30 Sep 2005 02:02:04 -0400

Related articles |
---|

Coloring a graph with exact N colors dot (tzuchienchiu <at> gmail <dot> com<tzuchien.chiu@gm) (2005-09-22) |

Re: Coloring a graph with exact N colors touati@prism.uvsq.fr (TOUATI Sid) (2005-09-27) |

Re: Coloring a graph with exact N colors nkavv@skiathos.physics.auth.gr (Uncle Noah) (2005-09-30) |

Re: Coloring a graph with exact N colors touati@prism.uvsq.fr (TOUATI Sid) (2005-09-30) |

From: | "Uncle Noah" <nkavv@skiathos.physics.auth.gr> |

Newsgroups: | comp.compilers |

Date: | 30 Sep 2005 02:02:04 -0400 |

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

References: | 05-09-08405-09-134 |

Keywords: | registers |

Posted-Date: | 30 Sep 2005 02:02:03 EDT |

TOUATI Sid wrote:

*> This is called q-coloring in graph theory (q=N in your message). The*

*> complexity is not exponential as with optimal coloring, but the constant*

*> of the complexity is high (=N!). In practice, N <= 12 is an observed limit.*

Hi

Are you referring to a register allocation via graph coloring for

heterogeneous register set (i.e. different register files e.g. for

different data types or for some purpose that needs register

specialization?)

I mean: does q-coloring target such problem?

Nikolaos Kavvadias

aka "Uncle Noah"

Post a followup to this message

Return to the
comp.compilers page.

Search the
comp.compilers archives again.