Changes between Version 9 and Version 10 of 2017WP/ENHANCE-05_Gurvan-RK3


Ignore:
Timestamp:
2017-07-25T10:07:38+02:00 (4 years ago)
Author:
gm
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • 2017WP/ENHANCE-05_Gurvan-RK3

    v9 v10  
    55|| '''Action''' || ENHANCE-5_Gurvan — RK-3 time stepping || 
    66|| '''PI(S)''' || Gurvan Madec || 
    7 {{{#!td 
    8 '''Digest''' 
    9 }}} 
    10 {{{#!td 
    11 Introduce an optional RK3 time-stepping scheme. The scheme will: 
     7 
     8{{{#!td '''Digest''' }}} {{{#!td Introduce an optional RK3 time-stepping scheme. The scheme will: 
    129 
    1310 1. be a valuable alternative to the current Modified Leap-Frog (MLF, Leclair and Madec OM2009) scheme, 
     
    1714 1. make feasible to properly implement an implicit computation of top and bottom friction (properly means consistent with the ssh time splitting scheme) 
    1815 
    19 As such a change will impact many module of OPA, it has been preferred to develop the RK3 version in a separate directory, RK3_SRC, put at the same level as OPA_SRC and being a copy of OPA_SRC as starting point.  
     16As such a change will impact many module of OPA, it has been preferred to develop the RK3 version in a separate directory, RK3_SRC, put at the same level as OPA_SRC and being a copy of OPA_SRC as starting point. 
    2017 
    2118In addition to the RK3 scheme, a Forward-Bacward Scheme (FBS) will be introduced. The goals associated with FBS are the following: 
    2219 
    23    ''(i)'' FB scheme is also a two levels time-stepping scheme but much simpler than a RK3. Its implementation will prepare all the required environment for the RK3 scheme, while been much easier to develop and test. 
     20  '' (i)'' FB scheme is also a two levels time-stepping scheme but much simpler than a RK3. Its implementation will prepare all the required environment for the RK3 scheme, while been much easier to develop and test. 
    2421 
    25    ''(ii)'' FB scheme may be a valuable alternative for low resolution simulations (in particular ORCA 2° and 1°) which do not necessarily need a high order time-stepping scheme, although FB scheme requires to used UBS scheme on advection of both dynamics and tracers. 
     22  '' (ii)'' FB scheme may be a valuable alternative for low resolution simulations (in particular ORCA 2° and 1°) which do not necessarily need a high order time-stepping scheme, although FB scheme requires to used UBS scheme on advection of both dynamics and tracers. 
    2623 
    27 In practice the work is to : 
     24In practice the work is to :[[BR]]• start from the trunk, updated with HPC09 branch (i.e. ZDF modifications)[[BR]]• create RK3_SRC directory a a copy of OPA_SRC[[BR]]• implement the FB scheme in a stp_fbs routine (found either in step.F90 or in stpfbs.F90), [[BR]]   implying changes in most of SBC, TRA, DYN and ZDF modules[[BR]]• implement the RK3 scheme in a stp_rk3 routine (found either in step.F90 or stprk3.F90) 
    2825 
    29  * start from the trunk, updated with HPC09 branch (i.e. ZDF modifications) 
     26}}} |- 
    3027 
    31  * create RK3_SRC directory a a copy of OPA_SRC 
    32  
    33  * implement the FB scheme in a stp_fbs routine (found either in step.F90 or in stpfbs.F90), implying changes in most of SBC, TRA, DYN and ZDF modules 
    34  
    35  * implement the RK3 scheme in a stp_rk3 routine (found either in step.F90 or stprk3.F90) 
    36 }}} 
    37 |- 
    3828|| '''Dependencies''' || no || 
    3929|| '''Target''' || 2017 merge party ||