14 Apr 2006 14:13:38 -0400

Related articles |
---|

Computing maximal register pressure of a DAG touati@prism.uvsq.fr (Touati Sid) (2006-04-14) |

Is a reducable flow graph necessary? derek@knosof.co.uk (Derek M. Jones) (2006-04-14) |

From: | "Derek M. Jones" <derek@knosof.co.uk> |

Newsgroups: | comp.compilers |

Date: | 14 Apr 2006 14:13:38 -0400 |

Organization: | ntl Cablemodem News Service |

References: | 06-04-097 |

Keywords: | optimize |

Posted-Date: | 14 Apr 2006 14:13:38 EDT |

All,

One of the reasons given for why goto statements should

not be used is that their use can result in a function's

control flow graph being irreducible.

Irreducible? So what you say. Well some (many?) graph

transformation algorithms only work for reducible graphs.

Why do we need to perform graph transformations? Well, they

are used by various optimization and static analysis algorithms.

Why don't these algorithms use one of the iterative techniques

(which don't require reducible graphs)? Inertia, the courses they

were taught only used the graph transformation algorithms, ...

Does anybody know of an optimization/code analysis algorithm

that cannot be implemented using an iterative technique?

Are any of the graph transformation approaches significantly

better than the equivalent iterative technique (ie, in storage

space requirements or running time)?

It seems to me that for practical purposes iterative techniques

are as good anything else out there.

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.