G.: A new framework for analysis and optimization of shared memory parallel programs

In this paper we present a new framework for analysis and optimization of shared memory parallel programs. Our framework is based on two key concepts: (1) concurrency relation and (2) isolation semantics. Two statements are said to be concurrent if there is an execution in which two threads can execute the two statements simultaneously. Isolation semantics essentially treats critical sections as being isolated or atomic. The main contributions of this paper are as follows: • We present a simple algorithm for computing concurrency relation among two or more statements in a parallel program that includes cobegin/coend, parallel for, barrier, post/wait, and critical sections. • We propose a new framework for solving data flow problems by reifying two key concepts: concurrency relation and isolation semantics for critical sections. As an example we solve the pointer analysis problem using our framework To the best of our knowledge ours is the first work that directly incorporates isolation semantics for solving data flow analysis