Dear Lisa, Francois; I confirm that GTX 285 went fine on Linux. So I suppose we really have a bug/issue in the fragment program on Mac.
Best Regards Fauze On Sun, Feb 21, 2010 at 4:34 PM, Fauze Polpeta <[email protected]> wrote: > Hi Lisa, > > Thank you so much for contacting me. > > What I will do next is pickup the GTX 285 from the Mac Pro and try it > on a Linux station. But I confirm that I have faced the same > problem(s) in a MacBook Pro with a 9400M. Indeed, I have other > findings in respect to the deployment of GPU-based Ray Cast in Mac OS > X: > > > a) GPU-based Ray Cast is not working for DICOM CT datasets that have > more than 170 images (aprox.) and are 512x512 in size. Resultant > volume is very "scratched" - see atached PNG. > > b) As I mentioned, shading doesn't work as well. Even if setting > AutoAdjustSampleDistances on. Interacting with the volume is very > compromised, I can say impossible. > > c) VolView 3.2, which I supposed relies on the VTKEdge approach for > implementing Ray Cast, crashes when GPU support is enabled in Mac OS > X. > > > If you from Kitware concluded there is a bug in the fragment program, > how can I try/help to solve it? Or this is something that may need > some work in the NVidia drivers for Mac OS X? > > Thanks in advance for any help. I will let you know about the GTX 285 on > Linux. > > Best Regards > > Fauze > > Ps.: Always using Leopard since Snow Leopard has a NVidia driver bug. > Also, as suggested by Francois, I have opt by the vtkKWEVolumeMapper, > which, in turn, has deployed the GPU-based Ray Cast mapper. > > On Sat, Feb 20, 2010 at 4:27 AM, Lisa Avila <[email protected]> wrote: >> Hello Fauze, >> >> Do you have AutoAdjustSampleDistances on? It should drop screen space >> resolution of the computed image to get the desired update rate. So you'll >> get about the same rendering rate but quality will be lower when shading is >> on. If you have this on and it is still far below your desired update rate >> then there must be some bug / issue in the fragment program on that >> platform. Do you happen to have a system with that (or a similar card) on >> something other than a Mac? >> >> Lisa >> >> >> On Fri, Feb 12, 2010 at 4:03 PM, Fauze Polpeta <[email protected]> >> wrote: >>> >>> Hi Francois; >>> The VRAM detection is working fine. >>> But interact with the volume when using shaders are still very >>> compromised. I've got the first rendering result but if I try any kind o >>> interaction render window becomes blank. This also happens when executing >>> the VTKEdgeGPURenderDemo example. >>> I diagnosed that the vtkKWEGPUVolumeRayCastMapper has been deployed by >>> the vtkKWEVolumeMapper proxy, so I'm not sure if this should be classified >>> as a performance issue (shading is heavy) or a bad function of GPU-based Ray >>> Casting on Leopard. >>> Please, how could I help you on helping me? :) >>> Best Regards >>> Fauze >>> On Fri, Feb 12, 2010 at 2:37 PM, Francois Bertel >>> <[email protected]> wrote: >>>> >>>> Hello, >>>> >>>> I just fixed VTKEdge/Common/vtkKWECoreGraphicsGPUInfoList.mm >>>> (svn revision 1481). >>>> >>>> I tried it on Tiger with an ATI 128MB and on Snow Leopard on a MacBook >>>> Pro in both energy modes (9400M 256MB+9600M GT 512MB) >>>> >>>> Please give it a try without forcing SetMaxMemoryInBytes anymore and >>>> let more if it works for you. >>>> >>>> For the second question, try to not use vtkKWEGPUVolumeRayCastMapper >>>> directly but instead use vtkKWEVolumeMapper which is a volume mapper >>>> chooser. Notably, it can switch automatically between a low res mapper >>>> and high res GPUVolumeRayCastMapper to allow good interaction >>>> experience. >>>> >>>> Regards. >>>> >>>> >>>> On Tue, Feb 9, 2010 at 7:52 AM, Fauze Polpeta <[email protected]> >>>> wrote: >>>> > Hi Francois; >>>> > On Leopard everything went fine, except two things: >>>> > 1) I had to SetMaxMemoryInBytes since the amount of memory of my NVidia >>>> > was >>>> > not identified correctly. It's a minor issue