BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Beyond the embarrassingly parallel – New languages\, compilers\,
  and runtimes for big-data processing
DTSTART:20170412T101500
DTEND:20170412T113000
DTSTAMP:20260407T093227Z
UID:2111e9f784d1d6a1a23e680701fa4bbb9fbe1c2c63aeb75c91e0f5ef
CATEGORIES:Conferences - Seminars
DESCRIPTION:By : Madan Musuvathi - Microsoft Research\n\nAbstract :\nLarge
 -scale data processing requires large-scale parallelism. Data-processing s
 ystems from traditional databases to Hadoop and Spark rely on embarrassing
 ly-parallel relational primitives (e.g. map\, reduce\, filter\, and join) 
 to extract parallelism from input programs. But many important application
 s\, such as machine learning and log processing\, iterate over large data 
 sets with true loop-carried dependences across iterations. As such\, these
  applications are not readily parallelizable in current data-processing sy
 stems. \n \nIn this talk\, I will challenge the premise that parallelism
  requires independent computations. In particular\, I will describe a gene
 ral methodology for extracting parallelism from dependent computations. Th
 e basic idea is replace dependences with symbolic unknowns and execute the
  dependent computations symbolically in parallel. The challenge of paralle
 lization now becomes a\, hopefully mechanizable\, task of performing the r
 esulting symbolic execution efficiently. This methodology opens up the pos
 sibility of designing new languages for data-processing computations\, com
 pilers that automatically parallelize such computations\, and runtimes tha
 t exploit the additional parallelism. I will describe our initial successe
 s with this approach and the research challenges that lie ahead. \nLarge-
 scale data processing requires large-scale parallelism. Data-processing sy
 stems from traditional databases to Hadoop and Spark rely on embarrassingl
 y-parallel relational primitives (e.g. map\, reduce\, filter\, and join) t
 o extract parallelism from input programs. But many important applications
 \, such as machine learning and log processing\, iterate over large data s
 ets with true loop-carried dependences across iterations. As such\, these 
 applications are not readily parallelizable in current data-processing sys
 tems. \n \nIn this talk\, I will challenge the premise that parallelism 
 requires independent computations. In particular\, I will describe a gener
 al methodology for extracting parallelism from dependent computations. The
  basic idea is replace dependences with symbolic unknowns and execute the 
 dependent computations symbolically in parallel. The challenge of parallel
 ization now becomes a\, hopefully mechanizable\, task of performing the re
 sulting symbolic execution efficiently. This methodology opens up the poss
 ibility of designing new languages for data-processing computations\, comp
 ilers that automatically parallelize such computations\, and runtimes that
  exploit the additional parallelism. I will describe our initial successes
  with this approach and the research challenges that lie ahead. \n\nBio :
 \nMadan Musuvathi is a Principal Researcher at Microsoft Research working 
 in the intersection of programming languages and systems\, with specific f
 ocus on concurrency and parallelism. His interests span program analysis\,
  systems\, model checking\, verification\, and theorem proving. His resear
 ch has led to several tools that improve the lives of software developers 
 both at Microsoft and at other companies. He received his Ph.D. from Stanf
 ord University in 2004.\n\nMore information\n 
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR