Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Andy Smith]

Ken,

I tested your changes and they are working for my more complicated
"production" case.  I am currently running more tests and will update you
if anything goes wrong.

I'm still adjusting my lighting settings from 4.4.0 to 5.x.x to get similar
images between the two versions.  In the past with 4.4.0 I had noticed that
some surface LIC images would seemingly randomly get darker.
Perhaps that is what you are seeing as well.

-Andy

On Thu, Feb 9, 2017 at 11:47 AM, Ken Martin  wrote:

> Thanks so much for the example. I have a VTK topic that I believe will fix
> the issue here:
>
> https://gitlab.kitware.com/vtk/vtk/merge_requests/2463
>
> It fails three of the current paraview LIC tests but with images that
> still look reasonable (although one looks dark to me). The changes only
> impact two files. Your example no longer hangs with these changes. If you
> have a chance to test it on your real data or other cases that would help
> me feel more confident about the change.
>
> Thanks
> Ken
>
>
>
>
> On Wed, Feb 8, 2017 at 9:21 AM, Andy Smith  wrote:
>
>> Ken,
>>
>> I've attached a small example file and script that demonstrates the
>> problem.
>>
>> The example dataset is a multiblock unstructured grid with two blocks.
>> The script attempts to create two images of slices through the dataset.  At
>> the first slice position data from both blocks are in the slice.  At the
>> second slice position only data from one block is part of the slice.  This
>> works in serial and parallel (two processors) using pvbatch when not using
>> surface LIC and in serial when using surface LIC with the latest ParaView
>> from the git repo.  In parallel, the surface LIC hangs when creating the
>> second image.  This script works in parallel with surface LIC when using
>> version 4.4.0.
>>
>> As an aside, I call the function to ensure that the surface LIC plugin is
>> loaded at the beginning of the script.  When I do this I get warnings about
>> "Replacing existing representation for key: Surface LIC".  I assume this is
>> because in my build I turned on the option to load the surface LIC plugin
>> by default.  Is there a way in Python to check if a plugin is loaded and
>> skip this step if that is the case?
>>
>> Let me know if I can provide any additional information.
>>
>> -Andy
>>
>> On Tue, Feb 7, 2017 at 8:32 AM, Ken Martin 
>> wrote:
>>
>>> Thanks Andy, just to make sure I understand the issue. Is the case that
>>> one processor has no data, or is it that it has data, but it is outside the
>>> current view?
>>>
>>> Thanks
>>> Ken
>>>
>>> On Thu, Feb 2, 2017 at 5:14 PM, Andy Smith  wrote:
>>>
 I've tested this with the latest ParaView and it does seem to run much
 faster.
 I am having a problem when running in parallel though.  If one
 processor does not have anything to display, the rendering hangs.
 Related to that, what routine actually calls
 vtkCompositeLICHelper::RenderPiece?
 The process that does not have anything to display never gets to that
 function but the other processes do.  In that function there are MPI calls
 where the rendering hangs.

 As a reply to Chuck:  I can use EGL in some instances but unfortunately
 not all of our machine have GPUs so software rendering is still required.

 Thanks again for all the help.

 On Mon, Jan 23, 2017 at 12:45 PM, Ken Martin 
 wrote:

> There is a VTK topic now that should fix this regression. With this
> change OpenGL2 LIC seems to run much faster (at least 10X).
>
> https://gitlab.kitware.com/vtk/vtk/merge_requests/2419
>
> Hopefully it will test out OK and get merged into VTK and then merged
> into ParaView.
>
> On Mon, Dec 12, 2016 at 11:38 AM, Andy Smith 
> wrote:
>
>> Ken,
>>
>> You are correct in your assumption about my dataset; it is multiblock
>> unstructured data.  Adding a MergeBlocks filter to my slice does improve
>> the 5.2.0 with OpenGL2 significantly.
>> With that change the performance of 5.2.0 with OpenGL2 is on the same
>> order of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0
>> variants are still slower than 4.4.0.
>>
>> I realized after posting that my comparison is not exactly apples to
>> apples - my build of 5.2.0 required updates to mesa (13.0.0) compared 
>> with
>> the version I used for 4.4.0 (11.2.0).
>> I will recompile 4.4.0 with the same mesa version I used with 5.2.0
>> and report back.
>>
>> Thank you for your input.
>>
>> -Andy
>>
>> On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin 
>> wrote:
>>
>>> I have not had time to look into this, but my quick guess is that
>>> you have a multiblock dataset with a lot of blocks. Can you try merging 
>>> 

