Compiler Optimizations for Industrial Unstructured Mesh CFD Applications on GPUs

详细信息    查看全文

• 作者：C. Bertolli (17)
  A. Betts (17)
  N. Loriant (17)
  G. R. Mudalige (18)
  D. Radford (21)
  D. A. Ham (17) (20)
  M. B. Giles (18) (19)
  P. H. J. Kelly (17)
  
• 关键词：Computational Fluid Dynamics ; Unstructured Meshes ; Graphical Processing Units ; Compilers
• 刊名：Lecture Notes in Computer Science
• 出版年：2013
• 出版时间：2013
• 年：2013
• 卷：7760
• 期：1
• 页码：127-142
• 全文大小：218KB
• 参考文献：1. Han, H., Tseng, C.-W.: A Comparison of Locality Transformations for Irregular Codes. In: Dwarkadas, S. (ed.) LCR 2000. LNCS, vol.聽1915, pp. 70鈥?4. Springer, Heidelberg (2000) CrossRef
  2. Burgess, D.A., Giles, M.B.: Renumbering unstructured grids to improve the performance of codes on hierarchical memory machines. Adv. Eng. Softw.聽28(3), 189鈥?01 (1997) CrossRef
  3. Chevalier, C., Pellegrini, F.: PT-Scotch: A tool for efficient parallel graph ordering. Parallel Comput.聽34(6-8), 318鈥?31 (2008) CrossRef
  4. Karypis, G., Kumar, V.: A fast and high quality multilevel scheme for partitioning irregular graphs. SIAM J. Sci. Comput.聽20(1), 359鈥?92 (1998) CrossRef
  5. Strout, M.M., Carter, L., Ferrante, J.: Compile-time composition of run-time data and iteration reorderings. In: Procs. of the PLDI 2003 (June 2003)
  6. DeVito, Z., Joubert, N., Palacios, F., Oakley, S., Medina, M., Barrientos, M., Elsen, E., Ham, F., Aiken, A., Duraisamy, K., Darve, E., Alonso, J., Hanrahan, P.: Liszt: a domain specific language for building portable mesh-based PDE solvers. In: Procs. of SC, pp. 9:1鈥?:12. ACM, New York (2011)
  7. Giles, M.B.: OP2 User鈥檚 Manual (April 2012), gilesm/op2/user.pdf"> http://people.maths.ox.ac.uk/gilesm/op2/user.pdf
  8. Giles, M.B., Mudalige, G.R., Sharif, Z., Markall, G., Kelly, P.H.: Performance analysis and optimization of the OP2 framework on many-core architectures. The Computer Journal聽55(2), 168鈥?80 (2012) CrossRef
  
• 作者单位：C. Bertolli (17)
  A. Betts (17)
  N. Loriant (17)
  G. R. Mudalige (18)
  D. Radford (21)
  D. A. Ham (17) (20)
  M. B. Giles (18) (19)
  P. H. J. Kelly (17)
  
  17. Dept. of Computing, Imperial College London, UK
  18. Oxford e-Research Centre, University of Oxford, UK
  21. Rolls Royce Plc., UK
  20. Grantham Institute for Climate Change, Imperial College London, UK
  19. Mathematical Institute, University of Oxford, UK
  
• ISSN：1611-3349

文摘

Graphical Processing Units (GPUs) have shown acceleration factors over multicores for structured mesh-based Computational Fluid Dynamics (CFD). However, the value remains unclear for dynamic and irregular applications. Our motivating example is HYDRA, an unstructured mesh application used in production at Rolls-Royce for the simulation of turbomachinery components of jet engines. We describe three techniques for GPU optimization of unstructured mesh applications: a technique able to split a highly complex loop into simpler loops, a kernel specific alternative code synthesis, and configuration parameter tuning. Using these optimizations systematically on HYDRA improves the GPU performance relative to the multicore CPU. We show how these optimizations can be automated in a compiler, through user annotations. Performance analysis of a large number of complex loops enables us to study the relationship between optimizations and resource requirements of loops, in terms of registers and shared memory, which directly affect the loop performance.