns exposed in GPU-specific OpenCL code are usually inherited without analysis

which may give side-effects on the CPU performance. Typically

the use of OpenCL's local memory on multi-core/many-core CPUs may lead to an opposite performance effect

because local-memory arrays no longer match well with the hardware and the associated synchronizations are costly. To solve this dilemma

we actively analyze the memory access patterns using array-access descriptors derived from GPU-specific kernels

which can thus be adapted for CPUs by (1) removing all the unwanted local-memory arrays together with the obsolete barrier statements and (2) optimizing the coalesced kernel code with vectorization and locality re-exploitation. Moreover

we have developed an automated tool chain that makes this transformation of GPU-specific OpenCL kernels into a CPU-friendly form

which is accompanied with a scheduler that forms a new OpenCL runtime. Experiments show that the automated transformation can improve OpenCL kernel performance on a multi-core CPU by an average factor of 3.24. Satisfactory performance improvements are also achieved on Intel's many-integrated-core coprocessor. The resultant performance on both architectures is better than or comparable with the corresponding OpenMP performance.

关键词

Keywords

references

R. Allen , K. Kennedy . Optimizing Compilers for Modern Architectures: a Dependence-Based Approach . Morgan Kaufmann, San Francisco , 2002 .

V. Balasundaram , K. Kennedy . A technique for summarizing data access and its use in parallelism enhancing transformations . ACM SIGPLAN Not , 1989 . 24 ( 7 ): 41 - 53 . DOI: 10.1145/74818.74822 http://doi.org/10.1145/74818.74822 .

M.M. Baskaran , U. Bondhugula , S. Krishnamoorthy , 等 . A compiler framework for optimization of affine loop nests for GPGPUs . Proc. 22nd Annual Int. Conf. on Supercomputing , 2008 . p.225 - 234 . DOI: 10.1145/1375527.1375562 http://doi.org/10.1145/1375527.1375562 .

C. Bastoul . Code generation in the polyhedral model is easier than you think . Proc. 13th Int. Conf. on Parallel Architectures and Compilation Techniques , 2004 . p.7 - 16 . DOI: 10.1109/PACT.2004.1342537 http://doi.org/10.1109/PACT.2004.1342537 .

A. Danalis , G. Marin , C. McCurdy , 等 . The scalable heterogeneous computing (SHOC) benchmark suite . Proc. 3rd Workshop on General-Purpose Computation on Graphics Processing Units , 2010 . 63 - 74 . DOI: 10.1145/1735688.1735702 http://doi.org/10.1145/1735688.1735702 .

H. Dong , D. Ghosh , F. Zafar , 等 . Cross-platform OpenCL code and performance portability for CPU and GPU architectures investigated with a climate and weather physics model . Proc. 41st Int. Conf. on Parallel Processing Workshops , 2012 . p.126 - 134 . DOI: 10.1109/ICPPW.2012.19 http://doi.org/10.1109/ICPPW.2012.19 .

P. Du , R. Weber , P. Luszczek , 等 . From CUDA to OpenCL: towards a performance-portable solution for multi-platform GPU programming . Parall. Comput , 2012 . 38 ( 8 ): 391 - 407 . DOI: 10.1016/j.parco.2011.10.002 http://doi.org/10.1016/j.parco.2011.10.002 .

J. Fang , H. Sips , P. Jaaskelainen , 等 . Grover: looking for performance improvement by disabling local memory usage in OpenCL kernels . Proc. 43rd Int. Conf. on Parallel Processing , 2014a . p.162 - 171 . DOI: 10.1109/ICPP.2014.25 http://doi.org/10.1109/ICPP.2014.25 .

J. Fang , H. Sips , A.L. Varbanescu . Aristotle: a performance impact indicator for the OpenCL kernels using local memory . Sci. Progr , 2014b . 22 ( 3 ): 239 - 257 . DOI: 10.3233/SPR-140390 http://doi.org/10.3233/SPR-140390 .

Freeocl . FreeOCL: multi-platform implementation of OpenCL 1.2 targeting CPUs . Available from , 2012 . https://code.google.com/p/freeocl https://code.google.com/p/freeocl , [Accessed on Apr. 13, 2014] .

J. Gummaraju , L. Morichetti , M. Houston , 等 . Twin peaks: a software platform for heterogeneous computing on general-purpose and graphics processors . Proc. 19th Int. Conf. on Parallel Architectures and Compilation Techniques , 2010 . p.205 - 216 . DOI: 10.1145/1854273.1854302 http://doi.org/10.1145/1854273.1854302 .

D. Huang , M. Wen , C. Xun , 等 . Automated transformation of GPU-specific OpenCL kernels targeting performance portability on multi-core/many-core CPUs . Proc. Euro-Par , 2014 . p.210 - 221 . DOI: 10.1007/978-3-319-09873-9_18 http://doi.org/10.1007/978-3-319-09873-9_18 .

Intel Corporation . Guide to Vectorization with Intel C++ Compilers , 2012 .

Intel Corporation . Intel C++ Intrinsic Reference . Available from , 2013a . https://software.intel.com/sites/default/files/a6/22/18072-347603.pdf https://software.intel.com/sites/default/files/a6/22/18072-347603.pdf , [Accessed on Feb. 9, 2014] .

Corporation Intel . Intel SDK for OpenCL Applications XE 2013 Optimization Guide . Available from , 2013b . http://software.intel.com/en-us/vcsource/tools/opencl-sdk-xe/ http://software.intel.com/en-us/vcsource/tools/opencl-sdk-xe/ , [Accessed on Feb. 9, 2014] .

