[Paraview] caching images (WAS: messy legend)

Moreland, Kenneth kmorel at sandia.gov
Mon Sep 28 12:13:05 EDT 2009


Rather than preventing renderings for the specific case of delivering images from the server, I am an advocate of caching the rendering in all cases.  The interaction with scalar bars and other 2D widgets is a big reason for why I would like to see this.

We have talked about doing this for a long time (see bug #3924) and actually had an implementation for a brief period of time.  The problem was that we kept grabbing things that we did not render, for example other windows on top of the render window.  Now might be a good time to readdress this though.  We have already spent a lot of work getting the screen capturing working for saving screen shots.  Perhaps we could apply this work to caching images for every still render to save on re-rendering.

-Ken


On 9/27/09 1:11 PM, "Utkarsh Ayachit" <utkarsh.ayachit at kitware.com> wrote:

Well, don't forget the scalar bar needs to be rendered for tile
displays. I don't think a real fix will have anything to do with the
scalar bar, but more like when a widget interaction happens, instead
of triggering a render request on the view, it has to now decide which
renderer this widget the user is interacting with lies in (since in
case of 3D widgets like implicit plane widget which is rendered on the
server side we need the regular rendering) and if it's the 2D
renderer, then do some fancy stuff to simply render locally (except in
tile-display mode) in the 2D renderer while reusing the previously
delivered image for the 3D renderer. This may just be a tad too
complex to simply handle the interaction with the scalar bar widget
use-case.

A more general purpose solution would be a mechanism to avoid
server-side renders and deliveries unless the server-side scene in the
3D renderer has changed. This will also help avoiding the re-renders
when another window that covered the ParaView window is moved or such.
But there possibly, the complexity lies in truthfully determining that
the last rendered scene is still valid.

Utkarsh

2009/9/27 Berk Geveci <berk.geveci at kitware.com>:
> Hi Burlen,
>
> I totally agree. If possible, we should change it so that there is no
> server side render and image delivery when interacting with the scalar
> bar widget. I am guessing it will involve subclassing the scalar bar
> widget and changes to some other stuff. Utkarsh, what do you think?
>
> -berk
>
> 2009/9/27 burlen <burlen.loring at gmail.com>:
>> Hi Berk,
>>
>> I don't doubt you but I am surprised to learn that the scalar bar widget
>> is being rendered client side. In client-server mode as I interact with
>> the scalar bar widget the scene is re-render server side and an image is
>> shipped with every mouse move. To see that this is so put a print
>> statements in vtkPVDesktopDeliveryClient::ReceiveImageFromServer. Since
>> the scalar bar widget is rendered client side, is there really a need to
>> re-render and ship the image form the server side with each scalar bar
>> widget interaction? Preventing these server side render/image transfers
>> would certainly improve remote interactivity.
>>
>> By the way can you point me to where PV renders the scalar bar widget?
>>
>> Burlen
>>
>> Berk Geveci wrote:
>>> I believe that the legend is always rendered on the client so changing
>>> to remote render would probably have no effect.
>>>
>>> 2009/9/26 burlen <burlen.loring at gmail.com>:
>>>
>>>>> Compiling mesa3d is difficult to me.
>>>>>
>>>> The mesa autoconf files never have worked for me, but, the pre-packaged make
>>>> files always have. "make linux".
>>>>
>>>>
>>>>> If I run paraview on the server and display it on the server,then
>>>>> everything is ok. But if I use my laptop ,log into the server,then plot with
>>>>> paraview on the server and display the result on the laptop,then I get the
>>>>> bad legend again.
>>>>>
>>>> When you connect to the server from your laptop: Go to
>>>> "Edit->Settings->Render View->Server". See that "remote Render Threshold" is
>>>> checked and set the value to 0. That will make sure rendering happens on the
>>>> server and not on your laptop.
>>>>
>>>> linzhenhua wrote:
>>>>
>>>>> Hi,Thanks for your reply. I try to narrow the problem as follows. I wonder
>>>>> if I could get some other hints. Compiling mesa3d is difficult to me.
>>>>>
>>>>> I install linux on a PC today,let's call it server. If I run paraview on
>>>>> the server and display it on the server,then everything is ok. But if I use
>>>>> my laptop ,log into the server,then plot with paraview on the server and
>>>>> display the result on the laptop,then I get the bad legend again.
>>>>>
>>>>> Any hints?
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> 在2009-09-25,"Berk Geveci" <berk.geveci at kitware.com> 写道:
>>>>>
>>>>>> To verify whether your graphics driver/card is causing this issue, you
>>>>>> can compile Mesa (from mesa3d.org) yourself without any hardware
>>>>>> support and use it as your OpenGL. If you are still having this
>>>>>> problem, it is not your card.
>>>>>>
>>>>>> -berk
>>>>>>
>>>>>> 2009/9/25 linzhenhua <linzhenhua at 163.com>:
>>>>>>
>>>>>>> Hi,All
>>>>>>>
>>>>>>> I was bothering by this problem for several days. The legend is very
>>>>>>> bad.
>>>>>>>
>>>>>>> I tried two or three Fedora versions, 10,11,rawhide. For Fedora 11,I
>>>>>>> have
>>>>>>> other problem,namely paraview freeze the computer,so I don't notice
>>>>>>> whether
>>>>>>> this version also have this problem.
>>>>>>>
>>>>>>> For all version I tried, I test both the version coming with the distro
>>>>>>> and
>>>>>>> personally compiled version(I have tried both 3.4.0 and 3.6.1)
>>>>>>>
>>>>>>> Since I have tried different Fedora version,they should have different
>>>>>>> versions of graphic driver, I guess this is not the problem. Now I
>>>>>>> suspect
>>>>>>> whether this problem is related with the hardware. But I do not have
>>>>>>> any
>>>>>>> other problem with the graphic card.
>>>>>>>
>>>>>>> Can any one have any idea which is the next possible direction I should
>>>>>>> try
>>>>>>> to solve this problem?
>>>>>>>
>>>>>>>
>>>>>>> Best wishes.
>>>>>>>
>>>>>>>
>>>>>>> --
>>>>>>>
>>>>>>>
>>>>>>> ________________________________
>>>>>>> "中国制造",讲述中国60年往事
>>>>>>> _______________________________________________
>>>>>>> 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
>>>>>>>
>>>>>>>
>>>>>>>
>>>>> ------------------------------------------------------------------------
>>>>> "中国制造 ",讲述中国60年往事
>>>>> <http://news.163.com/madeinchina/index.html?from=mailfooter>
>>>>> ------------------------------------------------------------------------
>>>>>
>>>>> _______________________________________________
>>>>> 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
>>>>>
>>>>>
>>>>
>>
>>
>
_______________________________________________
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



   ****      Kenneth Moreland
    ***      Sandia National Laboratories
***********
*** *** ***  email: kmorel at sandia.gov
**  ***  **  phone: (505) 844-8919
    ***      web:   http://www.cs.unm.edu/~kmorel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20090928/632e4eff/attachment.htm>


More information about the ParaView mailing list