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Ken Martin]

Thanks so much for the example. I have a VTK topic that I believe will fix
the issue here:

https://gitlab.kitware.com/vtk/vtk/merge_requests/2463

It fails three of the current paraview LIC tests but with images that still
look reasonable (although one looks dark to me). The changes only impact
two files. Your example no longer hangs with these changes. If you have a
chance to test it on your real data or other cases that would help me feel
more confident about the change.

Thanks
Ken




On Wed, Feb 8, 2017 at 9:21 AM, Andy Smith  wrote:

> Ken,
>
> I've attached a small example file and script that demonstrates the
> problem.
>
> The example dataset is a multiblock unstructured grid with two blocks.
> The script attempts to create two images of slices through the dataset.  At
> the first slice position data from both blocks are in the slice.  At the
> second slice position only data from one block is part of the slice.  This
> works in serial and parallel (two processors) using pvbatch when not using
> surface LIC and in serial when using surface LIC with the latest ParaView
> from the git repo.  In parallel, the surface LIC hangs when creating the
> second image.  This script works in parallel with surface LIC when using
> version 4.4.0.
>
> As an aside, I call the function to ensure that the surface LIC plugin is
> loaded at the beginning of the script.  When I do this I get warnings about
> "Replacing existing representation for key: Surface LIC".  I assume this is
> because in my build I turned on the option to load the surface LIC plugin
> by default.  Is there a way in Python to check if a plugin is loaded and
> skip this step if that is the case?
>
> Let me know if I can provide any additional information.
>
> -Andy
>
> On Tue, Feb 7, 2017 at 8:32 AM, Ken Martin  wrote:
>
>> Thanks Andy, just to make sure I understand the issue. Is the case that
>> one processor has no data, or is it that it has data, but it is outside the
>> current view?
>>
>> Thanks
>> Ken
>>
>> On Thu, Feb 2, 2017 at 5:14 PM, Andy Smith  wrote:
>>
>>> I've tested this with the latest ParaView and it does seem to run much
>>> faster.
>>> I am having a problem when running in parallel though.  If one processor
>>> does not have anything to display, the rendering hangs.
>>> Related to that, what routine actually calls
>>> vtkCompositeLICHelper::RenderPiece?
>>> The process that does not have anything to display never gets to that
>>> function but the other processes do.  In that function there are MPI calls
>>> where the rendering hangs.
>>>
>>> As a reply to Chuck:  I can use EGL in some instances but unfortunately
>>> not all of our machine have GPUs so software rendering is still required.
>>>
>>> Thanks again for all the help.
>>>
>>> On Mon, Jan 23, 2017 at 12:45 PM, Ken Martin 
>>> wrote:
>>>
 There is a VTK topic now that should fix this regression. With this
 change OpenGL2 LIC seems to run much faster (at least 10X).

 https://gitlab.kitware.com/vtk/vtk/merge_requests/2419

 Hopefully it will test out OK and get merged into VTK and then merged
 into ParaView.

