Title
Soundness of data flow analyses for weak memory models
Abstract
Modern multi-core microprocessors implement weak memory consistency models; programming for these architectures is a challenge. This paper solves a problem open for ten years, and originally posed by Rinard: we identify sufficient conditions for a data flow analysis to be sound w.r.t. weak memory models. We first identify a class of analyses that are sound, and provide a formal proof of soundness at the level of trace semantics. Then we discuss how analyses unsound with respect to weak memory models can be repaired via a fixed point iteration, and provide experimental data on the runtime overhead of this method.
Year
DOI
Venue
2011
10.1007/978-3-642-25318-8_21
APLAS
Keywords
Field
DocType
sufficient condition,formal proof,weak memory consistency model,weak memory model,modern multi-core microprocessors,runtime overhead,fixed point iteration,experimental data,data flow analysis,sound w
Sequential consistency,Computer science,Fixed-point iteration,Data-flow analysis,Theoretical computer science,Memory model,Consistency model,Soundness,Formal proof,Data flow diagram
Conference
Citations 
PageRank 
References 
9
0.56
27
Authors
5
Name
Order
Citations
PageRank
Jade Alglave160826.53
Daniel Kroening23084187.60
John Lugton390.56
Vincent Nimal4752.51
Michael Tautschnig542525.84