B. Jang , D. Schaa , P. Mistry , 等 . Exploiting memory access patterns to improve memory performance in data-parallel architectures . IEEE Trans. Parall. Distr. Syst , 2011 . 22 ( 1 ): 105 - 118 . DOI: 10.1109/TPDS.2010.107 http://doi.org/10.1109/TPDS.2010.107 .

C. Lattner , V. Adve . The LLVM compiler framework and infrastructure tutorial . In: Eigenmann R., Li Z.Y., Midkiff, S.P. (Eds.), Languages and Compilers for High Performance Computing. Springer , 2005 . p.15 - 16 . .

J. Lee , J. Kim , S. Seo , 等 . An OpenCL framework for heterogeneous multicores with local memory . Proc. 19th Int. Conf. on Parallel Architectures and Compilation Techniques , 2010 . p.193 - 204 . DOI: 10.1145/1854273.1854301 http://doi.org/10.1145/1854273.1854301 .

Team and others LLVM . Clang: a C language family frontend for LLVM . Available from , 2012 . http://clang.llvm.org/[AccessedonApr.13,2014] http://clang.llvm.org/[AccessedonApr.13,2014] , .

A. Munshi . The OpenCL specification . Available from , 2011 . http://www.khronos.org/opencl http://www.khronos.org/opencl , [Accessed on Apr. 12, 2014] .

Corporation Nvidia . OpenCL Best Practices Guide . Available from , 2011a . https://hpc.oit.uci.edu/nvidia-doc/sdk-cuda-doc/OpenCL/doc/OpenCL_Best_Practices_Guide.pdf https://hpc.oit.uci.edu/nvidia-doc/sdk-cuda-doc/OpenCL/doc/OpenCL_Best_Practices_Guide.pdf , [Accessed on Feb. 10, 2014] .

Corporation Nvidia . OpenCL Programming Guide for the CUDA Architecture . Available from , 2011b . https://hpc.oit.uci.edu/nvidia-doc/sdk-cuda-doc/OpenCL/doc/OpenCL_Programming_Guide.pdf https://hpc.oit.uci.edu/nvidia-doc/sdk-cuda-doc/OpenCL/doc/OpenCL_Programming_Guide.pdf , [Accessed on Feb. 10, 2014] .

Y. Paek , J. Hoeflinger , D. Padua . Efficient and precise array access analysis . ACM Trans. Progr. Lang. Syst , 2002 . 24 ( 1 ): 65 - 109 . DOI: 10.1145/509705.509708 http://doi.org/10.1145/509705.509708 .

S.J. Pennycook , S.D. Hammond , S.A. Wright , 等 . An investigation of the performance portability of OpenCL . J. Parall. Distr. Comput , 2013 . 73 ( 11 ): 1439 - 1450 . DOI: 10.1016/j.jpdc.2012.07.005 http://doi.org/10.1016/j.jpdc.2012.07.005 .

P.M. Phothilimthana , J. Ansel , J. Ragan-Kelley , 等 . Portable performance on heterogeneous architectures . Proc. 18th Int. Conf. on Architechtural Support for Programming Languages and Operating Systems , 2013 . p.431 - 444 . DOI: 10.1145/2451116.2451162 http://doi.org/10.1145/2451116.2451162 .

S. Rul , H. Vandierendonck , J. D'Haene , 等 . An experimental study on performance portability of OpenCL kernels . Symp. on Application Accelerators in High Performance Computing. Available from , 2010 . https://biblio.ugent.be/publication/1016024 https://biblio.ugent.be/publication/1016024 , .

Z. Shen , Z. Li , P. Yew . An empirical study of Fortran programs for parallelizing compilers . IEEE Trans. Parall. Distr. Syst , 1990 . 1 ( 3 ): 356 - 364 . DOI: 10.1109/71.80162 http://doi.org/10.1109/71.80162 .

S.M. Steven . Advanced Compiler Design and Implementation . Morgan Kaufmann, San Francisco , 1997 .

J.A. Stratton , S.S. Stone , W.M.W. Hwu . MCUDA: an effective implementation of CUDA kernels for multi-core CPUs . Proc. 21st Int. Workshop on Languages and Compilers for Parallel Computing , 2008 . p.16 - 30 . DOI: 10.1007/978-3-540-89740-8_2 http://doi.org/10.1007/978-3-540-89740-8_2 .

J.A. Stratton , V. Grover , J. Marathe , 等 . Efficient compilation of fine-grained SPMD threaded programs for multicore CPUs . Proc. 8th Annual IEEE/ACM Int. Symp. on Code Generation and Optimization , 2010 . p.111 - 119 . DOI: 10.1145/1772954.1772971 http://doi.org/10.1145/1772954.1772971 .

J.A. Stratton , H. Kim , T.B. Jablin , 等 . Performance portability in accelerated parallel kernels . Technical Report No. IMPACT-13-01, Center for Reliable and High-Performance Computing, University of Illinois at Urbana-Champaign, IL , 2013 .

TOP500.org . TOP500 lists: November 2014 . Available from , 2014 . http://top500.org/lists/2014/11/ http://top500.org/lists/2014/11/ , [AccessedonNov. 29,2014] .

R. Triolet , F. Irigoin , P. Feautrier . Direct parallelization of call statements . ACM SIGPLAN Not , 1986 . 21 ( 7 ): 176 - 185 . DOI: 10.1145/13310.13329 http://doi.org/10.1145/13310.13329 .

浏览量

Downloads

CSCD

文章被引用时，请邮件提醒。

Submit

工具集

关联资源

暂无数据