 On Mon, Dec 12, 2016 at 11:38 AM, Andy Smith 
 wrote:

> Ken,
>
> You are correct in your assumption about my dataset; it is multiblock
> unstructured data.  Adding a MergeBlocks filter to my slice does improve
> the 5.2.0 with OpenGL2 significantly.
> With that change the performance of 5.2.0 with OpenGL2 is on the same
> order of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0
> variants are still slower than 4.4.0.
>
> I realized after posting that my comparison is not exactly apples to
> apples - my build of 5.2.0 required updates to mesa (13.0.0) compared with
> the version I used for 4.4.0 (11.2.0).
> I will recompile 4.4.0 with the same mesa version I used with 5.2.0
> and report back.
>
> Thank you for your input.
>
> -Andy
>
> On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin 
> wrote:
>
>> I have not had time to look into this, but my quick guess is that you
>> have a multiblock dataset with a lot of blocks. Can you try merging the
>> blocks all together somehow into one block and see if the rendering speed
>> improves?
>>
>> On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber 
>> wrote:
>>
>>> Hi,
>>>
>>>
>>> i have a problem which is there in all RCs (1-4) of current
>>> ParaView: if i compile the RenderingBackend to "OpenGL2" and use 
>>> SurfaceLic
>>> the visualization time goes up from a few seconds (with "OpenGL") to 
>>> many
>>> minutes and the memory consumption from a few GByte to approx 100 GByte.
>>> Switching back to OpenGL any thing is fine. Without SurfaceLIC the 
>>> 

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Andy Smith]

Ken,

I've attached a small example file and script that demonstrates the problem.

The example dataset is a multiblock unstructured grid with two blocks.  The
script attempts to create two images of slices through the dataset.  At the
first slice position data from both blocks are in the slice.  At the second
slice position only data from one block is part of the slice.  This works
in serial and parallel (two processors) using pvbatch when not using
surface LIC and in serial when using surface LIC with the latest ParaView
from the git repo.  In parallel, the surface LIC hangs when creating the
second image.  This script works in parallel with surface LIC when using
version 4.4.0.

As an aside, I call the function to ensure that the surface LIC plugin is
loaded at the beginning of the script.  When I do this I get warnings about
"Replacing existing representation for key: Surface LIC".  I assume this is
because in my build I turned on the option to load the surface LIC plugin
by default.  Is there a way in Python to check if a plugin is loaded and
skip this step if that is the case?

Let me know if I can provide any additional information.

-Andy

On Tue, Feb 7, 2017 at 8:32 AM, Ken Martin  wrote:

> Thanks Andy, just to make sure I understand the issue. Is the case that
> one processor has no data, or is it that it has data, but it is outside the
> current view?
>
> Thanks
> Ken
>
> On Thu, Feb 2, 2017 at 5:14 PM, Andy Smith  wrote:
>
>> I've tested this with the latest ParaView and it does seem to run much
>> faster.
>> I am having a problem when running in parallel though.  If one processor
>> does not have anything to display, the rendering hangs.
>> Related to that, what routine actually calls
>> vtkCompositeLICHelper::RenderPiece?
>> The process that does not have anything to display never gets to that
>> function but the other processes do.  In that function there are MPI calls
>> where the rendering hangs.
>>
>> As a reply to Chuck:  I can use EGL in some instances but unfortunately
>> not all of our machine have GPUs so software rendering is still required.
>>
>> Thanks again for all the help.
>>
>> On Mon, Jan 23, 2017 at 12:45 PM, Ken Martin 
>> wrote:
>>
>>> There is a VTK topic now that should fix this regression. With this
>>> change OpenGL2 LIC seems to run much faster (at least 10X).
>>>
>>> https://gitlab.kitware.com/vtk/vtk/merge_requests/2419
>>>
>>> Hopefully it will test out OK and get merged into VTK and then merged
>>> into ParaView.
>>>
>>> On Mon, Dec 12, 2016 at 11:38 AM, Andy Smith 
>>> wrote:
>>>
 Ken,

 You are correct in your assumption about my dataset; it is multiblock
 unstructured data.  Adding a MergeBlocks filter to my slice does improve
 the 5.2.0 with OpenGL2 significantly.
 With that change the performance of 5.2.0 with OpenGL2 is on the same
 order of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0
 variants are still slower than 4.4.0.

