| 41 | However, it seems for WD_def we do not vary lrefine_max. |

| 42 | Chris needs to speak to his work colleague to understand this |

| 43 | more thoroughly. For the time being here is the email |

| 44 | describing the weak scaling for WD_Def: |

| 45 | {{{ |

| 46 | On Wed, 1 Oct 2008, Chris Daley wrote: |

| 47 | |

| 48 | > Also can you please explain why we change: |

| 49 | > r_match |

| 50 | > refine_uni_radius |

| 51 | > in each flash.par? |

| 52 | |

| 53 | THAT is how the size of the problem (i.e., muber of blocks) was |

| 54 | changed. Note the lrefine_min and lrefine_max do not change! |

| 55 | |

| 56 | > How do you know the correct value? |

| 57 | |

| 58 | It required quite a bit of work to come up with these values. |

| 59 | |

| 60 | They were chosen to give problem sizes of (roughly) 100 * 64 * 2^n |

| 61 | total blocks, for integer n, n is the number in scaling<n> I believe. |

| 62 | |

| 63 | Klaus |

| 64 | }}} |