BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260522T150115Z
LOCATION:D165
DTSTART;TZID=America/Chicago:20181112T153000
DTEND;TZID=America/Chicago:20181112T160000
UID:submissions.supercomputing.org_SC18_sess161_ws_pmbsf107@linklings.com
SUMMARY:Is Data Placement Optimization Still Relevant on Newer GPUs?
DESCRIPTION:Md Abdullah Shahneous Bari (Stony Brook University); Larisa St
 oltzfus (University of Edinburgh); Pei-Hung Lin, Chunhua Liao, and Murali 
 Emani (Lawrence Livermore National Laboratory); and Barbara Chapman (Stony
  Brook University, Brookhaven National Laboratory)\n\nModern supercomputer
 s often use Graphic Processing Units (or GPUs) to meet the evergrowing dem
 ands for energy efficient high performance computing. GPUs have a complex 
 memory architecture with various types of memories and caches, such as glo
 bal memory, shared memory, constant memory, and texture memory. Data place
 ment optimization, i.e. optimizing the placement of data among these diffe
 rent memories, has a significant impact on the performance of HPC applicat
 ions running on early generations of GPUs. However, newer generations of G
 PUs have new memory features. They also implement the same high-level memo
 ry hierarchy differently.\n\nIn this paper, we design a set of experiments
  to explore the relevance of data placement optimizations on several gener
 ations of NVIDIA GPUs, including Kepler, Maxwell, Pascal, and Volta. Our e
 xperiments include a set of memory microbenchmarks, CUDA kernels and a pro
 xy application. The experiments are configured to include different CUDA t
 hread blocks, data input sizes, and data placement choices. The results sh
 ow that newer generations of GPUs are less sensitive to data placement opt
 imization compared to older ones, mostly due to improvements to caches of 
 the global memory.\n\nTag: Benchmarks, Parallel Programming Languages, Lib
 raries, and Models, Performance, Simulation\n\nRegistration Category: Work
 shop Reg Pass\n\nSession Chair: Steven A. Wright (University of York, Engl
 and)\n\n
END:VEVENT
END:VCALENDAR