[MITgcm-devel] something to report
Gael Forget
gforget at MIT.EDU
Mon Oct 21 22:57:04 EDT 2013
hi Manfredi,
I am copying the answer I gave you last time on support :
>> If so, is this compatible with output in netcdf format or does it work only with binary files ?
> I dont use mnc. But to the extent that mnc orders the data
> as mds does, all you need to do is use rdmnc.m to read the output
> data to memory, then call convert2gcmfaces.m in analogy with
> rdmds2gcmfaces.m
Of course there is always the option of using binary outputs.
> And just one more thing: not sure how to modify/adapt Gael's script
> for mnc + rdmnc.m output if you are using blank-tiles.
Hi Jean Michel,
I think the only piece that is needed is an adequate IO routine.
I think I recall that mnc (i.e. Ed) did not follow the naming
conventions from mds, which I use in matlab. That could get
in the way. I also think that mnc used to add one raw
or column to some arrays. It is possible that both issues have
long been resolved, although you probably know better than me.
I have no idea how blank tiles are handled by mnc, which I dont use.
If there are still such issues, and if Manfredi wants to use my
matlab codes, then my advice is to use mds for IO.
Cheers,
Gael
On Oct 21, 2013, at 9:33 PM, Jean-Michel Campin wrote:
> Hi Manfredi,
>
> On Mon, Oct 21, 2013 at 03:41:35PM -0400, Manfredi Manizza wrote:
>>
>> Jean-Michel,
>>
>> Just to clarify, is the nc file that is a bit larger
>> than the file in binray format whihc I think is the default..
>
>>> and with some blank tiles ?
> Is it the case ?
>
>> I tried to impelment your sugggestion,
>> It blows the model up becuase it looks for a 6th tile
>> that is not there given that the domain is set for 5 tiles only, see
>> below from log file.
>
> Are you sure that you only changed W2_useE2ioLayOut
> from TRUE (= the default) to FALSE ?
> Because I've just checked and W2_useE2ioLayOut has no effect on loading
> the grid-files so I am little bit surprised (if reading grid-files
> was working before with W2_useE2ioLayOut=T it should work the same way now
> with W2_useE2ioLayOut=F).
>
> And just one more thing: not sure how to modify/adapt Gael's script
> for mnc + rdmnc.m output if you are using blank-tiles.
>
> Cheers,
> Jean-Michel
>
>> I will try to see if I use the grid nc files generated by the run
>> what happens to the matlab program that converts the output into
>> lat/lon.
>> It may work, not sure yet...
>>
>> forrtl: severe (29): file not found, unit 9, file
>> /nobackupp8/mmanizza/llc90/run_nocost/tile006.mitgrid
>> Image PC Routine Line Source
>> libintlc.so.5 00002AAAAC49CF6A Unknown Unknown
>> Unknown
>> libintlc.so.5 00002AAAAC49BAE5 Unknown Unknown
>> Unknown
>> libifcore.so.5 00002AAAAB62DE12 Unknown Unknown
>> Unknown
>> libifcore.so.5 00002AAAAB5A35CB Unknown Unknown
>> Unknown
>> libifcore.so.5 00002AAAAB5A2CDE Unknown Unknown
>> Unknown
>> libifcore.so.5 00002AAAAB5C0848 Unknown Unknown
>> Unknown
>> mitgcmuv 00000000005A2B00 mds_facef_read_rs 1081
>> mdsio_facef_read.f
>> mitgcmuv 0000000000739A5C ini_curvilinear_g 3482
>> ini_curvilinear_grid.f
>> mitgcmuv 0000000000741804 ini_grid_ 3330
>> ini_grid.f
>> mitgcmuv 000000000075D1F1 initialise_fixed_ 1904
>> initialise_fixed.f
>> mitgcmuv 00000000007721F2 the_model_main_ 3119
>> the_model_main.f
>> mitgcmuv 00000000006DCF82 MAIN__ 4010 main.f
>> mitgcmuv 0000000000409ABC Unknown Unknown
>> Unknown
>> libc.so.6 00002AAAACA1CBC6 Unknown Unknown
>> Unknown
>> mitgcmuv 00000000004099B9 Unknown Unknown
>> Unknown
>>
>>
>>
>> Manfredi
>>
>>
>>
>>
>> On Sun, 20 Oct 2013, Jean-Michel Campin wrote:
>>
>>> Hi Manfredi,
>>>
>>> I don't know what is your "data.exch2" parameter file,
>>> but I assume that you have:
>>> W2_useE2ioLayOut=T (this is the default)
>>> and:
>>> W2_mapIO = 1 (this is not the default, which is -1)
>>> and with some blank tiles ?
>>>
>>> Now, the option "W2_useE2ioLayOut" has not been implemented
>>> in pkg/mnc, and this could explain that you have different
>>> size between mnc + rdmnc.m and mdsio + rdmds.m
>>>
>>> I think that if you set:
>>> W2_useE2ioLayOut=F
>>> the plain binary output file (pkg/mdsio + rdmds.m) should have the
>>> same size the current output that you get with mnc + rdmnc.m
>>>
>>> Cheers,
>>> Jean-Michel
>>>
>>> On Sat, Oct 19, 2013 at 12:21:41PM -0400, Manfredi Manizza wrote:
>>>>
>>>> Hi,
>>>>
>>>> I am reporting a feture I found comparing the output
>>>> of the ECCO lat-lon-cap global set-up provided by Patrick to me.
>>>>
>>>> The dimensions of the files are different depending on the format.
>>>> If in netcdf :
>>>>
>>>> S =
>>>>
>>>> attributes: [1x1 struct]
>>>> iters_from_file: [12x1 double]
>>>> si_AREAtave: [90x1082x12 double]
>>>>
>>>> If in binary :
>>>>
>>>>
>>>>> more pickup_seaice.0000008772.meta
>>>> nDims = [ 2 ];
>>>> dimList = [
>>>> 90, 1, 90,
>>>> 1170, 1, 1170
>>>> ];
>>>> dataprec = [ 'float64' ];
>>>> nrecords = [ 6 ];
>>>> timeStepNumber = [ 8772 ];
>>>> timeInterval = [ 3.157920000000E+07 ];
>>>> nFlds = [ 6 ];
>>>> fldList = {
>>>> 'siTICE ' 'siAREA ' 'siHEFF ' 'siHSNOW ' 'siUICE ' 'siVICE '
>>>> };
>>>>
>>>>
>>>> The grid quantities that I have , when loaded, have the dimensions
>>>> of 90 by 1170.
>>>>
>>>> In fact i can use the functions of gcmfaces only using the binary
>>>> files...
>>>>
>>>> I am not sure if there is something I am missing or if
>>>> there is something wrong in the code. I just felt I needed to
>>>> send this out.
>>>>
>>>> Thanks in advance for your attention.
>>>>
>>>> Cheers
>>>> Manfredi
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> MITgcm-devel mailing list
>>>> MITgcm-devel at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>>
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel
More information about the MITgcm-devel
mailing list