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:20211207T055412Z
LOCATION:Online
DTSTART;TZID=America/Chicago:20211114T160000
DTEND;TZID=America/Chicago:20211114T163000
UID:submissions.supercomputing.org_SC21_sess327_ws_hipar106@linklings.com
SUMMARY:PPIR: Parallel Pattern Intermediate Representation
DESCRIPTION:Workshop\n\nPPIR: Parallel Pattern Intermediate Representation
 \n\nSchmitz, Miller, Trümper, Müller\n\nHPC systems are becoming rapidly l
 arger, heterogeneous, and generally more complex to fulfill the growing de
 mand for computational resources in science and engineering. HPC system an
 d architecture-specific optimizations are necessary to leverage the full p
 otential of such systems. One approach is to develop a high-level, structu
 red source code by leveraging parallel patterns. Our previous work propose
 d a framework to optimize pattern-based codes for a target HPC system auto
 matically. Global optimizations can be carried out by the framework, optim
 izing the dataflow throughout the parallel algorithm and efficiently sched
 uling the application on the available resources.\n\nWe propose a compact 
 and fast intermediate representation (IR) of parallel pattern-based applic
 ations to enable such global optimizations. A tree-based hierarchical abst
 raction called abstract pattern tree (APT) is introduced to this end. It i
 s augmented with scheduling information provided by the optimization frame
 work. The resulting IR is implemented in a prototype compiler and evaluate
 d on the Rodinia benchmark suite. 17 out of the 19 benchmarks could be rep
 resented by a small set of parallel patterns and statically analyzed by th
 e prototype compiler. The original source code could be highly condensed b
 y using parallel patterns, which increases the development productivity. T
 he compiler prototype implementation showed a feasible compilation time be
 low one second for most benchmarks. The proposed PPIR was compared to LLVM
  IR.\n\nTag: Online Only, Algorithms, Architectures, Extreme Scale Computi
 ng, Heterogeneous Systems, Memory Systems, Parallel Programming Languages 
 and Models, Resource Management and Scheduling\n\nRegistration Category: W
 orkshop Reg Pass
END:VEVENT
END:VCALENDAR