[CIG-SHORT] Fault Output

[Brad Aagaard]

Bobby,

To run in serial, just use --nodes=1.

Brad

On 3/5/13 12:38 PM, Brad Aagaard wrote:
> Bobby,
>
> To run in parallel with 2 processes (you can try this on your laptop
> too) all you do is run with --nodes=2.
>
> For more info on running on a cluster, see the "Configuration on a
> Cluster" in the "Installation and Getting Help" chapter of the manual.
> Also see the PyLith Wiki
> (http://www.geodynamics.org/cig/community/workinggroups/short/workarea/pylith-wiki)
> for additional tips.
>
> Brad
>
> On 3/5/13 12:32 PM, bok10 at pitt.edu wrote:
>> Is there a way to adjust the input to have it run in serial?
>>
>> Bobby
>>
>>
>>> Bobby,
>>>
>>> It wasn't clear from your description if you are comparing serial and
>>> parallel runs on the cluster or a serial run on a laptop with a parallel
>>> run on the cluster. I think the first step is to verify that a serial
>>> run on the cluster matches a serial run on your laptop. The next thing I
>>> would do is to test serial and parallel runs for simple examples like
>>> those in examples/3d/hex8 (or some of the other provided examples). You
>>> want to isolate the specific conditions/parameters when things may be
>>> going wrong.
>>>
>>> Regards,
>>> Brad
>>>
>>>
>>> On 3/5/13 10:16 AM, BOK10 at pitt.edu wrote:
>>>> Hi Brad,
>>>>
>>>> I clicked on the information tab and all the time-steps have a range of
>>>> 0
>>>> for the faults.
>>>>
>>>> I am using the PyLith binary on the cluster, and I am working on 1
>>>> computer node (which far outpowers my laptop).
>>>>
>>>> I'm writing the files as vtk formats, and I don't believe I am using
>>>> the
>>>> HDF5 writer.
>>>>
>>>> Bobby
>>>>
>>>>> Bobby,
>>>>>
>>>>> Click on the fault dataset and then on the Info tab. You will see the
>>>>> range of values for each of the fields. Then jump to the final time
>>>>> step. This is a more robust way of checking the range of values than
>>>>> viewing an image.
>>>>>
>>>>> Are you using a PyLith binary on the cluster? The binary may work on 1
>>>>> compute node, but in order to run on more than 1 compute node, you
>>>>> need
>>>>> to build from source because the binary contains MPI consistent with a
>>>>> desktop machine. For a cluster building from source (using the pylith
>>>>> installer) insures that the MPI being used is appropriate for your
>>>>> hardware.
>>>>>
>>>>> Another possibility is that you are using the HDF5 writer, but HDF5
>>>>> was
>>>>> not built for parallel I/O (the pylith installer should do this
>>>>> automatically). If that is the case, then the output from all
>>>>> processors
>>>>> is not being collected into the file.
>>>>>
>>>>> Regards,
>>>>> Brad
>>>>>
>>>>>
>>>>> On 3/5/13 9:59 AM, BOK10 at pitt.edu wrote:
>>>>>> I'm running my model on a cluster, and everything outputs fine,
>>>>>> except
>>>>>> for
>>>>>> the faults. In paraview I rescale to the temporal range, but it tells
>>>>>> me
>>>>>> my faults have 0 slip, but in the past they have had slip. Any
>>>>>> suggestions/hints?
>>>>>>
>>>>>> Bobby
>>>>>>
>>>>>> _______________________________________________
>>>>>> CIG-SHORT mailing list
>>>>>> CIG-SHORT at geodynamics.org
>>>>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> CIG-SHORT mailing list
>>>>> CIG-SHORT at geodynamics.org
>>>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> CIG-SHORT mailing list
>>>> CIG-SHORT at geodynamics.org
>>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>>>
>>>
>>> _______________________________________________
>>> CIG-SHORT mailing list
>>> CIG-SHORT at geodynamics.org
>>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>>
>>
>>
>> _______________________________________________
>> CIG-SHORT mailing list
>> CIG-SHORT at geodynamics.org
>> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>>
>