 I realized after posting that my comparison is not exactly apples to
 apples - my build of 5.2.0 required updates to mesa (13.0.0) compared with
 the version I used for 4.4.0 (11.2.0).
 I will recompile 4.4.0 with the same mesa version I used with 5.2.0 and
 report back.

 Thank you for your input.

 -Andy

 On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin 
 wrote:

> I have not had time to look into this, but my quick guess is that you
> have a multiblock dataset with a lot of blocks. Can you try merging the
> blocks all together somehow into one block and see if the rendering speed
> improves?
>
> On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber 
> wrote:
>
>> Hi,
>>
>>
>> i have a problem which is there in all RCs (1-4) of current ParaView:
>> if i compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
>> visualization time goes up from a few seconds (with "OpenGL") to many
>> minutes and the memory consumption from a few GByte to approx 100 GByte.
>> Switching back to OpenGL any thing is fine. Without SurfaceLIC the 
>> OpenGL2
>> works although fine for me.
>>
>> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and
>> has (the same)  problems?
>>
>> o Kernel 4.7.6-1 / x86-64
>> o Quadro 4000
>> o NVidia driver v367.57
>> o PV 5.2.0 RC1 - RC4
>>
>> Best regards,
>>
>>Stefan
>>
>>
>>
>> ___
>> 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
>>
>> Search the list archives 

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Ken Martin]

Thanks Andy, just to make sure I understand the issue. Is the case that one
processor has no data, or is it that it has data, but it is outside the
current view?

Thanks
Ken

On Thu, Feb 2, 2017 at 5:14 PM, Andy Smith  wrote:

> I've tested this with the latest ParaView and it does seem to run much
> faster.
> I am having a problem when running in parallel though.  If one processor
> does not have anything to display, the rendering hangs.
> Related to that, what routine actually calls vtkCompositeLICHelper::
> RenderPiece?
> The process that does not have anything to display never gets to that
> function but the other processes do.  In that function there are MPI calls
> where the rendering hangs.
>
> As a reply to Chuck:  I can use EGL in some instances but unfortunately
> not all of our machine have GPUs so software rendering is still required.
>
> Thanks again for all the help.
>
> On Mon, Jan 23, 2017 at 12:45 PM, Ken Martin 
> wrote:
>
>> There is a VTK topic now that should fix this regression. With this
>> change OpenGL2 LIC seems to run much faster (at least 10X).
>>
>> https://gitlab.kitware.com/vtk/vtk/merge_requests/2419
>>
>> Hopefully it will test out OK and get merged into VTK and then merged
>> into ParaView.
>>
>> On Mon, Dec 12, 2016 at 11:38 AM, Andy Smith 
>> wrote:
>>
>>> Ken,
>>>
>>> You are correct in your assumption about my dataset; it is multiblock
>>> unstructured data.  Adding a MergeBlocks filter to my slice does improve
>>> the 5.2.0 with OpenGL2 significantly.
>>> With that change the performance of 5.2.0 with OpenGL2 is on the same
>>> order of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0
>>> variants are still slower than 4.4.0.
>>>
>>> I realized after posting that my comparison is not exactly apples to
>>> apples - my build of 5.2.0 required updates to mesa (13.0.0) compared with
>>> the version I used for 4.4.0 (11.2.0).
>>> I will recompile 4.4.0 with the same mesa version I used with 5.2.0 and
>>> report back.
>>>
>>> Thank you for your input.
>>>
>>> -Andy
>>>
>>> On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin 
>>> wrote:
>>>
 I have not had time to look into this, but my quick guess is that you
 have a multiblock dataset with a lot of blocks. Can you try merging the
 blocks all together somehow into one block and see if the rendering speed
 improves?

