[Paraview] [vtk-developers] GPU volume rendering failure on Intel HD XXXX/Iris, Windows and recent drivers

Simon ESNEAULT simon.esneault at gmail.com
Mon Jul 20 10:11:57 EDT 2015


Hi Aashish,

Thanks you very much for looking into this issue. Our customers (and I
believe the community since the issue is visible in paraview) would really
benefit from such a fix. Quite a lot of new computers just have an intel
chipset nowadays.
This of course until the new OpenGL2 backend is ready !

Please let me know if there is anything I can do to help.

Thanks a lot
Simon

2015-07-20 15:21 GMT+02:00 Aashish Chaudhary <aashish.chaudhary at kitware.com>
:

> Hi Simon,
>
> On Mon, Jul 20, 2015 at 5:02 AM, Simon ESNEAULT <simon.esneault at gmail.com>
> wrote:
>
>> Hi Aashish,
>>
>> Thanks for the patchs.
>> However, I've tried them but without success, the GPU rendering won't
>> show up on an Intel HD chipset using the GPURenderDemo (the same exe/dll
>> works fine on a machine with an NVidia/Ati card). The warning disappeared
>> though ...
>>
>
> The old mapper had issues on some of the ATI cards (specifically on Mac)
> and Intel chipset as well. So I am not surprised that it didn't work.
>
>
>> The proposed patchs seems to affect only the TextureMapper 3D, shouldn't
>> there be modification to
>> the vtkOpenGLGPUVolumeRayCastMapper::LoadExtensions method as well?
>>
>
> That method calls the vtkGL method I believe which I took care of it. I
> will have a second look at it today.
>
>>
>> I've tried to uncomment all added lines in the GL_VERSION_1_2_DEPRECATED
>> part of vtkgl.cxx, build correctly everything, and on the problematic
>> machine the checkIfNull() fail for all the lines, the program prints out :
>> Function pointer vtkgl::ColorTable is invalid
>> Function pointer vtkgl::.. is invalid
>> For every one of them.
>>
>> Also the vtkgl.cxx file is generated by the vtkParseOGLExt exe, shouldn't
>> there be modification in the headers glext.h/glxext.h/wglext.h instead ?
>> Maybe this is causing some troubles elsewhere...
>>
>
> Yes and No. Basically what happened was  that newer drivers stopped
> reporting *1_2* variables and that's why you are running into issues. Since
> the newer backend uses GLEW, all of this mostly works correctly.
>
>>
>> If you have some other magical trick to solve this, I would be glad to
>> hear about it! Also I can provide more informations if you want
>>
>
> I will have a look at it but even if we can fix all the extension issue,
> things may not work on Intel.
>
> - Aashish
>
>>
>> Thank You
>> Simon
>>
>>
>> 2015-07-17 17:48 GMT+02:00 Aashish Chaudhary <
>> aashish.chaudhary at kitware.com>:
>>
>>> Hi Andras,
>>>
>>> Please find the patchs attached. If you run into issues let me know. I
>>> should mention that apply these patches at your discretion as I have not
>>> done enough testing to ensure the quality of the patches but the test
>>> passed last time I applied them.
>>>
>>>
>>>
>>> On Fri, Jul 17, 2015 at 11:40 AM, Andras Lasso <lasso at queensu.ca> wrote:
>>>
>>>>  Please share the patch publicly, we would need it for some of our
>>>> projects, too. Thank you!
>>>>
>>>>
>>>>
>>>> Andras
>>>>
>>>>
>>>>
>>>> *From:* vtk-developers [mailto:vtk-developers-bounces at vtk.org] *On
>>>> Behalf Of *Simon ESNEAULT
>>>> *Sent:* Friday, July 17, 2015 11:35 AM
>>>> *To:* Aashish Chaudhary
>>>> *Cc:* VTK Developers; vtkusers at vtk.org; ParaView list
>>>> *Subject:* Re: [vtk-developers] [Paraview] GPU volume rendering
>>>> failure on Intel HD XXXX/Iris, Windows and recent drivers
>>>>
>>>>
>>>>
>>>> Dear Aashish,
>>>>
>>>>
>>>>
>>>> Thank you very much for you answer.
>>>>
>>>> Yes you are right, we are waiting for a new VTK release that defaults
>>>> to the new OpenGL2 backend, for safety and because a change of VTK version
>>>> requires a lot of extra work with our current framework... But the green
>>>> dashboard indeed looks really promising, and we will probably make the move
>>>> as soon as it's ready !
>>>>
>>>> Meanwhile, I would be really glad if you could send me the patch that
>>>> solves this issue with the old backend, or guide me through the needed
>>>> modification to make it work.
>>>>
>>>> Thanks again
>>>>
>>>> Simon
>>>>
>>>>
>>>>
>>>> 2015-07-17 16:22 GMT+02:00 Aashish Chaudhary <
>>>> aashish.chaudhary at kitware.com>:
>>>>
>>>>  Dear Simon,
>>>>
>>>>
>>>>
>>>> On Fri, Jul 17, 2015 at 4:43 AM, Simon ESNEAULT <
>>>> simon.esneault at gmail.com> wrote:
>>>>
>>>>  Hi All,
>>>>
>>>> We're experiencing GPU volume rendering failure on a lot (maybe all ?)
>>>> Intel HD windows system with up to date drivers.
>>>>
