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:20211207T054806Z
LOCATION:220-221
DTSTART;TZID=America/Chicago:20211117T140000
DTEND;TZID=America/Chicago:20211117T143000
UID:submissions.supercomputing.org_SC21_sess147_pap571@linklings.com
SUMMARY:Hybrid, Scalable, Trace-Driven Performance Modeling of GPGPUs
DESCRIPTION:Paper\n\nHybrid, Scalable, Trace-Driven Performance Modeling o
 f GPGPUs\n\nArafa, Badawy, ElWazir, Barai, Eker...\n\nIn this paper, we pr
 esent PPT-GPU, a scalable performance prediction toolkit for GPUs. PPT-GPU
  achieves scalability through a hybrid high-level modeling approach where 
 some computations are extrapolated and multiple parts of the model are par
 allelized. The tool primary prediction models use pre-collected memory and
  instructions traces of the workloads to accurately capture the dynamic be
 havior of the kernels.\n\nPPT-GPU reports an extensive array of GPU perfor
 mance metrics accurately while being easily extensible. We use a broad set
  of benchmarks to verify predictions accuracy. We compare the results agai
 nst hardware metrics collected using vendor profiling tools and cycle-accu
 rate simulators. The results show that the performance predictions are hig
 hly correlated to the actual hardware (MAPE: < 16% and Correlation: > 0.98
 ). Moreover, PPT-GPU is orders of magnitude faster than cycle-accurate sim
 ulators. This comprehensiveness of the collected metrics can guide archite
 cts and developers to perform design space explorations.\n\nTag: Reproduci
 bility Badge, Performance\n\nRegistration Category: Tech Program Reg Pass\
 n\nReproducibility Badges: Artifact Available, Artifact Functional, Result
 s Reproduced
END:VEVENT
END:VCALENDAR
