Difference between revisions of "UCVM Review"

From SCECpedia
Jump to navigationJump to search
Line 11: Line 11:
 
* ucvmc Installation instructions need to be updated to describe how to create mpi version if running on usc hpc.
 
* ucvmc Installation instructions need to be updated to describe how to create mpi version if running on usc hpc.
  
 +
== notes on basin data =
 +
<pre>
 +
 +
When I receive values from Scott I will check them by plotting and comparing with previous versions. I need separate files for Z1.0 and Z2.5 in meters.
 +
 +
The binary files are 4 byte float, little endian, fast XY, no rotation. Here are the formulas for lat/lon at a given x/y index:
 +
 +
lon = xIndex * gridSpacing + minLon
 +
lat = yIndex * gridSpacing + minLat
 +
 +
and for calculating the file position for a given xIndex/yIndex:
 +
 +
filePos = 4 * (nx * yIndex + xIndex)
 +
 +
The first data point is bottom left (at minX, minY).
 +
 +
Kevin
 +
 +
On 7/11/2017 5:44 PM, Philip J. Maechling wrote:
 +
> Kevin,
 +
>  How are you checking the results you get form Scott? Maybe this cross check with UCVMC is not needed.
 +
>  Also, what parameters are you extracting?
 +
>  Also what is binary regular mesh format? binary 32bit floats? What order? fast x, y? or other?
 +
> Thanks,
 +
> Phil M.
 +
> On 7/11/17 7:26 PM, Kevin Milner wrote:
 +
>> Phil,
 +
>>
 +
>> Scott is working on generating these files for me with the python version of UCVM. I need binary regular meshes, which David generated for me from both codebases. If you want to proceed in parallel and you find the regular mesh generation utility in UCVMC, here are the parameters that specify the regions:
 +
>>
 +
>> CCA:
 +
>>    minLat = 33.35
 +
>>    minLon = -123
 +
>>    nx = 1551
 +
>>    ny = 1201
 +
>>    gridSpacing = 0.005
 +
>>
 +
>> 4.26:
 +
>>    minLat = 31;
 +
>>    minLon = -121;
 +
>>    nx = 1701;
 +
>>    ny = 1101;
 +
>>    gridSpacing = 0.005;
 +
>>
 +
>> Thanks,
 +
>> Kevin
 +
>>
 +
>> On 07/11/2017 05:12 PM, Philip J. Maechling wrote:
 +
>>> Scott and/or Kevin,
 +
>>>
 +
>>>  Christine reminded me that we want to re-extract CVM parameters. Can you send me the points of interest? And the parameter of interest for each point?
 +
>>>
 +
>>>  Although the last released version of UCVMC does not include CCA06, there may be some overlapping points I can try to query using UCVMC. I think the UCVMC software provides specialized queries that the UCVM does not have, queries that provide first occurrence, or last occurrence of the Vs value, as an option.
 +
>>>
 +
>>>  Also, we want to make sure that the changes to UCVM to support the depth of last occurance, are preserved in version control. Can either of your determine where those changes are saved in our source code?
 +
>>>
 +
>>>  Scott's version of UCVMC must support CCA06, since it was used to create the CyberShake 17.3 meshes, so presumably we could back engineer the addition of that model, if needed.
 +
>>>
 +
>>> Phil M.
 +
 +
 +
</pre>
  
 
== Introductory Presentations ==
 
== Introductory Presentations ==

Revision as of 05:01, 21 October 2017

These links are information about the current UCVM software.

basin_query z1000 first occurence or second occurence test

  • goal of this test is to evaluate the difference between results for the z1000 test using difference rules for which z1000 depth to return, the first occurence of the specified Vs, or the second occurence, or something else. We need to document the algorithms available in ucvmc, and show example return values for each.

basin_query serial and mpi tests

  • goal of this test is to evaluate the results from basin_query and basin_query_mpi to demonstrate the produce the same results.

mpi ucvm2mesh tests

  • goal of this test is to construct awp test meshes using both ucvm2mesh and ucvm2mesh_mpi. Then we want to compare the resulting meshes and make sure they are identical
  • ucvmc Installation instructions need to be updated to describe how to create mpi version if running on usc hpc.

= notes on basin data


When I receive values from Scott I will check them by plotting and comparing with previous versions. I need separate files for Z1.0 and Z2.5 in meters.

The binary files are 4 byte float, little endian, fast XY, no rotation. Here are the formulas for lat/lon at a given x/y index:

lon = xIndex * gridSpacing + minLon
lat = yIndex * gridSpacing + minLat

and for calculating the file position for a given xIndex/yIndex:

filePos = 4 * (nx * yIndex + xIndex)

The first data point is bottom left (at minX, minY).

Kevin

On 7/11/2017 5:44 PM, Philip J. Maechling wrote:
> Kevin,
>  How are you checking the results you get form Scott? Maybe this cross check with UCVMC is not needed.
>  Also, what parameters are you extracting?
>  Also what is binary regular mesh format? binary 32bit floats? What order? fast x, y? or other?
> Thanks,
> Phil M.
> On 7/11/17 7:26 PM, Kevin Milner wrote:
>> Phil,
>>
>> Scott is working on generating these files for me with the python version of UCVM. I need binary regular meshes, which David generated for me from both codebases. If you want to proceed in parallel and you find the regular mesh generation utility in UCVMC, here are the parameters that specify the regions:
>>
>> CCA:
>>     minLat = 33.35
>>     minLon = -123
>>     nx = 1551
>>     ny = 1201
>>     gridSpacing = 0.005
>>
>> 4.26:
>>     minLat = 31;
>>     minLon = -121;
>>     nx = 1701;
>>     ny = 1101;
>>     gridSpacing = 0.005;
>>
>> Thanks,
>> Kevin
>>
>> On 07/11/2017 05:12 PM, Philip J. Maechling wrote:
>>> Scott and/or Kevin,
>>>
>>>  Christine reminded me that we want to re-extract CVM parameters. Can you send me the points of interest? And the parameter of interest for each point?
>>>
>>>   Although the last released version of UCVMC does not include CCA06, there may be some overlapping points I can try to query using UCVMC. I think the UCVMC software provides specialized queries that the UCVM does not have, queries that provide first occurrence, or last occurrence of the Vs value, as an option.
>>>
>>>  Also, we want to make sure that the changes to UCVM to support the depth of last occurance, are preserved in version control. Can either of your determine where those changes are saved in our source code?
>>>
>>>   Scott's version of UCVMC must support CCA06, since it was used to create the CyberShake 17.3 meshes, so presumably we could back engineer the addition of that model, if needed.
>>>
>>> Phil M.


Introductory Presentations

Related Entries