 On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber 
 wrote:

> Hi,
>
>
> i have a problem which is there in all RCs (1-4) of current ParaView:
> if i compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
> visualization time goes up from a few seconds (with "OpenGL") to many
> minutes and the memory consumption from a few GByte to approx 100 GByte.
> Switching back to OpenGL any thing is fine. Without SurfaceLIC the OpenGL2
> works although fine for me.
>
> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and
> has (the same)  problems?
>
> o Kernel 4.7.6-1 / x86-64
> o Quadro 4000
> o NVidia driver v367.57
> o PV 5.2.0 RC1 - RC4
>
> Best regards,
>
>Stefan
>
>
>
> ___
> 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
>
> Search the list archives at: http://markmail.org/search/?q=ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/paraview
>
>


 --
 Ken Martin PhD
 Chairman & CFO
 Kitware Inc.
 28 Corporate Drive
 Clifton Park NY 12065
 518 371 3971 <(518)%20371-3971>

 This communication, including all attachments, contains confidential
 and legally privileged information, and it is intended only for the use of
 the addressee.  Access to this email by anyone else is unauthorized. If you
 are not the intended recipient, any disclosure, copying, distribution or
 any action taken in reliance on it is prohibited and may be unlawful. If
 you received this communication in error please notify us immediately and
 destroy the original message.  Thank you.

 ___
 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

 Search the list archives at: http://markmail.org/search/?q=ParaView

 Follow this link to subscribe/unsubscribe:
 http://public.kitware.com/mailman/listinfo/paraview

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Andy Smith]

I've tested this with the latest ParaView and it does seem to run much
faster.
I am having a problem when running in parallel though.  If one processor
does not have anything to display, the rendering hangs.
Related to that, what routine actually calls
vtkCompositeLICHelper::RenderPiece?
The process that does not have anything to display never gets to that
function but the other processes do.  In that function there are MPI calls
where the rendering hangs.

As a reply to Chuck:  I can use EGL in some instances but unfortunately not
all of our machine have GPUs so software rendering is still required.

Thanks again for all the help.

On Mon, Jan 23, 2017 at 12:45 PM, Ken Martin  wrote:

> There is a VTK topic now that should fix this regression. With this change
> OpenGL2 LIC seems to run much faster (at least 10X).
>
> https://gitlab.kitware.com/vtk/vtk/merge_requests/2419
>
> Hopefully it will test out OK and get merged into VTK and then merged into
> ParaView.
>
> On Mon, Dec 12, 2016 at 11:38 AM, Andy Smith  wrote:
>
>> Ken,
>>
>> You are correct in your assumption about my dataset; it is multiblock
>> unstructured data.  Adding a MergeBlocks filter to my slice does improve
>> the 5.2.0 with OpenGL2 significantly.
>> With that change the performance of 5.2.0 with OpenGL2 is on the same
>> order of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0
>> variants are still slower than 4.4.0.
>>
>> I realized after posting that my comparison is not exactly apples to
>> apples - my build of 5.2.0 required updates to mesa (13.0.0) compared with
>> the version I used for 4.4.0 (11.2.0).
>> I will recompile 4.4.0 with the same mesa version I used with 5.2.0 and
>> report back.
>>
>> Thank you for your input.
>>
>> -Andy
>>
>> On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin 
>> wrote:
>>
>>> I have not had time to look into this, but my quick guess is that you
>>> have a multiblock dataset with a lot of blocks. Can you try merging the
>>> blocks all together somehow into one block and see if the rendering speed
>>> improves?
>>>
>>> On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber 
>>> wrote:
>>>
 Hi,


 i have a problem which is there in all RCs (1-4) of current ParaView:
 if i compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
 visualization time goes up from a few seconds (with "OpenGL") to many
 minutes and the memory consumption from a few GByte to approx 100 GByte.
 Switching back to OpenGL any thing is fine. Without SurfaceLIC the OpenGL2
 works although fine for me.

 Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and has
 (the same)  problems?

 o Kernel 4.7.6-1 / x86-64
 o Quadro 4000
 o NVidia driver v367.57
 o PV 5.2.0 RC1 - RC4

 Best regards,

Stefan



 ___
 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

 Search the list archives at: http://markmail.org/search/?q=ParaView

 Follow this link to subscribe/unsubscribe:
 http://public.kitware.com/mailman/listinfo/paraview


>>>
>>>
>>> --
>>> Ken Martin PhD
>>> Chairman & CFO
>>> Kitware Inc.
>>> 28 Corporate Drive
>>> Clifton Park NY 12065
>>> 518 371 3971 <(518)%20371-3971>
>>>
>>> This communication, including all attachments, contains confidential and
>>> legally privileged information, and it is intended only for the use of the
>>> addressee.  Access to this email by anyone else is unauthorized. If you are
>>> not the intended recipient, any disclosure, copying, distribution or any
>>> action taken in reliance on it is prohibited and may be unlawful. If you
>>> received this communication in error please notify us immediately and
>>> destroy the original message.  Thank you.
>>>
>>> ___
>>> 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
>>>
>>> Search the list archives at: http://markmail.org/search/?q=ParaView
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://public.kitware.com/mailman/listinfo/paraview
>>>
>>>
>>
>
>
> --
> Ken Martin PhD
> Chairman & CFO
> Kitware Inc.
> 28 Corporate Drive
> Clifton Park NY 12065
> 518 371 3971 <(518)%20371-3971>
>
> This communication, including all attachments, contains confidential and
> legally privileged information, and it is intended only for the use of the
> addressee.  Access to this email by anyone else is unauthorized. If you are
> not the intended 

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Chuck Atkins]

Hi Andy,

I realized after posting that my comparison is not exactly apples to apples
> - my build of 5.2.0 required updates to mesa (13.0.0) compared with the
> version I used for 4.4.0 (11.2.0).
> I will recompile 4.4.0 with the same mesa version I used with 5.2.0 and
> report back.
>
...

> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and has
>>> (the same)  problems?
>>> ...
>>> o NVidia driver v367.57
>>>
>>
With 5.2.0, you also have the option of building with EGL which will let
you directly render on your NVidia GPU w/o an x11 server, vs using OSMesa
for CPU based rendering.

- Chuck
___
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

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Andy Smith]

Ken,

You are correct in your assumption about my dataset; it is multiblock
unstructured data.  Adding a MergeBlocks filter to my slice does improve
the 5.2.0 with OpenGL2 significantly.
With that change the performance of 5.2.0 with OpenGL2 is on the same order
of magnitude as 5.2.0 with OpenGL and with 4.4.0, though the 5.2.0 variants
are still slower than 4.4.0.

I realized after posting that my comparison is not exactly apples to apples
- my build of 5.2.0 required updates to mesa (13.0.0) compared with the
version I used for 4.4.0 (11.2.0).
I will recompile 4.4.0 with the same mesa version I used with 5.2.0 and
report back.

Thank you for your input.

-Andy

On Fri, Dec 9, 2016 at 3:44 PM, Ken Martin  wrote:

> I have not had time to look into this, but my quick guess is that you have
> a multiblock dataset with a lot of blocks. Can you try merging the blocks
> all together somehow into one block and see if the rendering speed improves?
>
> On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber 
> wrote:
>
>> Hi,
>>
>>
>> i have a problem which is there in all RCs (1-4) of current ParaView: if
>> i compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
>> visualization time goes up from a few seconds (with "OpenGL") to many
>> minutes and the memory consumption from a few GByte to approx 100 GByte.
>> Switching back to OpenGL any thing is fine. Without SurfaceLIC the OpenGL2
>> works although fine for me.
>>
>> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and has
>> (the same)  problems?
>>
>> o Kernel 4.7.6-1 / x86-64
>> o Quadro 4000
>> o NVidia driver v367.57
>> o PV 5.2.0 RC1 - RC4
>>
>> Best regards,
>>
>>Stefan
>>
>>
>>
>> ___
>> 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
>>
>> Search the list archives at: http://markmail.org/search/?q=ParaView
>>
>> Follow this link to subscribe/unsubscribe:
>> http://public.kitware.com/mailman/listinfo/paraview
>>
>>
>
>
> --
> Ken Martin PhD
> Chairman & CFO
> Kitware Inc.
> 28 Corporate Drive
> Clifton Park NY 12065
> 518 371 3971 <(518)%20371-3971>
>
> This communication, including all attachments, contains confidential and
> legally privileged information, and it is intended only for the use of the
> addressee.  Access to this email by anyone else is unauthorized. If you are
> not the intended recipient, any disclosure, copying, distribution or any
> action taken in reliance on it is prohibited and may be unlawful. If you
> received this communication in error please notify us immediately and
> destroy the original message.  Thank you.
>
> ___
> 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
>
> Search the list archives at: http://markmail.org/search/?q=ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/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

