15 Mar 2004 09:35:23 -0500

Related articles |
---|

Data flow analysis question saswat@cc.gatech.edu (2004-03-15) |

From: | saswat@cc.gatech.edu (Saswat) |

Newsgroups: | comp.compilers |

Date: | 15 Mar 2004 09:35:23 -0500 |

Organization: | http://groups.google.com |

Keywords: | analysis, question |

Posted-Date: | 15 Mar 2004 09:35:22 EST |

Hello,

I am faced with this analysis problem that seems to be useful in

compilers and software engineering tools. So I am trying to find if it

has already been solved or any reference to similar problems. If the

problem sounds familiar please direct me to the reference.

Roughly, the analysis finds out all statement pairs (s1, s2) that use

the same variable (say x) such that, s2 always follows s1 and when s1

uses a particular definition of x, say p, the following s2 also use

the same definition of x (ie. p).

Formally:

For a procedure, suppose we have a trace of program statements of the

form $s_{i_1}, s_{i_2} ... s_{i_n}$ denoting the order in which

statements are executed. Let $(s_{p/q})^+$ denote a sequence of

statements in which $s_p$ occurs one or more number of times, but

$s_q$ does not occur. The goal of the analysis to find all statement

pairs $s_p$ and $s_q$ (that use the same variable) such that in the

trace:

1. every occurrence of a $(s_{p/q})^+$ sequence ($S_1$) in the trace

is followed by atleast one $s_q$

2. each occurrence of $s_p$ in $S_1$ and the following $s_q$ use the

same definition of $x$.

I was thinking of SSA. But does that take care all the cases?

Thanks

Saswat

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.