[Paraview] using vtkInformation in 3.10

Berk Geveci berk.geveci at kitware.com
Fri May 6 09:00:43 EDT 2011


Hi Burlen,

There is support for this in the pipeline. In the appropriate pipeline pass
(REQUEST_INFORMATION, REQUEST_UPDATE_EXTENT ro
REQUEST_DATA), add the keys you want copied to the request object
(the first argument to ProcessRequest) under the KEYS_TO_COPY key.
This will tell the executive to propagate those keys all the way upstream
or downstream depending on the pass. See vtkExecutive::CopyDefaultInformation()
for the implementation if something is not working right.

Best,
-berk

On Thu, May 5, 2011 at 4:09 PM, burlen <burlen.loring at gmail.com> wrote:
> Hi Utkarsh,
>
> Yes, I had connected the reader or source directly to the filter with no
> other filters intervening.
>
> I have used the keys for a number of things. In addition to passing meta
> data I also implemented dynamic load balancing in a couple of algorithms by
> passing vtkObjects. For example the demand loading strategy in our stream
> tracer expects to receive a reader object in the pipeline information. The
> reader object also implements a LRU block cache to reduce I/O footprint.
>
> We've used a similar strategy to generate seed cell geometry on demand,
> where our volume source passes a cell generator object downstream in the
> pipeline information which the stream tracer uses to access any cell in any
> order (dynamic load balancing part) with a minimum communication and memory
> footprint. We did that because we  ran our of memory when both the source
> and the stream tracer had a copy of the seed geometry.
>
> The use case I'm describing is the generation of topographic maps of the
> magnetosphere which in the 3D case requires 10's of millions of stream
> traces and the seed geometry is colored in the mapper output. Dynamic load
> balancing of the stream tracing provides a nice speed up.
>
> The vtkPVPostFilter looks to add nice new features to extracting components
> form vectors and converting cell to point data and so on. I do see why PV
> can't copy keys from input to output blindly.  But, do you think
> vtkPVPostFilter could copy the information from input to output as well? It
> doesn't after all modify the existing arrays or geometry (if I'm reading it
> right it only adds arrays if they are requested).? Or is there another work
> around that would get our filters working again?
>
> Thanks
> Burlen
>
> On 05/05/2011 05:51 AM, Utkarsh Ayachit wrote:
>>
>> Burlen,
>>
>> Keys are not propagated by pipeline by default. So your approach would
>> only work when there's no other filter between the reader and your
>> filters. With ParaView 3.10, vtkPVPostFilter get inserted always into
>> the pipeline after every algorithm. As a result your keys maybe
>> getting eaten up by the post-filter.
>>
>> What keys are these?
>>
>> Utkarsh
>>
>> On Wed, May 4, 2011 at 10:35 PM, Burlen Loring<bloring at lbl.gov>  wrote:
>>>
>>> Hi All,
>>>
>>> I have a reader that gets the information object from it's output and
>>> adds
>>> some keys for downstream filters to use. This recently stopped working
>>> with
>>> version 3.10. Can anyone shed some light on what might have caused the
>>> change in behavior in version 3.10? Was this a deliberate change?
>>>
>>> Thanks
>>> Burlen
>>>
>>> _______________________________________________
>>> Powered bywww.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Please keep messages on-topic and check the ParaView Wiki at:
>>> http://paraview.org/Wiki/ParaView
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.paraview.org/mailman/listinfo/paraview
>>>
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ParaView Wiki at:
> http://paraview.org/Wiki/ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://www.paraview.org/mailman/listinfo/paraview
>


More information about the ParaView mailing list