> No, that's all legal, if unusual -- I am glad to see you are exercising the
> "Argonne/Northwestern parallel netcdf" driver feature of NetCDF-4.
Oh wait, you mean pnetcdf (as opposed to Parallel I/O?). The
configuration [1] on the machine which claims the file is not netCDF
is
```
Configuration Summary:
-- Building Shared Libraries: ON
-- Building netCDF-4: ON
-- Building DAP Support: ON
-- Building Utilities: ON
Building Parallel NetCDF
-- Using pnetcdf: OFF
-- Using Parallel IO: ON
```
I actually do *not* want pnetcdf for reading, I would like to use the
Parallel IO feature built into netCDF itself. What option would I have
to set for this?
Cheers,
Nico
[1]
https://launchpadlibrarian.net/196524671/buildlog_ubuntu-utopic-amd64.netcdf_1%3A4.3.3~20150204-utopic1_UPLOADING.txt.gz
On Wed, Feb 4, 2015 at 11:09 PM, Rob Latham <robl@xxxxxxxxxxx> wrote:
>
>
> On 02/04/2015 02:45 PM, Nico Schlömer wrote:
>>>
>>> What's the value of the mode argument you are passing to
>>>
>>> nc_open_par(const char *path, int mode, MPI_Comm comm, MPI_Info
>>> info,
>>> int *ncidp);
>>
>>
>> The call is from [1] with
>> ```
>> NC_NOWRITE == 0
>> NC_SHARE == 2048
>> pariomode == 32768
>
>
> 32768 is 0x8000 or NC_PNETCDF
>>
>> ```
>> so eventually the mode argument is
>> ```
>> NC_NOWRITE|NC_SHARE|pariomode == 34816
>> ```
>> Looks suspicious?
>
>
> No, that's all legal, if unusual -- I am glad to see you are exercising the
> "Argonne/Northwestern parallel netcdf" driver feature of NetCDF-4.
>
> These flags will create a classic netcdf file.
>
> Why would such a file be unreadable on another system?
>
> Well, if the other system is also requesting NC_PNETCDF, it's possible the
> Unidata-NetCDF library was not built with pnetcdf support, though
> nc_open_par should just detect the format.
>
> ==rob
>
>
>>
>> Cheers,
>> Nico
>>
>>
>> [1]
>> https://github.com/trilinos/trilinos/blob/master/packages/seacas/libraries/exodus/cbind/src/ex_open_par.c#L180
>>
>> On Wed, Feb 4, 2015 at 6:00 PM, Russ Rew <russ@xxxxxxxxxxxxxxxx> wrote:
>>>
>>> Hi Nico,
>>>
>>> What's the value of the mode argument you are passing to
>>>
>>> nc_open_par(const char *path, int mode, MPI_Comm comm, MPI_Info
>>> info,
>>> int *ncidp);
>>>
>>> ?
>>>
>>> --Russ
>>>
>>> On Tue, Feb 3, 2015 at 3:25 PM, Nico Schlömer <nico.schloemer@xxxxxxxxx>
>>> wrote:
>>>>
>>>>
>>>> Hi all,
>>>>
>>>> I'm trying to read the file
>>>> ```
>>>> http://win.ua.ac.be/~nschloe/other/pacman.e
>>>> ```
>>>> in parallel with `nc_open_par` [1]. This works on one machine, but
>>>> returns the error code `-51` ("Not a netCDF file") on another.
>>>> Could someone clarify this?
>>>>
>>>> Cheers,
>>>> Nico
>>>>
>>>>
>>>> [1]
>>>>
>>>> https://github.com/Unidata/netcdf-c/blob/master/libdispatch/dparallel.c#L38
>>>>
>>>> _______________________________________________
>>>> netcdfgroup mailing list
>>>> netcdfgroup@xxxxxxxxxxxxxxxx
>>>> For list information or to unsubscribe, visit:
>>>> http://www.unidata.ucar.edu/mailing_lists/
>>>>
>>>
>>
>> _______________________________________________
>> netcdfgroup mailing list
>> netcdfgroup@xxxxxxxxxxxxxxxx
>> For list information or to unsubscribe, visit:
>> http://www.unidata.ucar.edu/mailing_lists/
>>
>
> --
> Rob Latham
> Mathematics and Computer Science Division
> Argonne National Lab, IL USA
>
>
> _______________________________________________
> netcdfgroup mailing list
> netcdfgroup@xxxxxxxxxxxxxxxx
> For list information or to unsubscribe, visit:
> http://www.unidata.ucar.edu/mailing_lists/
