[MITgcm-support] 1D optimisation / optimization
nrogan at geomar.de
Wed May 27 09:46:43 EDT 2015
That's great. I'll see how it goes.
> On 30 Jan 2015, at 17:44, Martin Losch <Martin.Losch at awi.de> wrote:
> Hi Nick,
> many of your packages can be specified with macros, see pkg/pkg_groups
> your list would be simpler like this:
> # timeave # although I would not use that anymore, use diagnostics instead
> mnc # if you want netcdf output
> for a 1D configuration you probably do not want any advection, but vertical mixing. The advection can be turned off in data:
> tempAdvection = .FALSE.,
> saltAdvection = .FALSE.,
> momAdvection = .FALSE.,
> # actually you may want this: momStepping = .FALSE., to turn off all momentum time stepping (but this will offset the mixed layer model KPP, if you use it, because no vertical shear is allowed anymore, so I would leave it as TRUE)
> I don’t know how to turn of tracers advection, maybe not possible?
> I would still use an overlap of 1 (I think that’s actually necessary).
>> On 30 Jan 2015, at 16:21, Nicholas Rogan <nrogan at geomar.de> wrote:
>> Sorry, I should add that for now I’m happy to ignore advection, but would like vertical transport and diffusive fluxes to be active. I adapted from the 1D ice column verification scripts so there is some external forcing running. I tried to disable this but it flashed up with some errors I didn’t understand when trying to run. I have closed boundaries and 9 tracers which are all active in the DIC package.
>> My list of packages at compile time is here:
>> And at runtime I specify:
>> useKPP = .TRUE.,
>> # useCAL = .FALSE.,
>> useEXF = .TRUE.,
>> useSEAICE = .FALSE.,
>> useDiagnostics = .TRUE.,
>> # useECCO = .TRUE.,
>> useGrdchk = .FALSE.,
>> useMNC = .TRUE.,
>> useGCHEM = .TRUE.,
>> Sorry for omitting this first time round!
>>> On 30 Jan 2015, at 16:11, Nicholas Rogan <nrogan at geomar.de> wrote:
>>> Hi everyone,
>>> I was wondering if anyone had any advice on optimising model setups?
>>> I’m currently running a 1D water column that i have adapted from a much larger global simulation setup that I worked on during my PhD. I had a lot of custom code that I wanted to retain so I have ended up with a bit of a weird hybrid thing. I’m competent when running and adapting code but a little sketchy when building package layers from scratch. :)
>>> Does anyone have any advice on establishing what packages are necessary and what can be trimmed when moving to a 1D setup? I am using pTracers, DIC and whole bunch of other packages, which obviously have a lot of hefty routines, particularly the GAD routines,which are now largely redundant but could still be using a lot of model time.
>>> Any quick tips or links to prior threads? Couldn’t find much when searching site specific in google.
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
More information about the MITgcm-support