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:20181221T160726Z
LOCATION:D170
DTSTART;TZID=America/Chicago:20181111T121400
DTEND;TZID=America/Chicago:20181111T123000
UID:submissions.supercomputing.org_SC18_sess149_ws_mchpc108@linklings.com
SUMMARY:Data Placement Optimization in GPU Memory Hierarchy Using Predicti
 ve Modeling
DESCRIPTION:Workshop\nMemory, NVRAM, Parallel Programming Languages, Libra
 ries, and Models, Workshop Reg Pass\n\nData Placement Optimization in GPU 
 Memory Hierarchy Using Predictive Modeling\n\nStoltzfus, Emani, Lin, Liao\
 n\nModern supercomputers often use Graphic Processing Units (or GPUs) to m
 eet the ever-growing demands for high performance computing. GPUs typicall
 y have a complex memory architecture with various types of memories and ca
 ches, such as global memory, shared memory, constant memory, and texture m
 emory.The placement of data on these memories has a tremendous impact on t
 he performance of the HPC applications and identifying the optimal placeme
 nt location is non-trivial. \n\nIn this paper, we propose a machine learn
 ing-based approach to determine the best class of GPU memory that will min
 imize GPU kernel execution time. The machine learning process utilizes a s
 et of performance counters obtained from profiling runs and combines with 
 relevant hardware features to generate trained models. We evaluate our app
 roach on several generations of NVIDIA GPUs, including Kepler, Maxwell, Pa
 scal, and Volta on a set of benchmarks. The results show that the trained 
 models achieve prediction accuracies over 90%.
URL:https://sc18.supercomputing.org/presentation/?id=ws_mchpc108&sess=sess
 149
END:VEVENT
END:VCALENDAR