BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163600Z
LOCATION:501
DTSTART;TZID=America/Denver:20191118T144100
DTEND;TZID=America/Denver:20191118T150000
UID:submissions.supercomputing.org_SC19_sess117_ws_mchpc110@linklings.com
SUMMARY:Machine Learning Guided Optimal Use of GPU Unified Memory
DESCRIPTION:Workshop\n\nMachine Learning Guided Optimal Use of GPU Unified
  Memory\n\nXu, Emani, Lin, Hu, Liao\n\nNVIDIA's unified memory (UM) create
 s a pool of managed memory on top of physically separated CPU and GPU memo
 ries. UM automatically migrates page-level data on-demand so programmers c
 an quickly write CUDA codes on heterogeneous machines without tedious and 
 error-prone manual memory management. To improve performance, NVIDIA allow
 s advanced programmers to pass additional memory use hints to its UM drive
 r. However, it is extremely difficult for programmers to decide when and h
 ow to efficiently use unified memory, given the complex interactions betwe
 en applications and hardware. In this paper, we present a machine learning
 -based approach to choosing between discrete memory and unified memory, wi
 th additional consideration of different memory hints. Our approach utiliz
 es profiler-generated metrics of CUDA programs to train a model  offline, 
 which is later used to guide optimal use of UM for multiple applications a
 t runtime. We evaluate our approach on NVIDIA Volta GPU with a set of benc
 hmarks. Results show that the proposed model achieves 96% prediction accur
 acy in correctly identifying the optimal memory advice choice.\n\nTag: Wor
 kshop Reg Pass, HPC, Memory, OS and Runtime Systems, Runtime Systems\n\nRe
 gistration Category: Workshop Reg Pass, HPC, Memory, OS and Runtime System
 s, Runtime Systems
URL:https://sc19.supercomputing.org/presentation/?id=ws_mchpc110&sess=sess
 117
END:VEVENT
END:VCALENDAR