Parallel memory prediction for fused linear algebra kernels Journal Article uri icon



  • The performance of many scientific programs is limited by data movement. Loop fusion is one optimization used to increase the speed of memory bound operations. To automate loop fusion for matrix computations, we developed the Build to Order (BTO) compiler. Within BTO, an analytic memory model efficiently and accurately reduces the number of serial loop fusion options considered. In this paper, we extend the model to shared memory parallel machines. We detail the differences between parallel and serial memory use and runtime prediction and explain the changes made to include parallel machines in the model. Analysis of the parallel model's predictions show that when it is included in BTO it will reduce the search space of considered routines.

publication date

  • March 29, 2011

has restriction

  • closed

Date in CU Experts

  • October 17, 2013 10:46 AM

Full Author List

  • Karlin I; Jessup E; Belter G; Siek JG

author count

  • 4

Other Profiles

International Standard Serial Number (ISSN)

  • 0163-5999

Additional Document Info

start page

  • 43

end page

  • 49


  • 38


  • 4