BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Lock Inference in the Presence of Large Libraries - Khilan Gudka (
 University of Cambridge)
DTSTART:20120607T150000Z
DTEND:20120607T160000Z
UID:TALK37681@talks.cam.ac.uk
CONTACT:Eiko Yoneki
DESCRIPTION:Atomic sections can be implemented using lock inference. For l
 ock inference\nto be practically useful\, it is crucial that large librari
 es be analysed.\nHowever\, libraries are challenging for static analysis\,
  due to their\ncyclomatic complexity.\n\nExisting approaches either ignore
  libraries\, require library implementers to\nannotate which locks to take
  or only consider accesses performed upto one\nlevel deep in library call 
 chains. Thus\, some library accesses may go\nunprotected\, leading to atom
 icity violations that atomic sections are\nsupposed to eliminate.\n\nWe pr
 esent a lock inference approach for Java that analyses library methods\nin
  full. We achieve this by (i) formulating lock inference as an\nInterproce
 dural Distributive Environment dataflow problem\, (ii) using a\ngraph repr
 esentation for summary information and (iii) applying a number of\noptimis
 ations to our implementation to reduce space-time requirements and\nlocks 
 inferred. We demonstrate the scalability of our approach by analysing\nthe
  entire GNU Classpath library comprising 122KLOC.\n
LOCATION:FW26\, Computer Laboratory\, William Gates Builiding
END:VEVENT
END:VCALENDAR