>>>>
>>>> This is reproducible in Paraview or with the GPURenderDemo example from
>>>> VTK, just load a volume and try to render the volume... Here is a result
>>>> image from paraview on a Dell Latitude E7450 (Intel HD 5500 inside), with
>>>> GPU rendering [1] and the same volume with CPU VR rendering [2].
>>>> Interestingly, the volume rendering used to work with drivers older or
>>>> equal to the version 10.18.10.3496 [3]. As a work around, we ask our
>>>> customers to downgrade their driver version to this one, and they're
>>>> enjoying smooth GPU rendering. But this driver is quite old and does not
>>>> support recent chipset...
>>>>
>>>> I've built and tested the GPURenderDemo with the brand new OpenGL2
>>>> backend on an affected machine and of course it does work out of the box
>>>> (congratulations for this amazing work !).
>>>>
>>>>  Great!
>>>>
>>>>
>>>>
>>>>  But our application is FDA/CE marked and we need to wait that this
>>>> new backend is completely ready for prime time. So for now, we have to
>>>> stick with VTK 6.1 ...
>>>>
>>>>
>>>>
>>>> Most of the tests are passing for OpenGL2 (
>>>> https://open.cdash.org/index.php?project=VTK) on most of the
>>>> dashboards (various configurations etc) but it is a moving target just like
>>>> VTK itself is a moving target. I guess you are waiting for a VTK release
>>>> that defaults to OpenGL2 as oppose to OpenGL?
>>>>
>>>>
>>>> Anyone has any clue ? They're some warning that pop up from paraview
>>>> related to the extension "gl_version_1_2" that could not be loaded, not
>>>> sure if that could be related ? Following this thread [4] and the fix
>>>> proposed, I've tried to replace
>>>>
>>>> /*****************/
>>>>
>>>> extensions->LoadExtension("GL_VERSION_1_2")
>>>>
>>>> /*****************/
>>>>
>>>> by
>>>>
>>>> /*****************/
>>>>
>>>> extensions->LoadExtension("GL_EXT_texture3D");
>>>> extensions->LoadExtension( "GL_ARB_texture_non_power_of_two" );
>>>>
>>>> /*****************/
>>>>
>>>> in the vtkOpenGLGPUVolumeRayCastMapper.cxx, The warning disappears but
>>>> there is still no volume rendering (tested with GPURenderDemo). Maybe it
>>>> needs some others extensions ?
>>>>
>>>>
>>>>
>>>> Yes, there is a patch and I can send it to you. I have to talk to folks
>>>> here if we can fix the old mapper with those changes but as of now it is
>>>> not in the plan.
>>>>
>>>>
>>>>
>>>>
>>>> Should I fill a bug in paraview, vtk ? Are there any plan to maintain
>>>> and maybe fix this for the old backend ?
>>>>
>>>>
>>>>
>>>> - Aashish
>>>>
>>>>
>>>>
>>>>
>>>> Thanks,
>>>> --
>>>>
>>>> ------------------------------------------------------------------
>>>> Simon Esneault
>>>>
>>>> Rennes, France
>>>> ------------------------------------------------------------------
>>>>
>>>>
>>>> [1] http://picpaste.com/pics/Paraview-4-3-1-GPU-VR_fail.1437122217.PNG
>>>>
>>>> [2] http://picpaste.com/pics/Paraview-4-3-1-CPU-VR_win.1437122267.PNG
>>>>
>>>> [3] https://downloadcenter.intel.com/Detail_Desc.aspx?DwnldID=23339
>>>> [4] http://public.kitware.com/pipermail/vtkusers/2014-April/083584.html
>>>>
>>>>
>>>> _______________________________________________
>>>> 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
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>>
>>>>
>>>>
>>>> *| Aashish Chaudhary | Technical Leader          | Kitware Inc.
>>>>     *
>>>>
>>>> *| http://www.kitware.com/company/team/chaudhary.html
>>>> <http://www.kitware.com/company/team/chaudhary.html>*
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>>
>>>> ------------------------------------------------------------------
>>>> Simon Esneault
>>>>
>>>> Rennes, France
>>>> ------------------------------------------------------------------
>>>>
>>>
>>>
>>>
>>> --
>>>
>>>
>>>
>>> *| Aashish Chaudhary | Technical Leader         | Kitware Inc.
>>>   *
>>> *| http://www.kitware.com/company/team/chaudhary.html
>>> <http://www.kitware.com/company/team/chaudhary.html>*
>>>
>>
>>
>>
>> --
>> ------------------------------------------------------------------
>> Simon Esneault
>> Rennes, France
>> ------------------------------------------------------------------
>>
>
>
>
> --
>
>
>
> *| Aashish Chaudhary | Technical Leader         | Kitware Inc.            *
> *| http://www.kitware.com/company/team/chaudhary.html
> <http://www.kitware.com/company/team/chaudhary.html>*
>



-- 
------------------------------------------------------------------
Simon Esneault
Rennes, France
------------------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20150720/fd3c41e7/attachment.html>


More information about the ParaView mailing list