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:20240116T191657Z
LOCATION:301-302-303
DTSTART;TZID=America/Denver:20231114T113000
DTEND;TZID=America/Denver:20231114T120000
UID:submissions.supercomputing.org_SC23_sess172_pap389@linklings.com
SUMMARY:Automated Mapping of Task-Based Programs onto Distributed and Hete
 rogeneous Machines
DESCRIPTION:Paper\n\nThiago S. F. X. Teixeira (Stanford University), Alexa
 ndra Henzinger (Massachusetts Institute of Technology (MIT)), and Rohan Ya
 dav and Alex Aiken (Stanford University)\n\nIn a parallel and distributed 
 application, a mapping is a selection of a processor for each computation 
 or task and memories for the data collections that each task accesses.  Fi
 nding high-performance mappings is challenging, particularly on heterogene
 ous hardware with multiple choices for processors and memories.  We show t
 hat fast mappings are sensitive to the machine, application, and input.  P
 orting to a new machine, modifying the application, or using a different i
 nput size may necessitate re-tuning the mapping to maintain the best possi
 ble performance.\n\nWe present AutoMap, a system that automatically tunes 
 the mapping to the hardware used and finds fast mappings without user inte
 rvention or code modification.  In contrast, hand-written mappings often r
 equire days of experimentation.  AutoMap utilizes a novel constrained coor
 dinate-wise descent search algorithm that balances the trade-off between r
 unning computations quickly and minimizing data movement.  AutoMap discove
 rs mappings up to 2.41x faster than custom, hand-written mappers.\n\nTag: 
 Heterogeneous Computing, Programming Frameworks and System Software, Task 
 Parallelism\n\nRegistration Category: Tech Program Reg Pass\n\nReproducibi
 lity Badges: Artifact Available\n\nSession Chair: Claudia Fohry (Universit
 y of Kassel)
END:VEVENT
END:VCALENDAR
