BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111116T011500Z
DTEND:20111116T030000Z
LOCATION:WSCC North Galleria 2nd/3rd Floors
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Graphics Processing Units (GPUs) represent a possible means of exploiting large-scale parallelism. Source-to-source transformation tools mapping CPU code to GPU code (e.g. PGI accelerator) are available. But identification of code segments that, when run on a GPU will create a significant performance enhancement, requires expert knowledge of algorithms, architectures, compilers and the program structure. In this work, we demonstrate a reasonably accurate prediction of code segments that can benefit from GPU execution and compare our identified code segments with those that have been manually selected by experts for mapping to GPUs. The characterization uses the PerfExpert tool for multicore chip optimization to obtain the measurements on the code segments while a new tool, MACPO, is used for characterization of data structure execution properties. The poster describes the characterization process, gives the results of applying the process to the Rodina parallel benchmarks and gives the assumptions and limitations of the process.
SUMMARY:Determining code segments that can benefit from execution on GPUs
PRIORITY:3
END:VEVENT
END:VCALENDAR