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Ken Martin]

I have not had time to look into this, but my quick guess is that you have
a multiblock dataset with a lot of blocks. Can you try merging the blocks
all together somehow into one block and see if the rendering speed improves?

On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber  wrote:

> Hi,
>
>
> i have a problem which is there in all RCs (1-4) of current ParaView: if i
> compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
> visualization time goes up from a few seconds (with "OpenGL") to many
> minutes and the memory consumption from a few GByte to approx 100 GByte.
> Switching back to OpenGL any thing is fine. Without SurfaceLIC the OpenGL2
> works although fine for me.
>
> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and has
> (the same)  problems?
>
> o Kernel 4.7.6-1 / x86-64
> o Quadro 4000
> o NVidia driver v367.57
> o PV 5.2.0 RC1 - RC4
>
> Best regards,
>
>Stefan
>
>
>
> ___
> 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
>
> Search the list archives at: http://markmail.org/search/?q=ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/paraview
>
>


-- 
Ken Martin PhD
Chairman & CFO
Kitware Inc.
28 Corporate Drive
Clifton Park NY 12065
518 371 3971

This communication, including all attachments, contains confidential and
legally privileged information, and it is intended only for the use of the
addressee.  Access to this email by anyone else is unauthorized. If you are
not the intended recipient, any disclosure, copying, distribution or any
action taken in reliance on it is prohibited and may be unlawful. If you
received this communication in error please notify us immediately and
destroy the original message.  Thank you.
___
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

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview

Re: [Paraview] Rendering Backend OpenGL2 & PV 5.2.0-RC4 & SurfaceLic --> Problem...

[Andy Smith]

Stefan,

I am testing out 5.2.0 with offscreen rendering and have observed similar
results.  My version built with OpenGL2 is slightly faster than the OpenGL
version for non surface LIC operations, but massively slower when using
Surface LIC.

Comparing back to version 4.4.0, version 5.2.0 with OpenGL2 is faster for
non-Surface LIC operations but much slower for surface LIC.  In fact when
using Surface LIC, version 4.4.0 is faster then 5.2.0 with OpenGL.

My test case involves creating a slice in an unstructured dataset and
sweeping this slice in one direction taking multiple snapshots.

Hopefully someone more knowledgeable on this subject than me has some ideas
on the problem and how to address it.

-Andy

On Tue, Nov 15, 2016 at 5:35 AM, Stefan Melber  wrote:

> Hi,
>
>
> i have a problem which is there in all RCs (1-4) of current ParaView: if i
> compile the RenderingBackend to "OpenGL2" and use SurfaceLic the
> visualization time goes up from a few seconds (with "OpenGL") to many
> minutes and the memory consumption from a few GByte to approx 100 GByte.
> Switching back to OpenGL any thing is fine. Without SurfaceLIC the OpenGL2
> works although fine for me.
>
> Has any one tested OpenGL2 with SurfaceLIC-Plugin the last time and has
> (the same)  problems?
>
> o Kernel 4.7.6-1 / x86-64
> o Quadro 4000
> o NVidia driver v367.57
> o PV 5.2.0 RC1 - RC4
>
> Best regards,
>
>Stefan
>
>
>
> ___
> 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
>
> Search the list archives at: http://markmail.org/search/?q=ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/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

Search the list archives at: http://markmail.org/search/?q=ParaView

Follow this link to subscribe/unsubscribe:
http://public.kitware.com/mailman/listinfo/paraview