I think if I consider that >>>> > the >>>> > video board will always be the same. >>>> > 2) When using Shading it becomes almost impossible to interact with the >>>> > volume. I know shading requires more memory and processing power but >>>> > I'm on >>>> > a GTX 285 with 1GB. Am I missing something or doing something wrong?? >>>> > Best Regards >>>> > Fauze >>>> > On Mon, Feb 8, 2010 at 1:56 PM, Francois Bertel >>>> > <[email protected]> wrote: >>>> >> >>>> >> Hello, >>>> >> >>>> >> Yes, the comment is valid: the nVidia driver has a bug on Snow Leopard >>>> >> that prevents to run the GPU volume mapper. >>>> >> >>>> >> I don't have Leopard. You can try to remove the lines you are >>>> >> pointing at on your leopard installation and see what kind of result >>>> >> you have. If it works then I will put a statement specific to just >>>> >> snow leopard and not leopard. >>>> >> >>>> >> Regards. >>>> >> >>>> >> On Mon, Feb 8, 2010 at 8:41 AM, Fauze Polpeta >>>> >> <[email protected]> >>>> >> wrote: >>>> >> > Dear Team; >>>> >> > Via debug I found the following strange piece of code in >>>> >> > vtkKWEOpenGLGPUVolumeRayCastMapper.cxx: >>>> >> > // It does not work on Apple OS X Snow Leopard with nVidia. >>>> >> > // There is a bug in the OpenGL driver with an error in the >>>> >> > // Cg compiler about an infinite loop. >>>> >> > #ifdef __APPLE__ >>>> >> > this->LoadExtensionsSucceeded=0; >>>> >> > return; >>>> >> > #endif >>>> >> > This belongs to the function void >>>> >> > >>>> >> > """vtkKWEOpenGLGPUVolumeRayCastMapper::LoadExtensions(vtkRenderWindow >>>> >> > *window)""", which, in turn is called by """void >>>> >> > vtkKWEOpenGLGPUVolumeRayCastMapper::PreRender(...). >>>> >> > What I conclude is: by setting """this->LoadExtensionsSucceeded=0""" >>>> >> > will >>>> >> > always lead """LoadExtensions""" to fail on Mac OS X apps that rely >>>> >> > on >>>> >> > vtkKWEOpenGLGPUVolumeRayCastMapper. Such fail will not only occurs >>>> >> > on >>>> >> > Snow >>>> >> > Leopard but also in Leopard. >>>> >> > Please, can someone point me out if my findings makes sense? >>>> >> > Best Regards >>>> >> > Fauze >>>> >> > >>>> >> > On Sat, Feb 6, 2010 at 2:03 AM, Fauze Polpeta >>>> >> > <[email protected]> >>>> >> > wrote: >>>> >> >> >>>> >> >> Dear Edge Team; >>>> >> >> >>>> >> >> I'm trying to build a simple volume rendering application on top of >>>> >> >> Cocoa >>>> >> >> in order to evaluate the GPU Ray Casting support, but when I run >>>> >> >> the >>>> >> >> application I've got the following error: >>>> >> >> >>>> >> >> --- >>>> >> >> >>>> >> >> ERROR: In >>>> >> >> >>>> >> >> >>>> >> >> /Users/fauze/Development/toolkits/vtkedge/source/VolumeRendering/vtkKWEOpenGLGPUVolumeRayCastMapper.cxx, >>>> >> >> line 4139 >>>> >> >> vtkKWEOpenGLGPUVolumeRayCastMapper (0x768d600): Rendering failed >>>> >> >> because >>>> >> >> the following OpenGL extensions are required but not supported: >>>> >> >> >>>> >> >> Program received signal: “EXC_BAD_ACCESS”. >>>> >> >> >>>> >> >> --- >>>> >> >> >>>> >> >> I'm sure the program is ok since it works with the ordinary mappers >>>> >> >> of >>>> >> >> VTK >>>> >> >> and also if I let VTKEdge to load itself the mapper >>>> >> >> (vtkKWEVolumeMapper). >>>> >> >> The error occurs when I opt by vtkKWEGPUVolumeRayCastMapper which >>>> >> >> leads >>>> >> >> to >>>> >> >> the deployment of the vtkKWEOpenGLGPUVolumeRayCastMapper. The point >>>> >> >> is >>>> >> >> that >>>> >> >> the "following OpenGL extensions" are not listed and thus I can't >>>> >> >> figure out >>>> >> >> what is happening. >>>> >> >> >>>> >> >> Please, can someone give me so help? Some framework I'm missing? >>>> >> >> I'm >>>> >> >> running on top of Mac Pro Core i7 - NVidia GTX 285 - Snow Leopard / >>>> >> >> Leopard. >>>> >> >> >>>> >> >> Thanks in advance >>>> >> >> >>>> >> >> Fauze >>>> >> >> >>>> >> >> >>>> >> >> Ps.: The VTKEdgeVolumeRenderingCxxTests have printed out: Required >>>> >> >> extensions not supported. >>>> >> >> >>>> >> >> >>>> >> >> >>>> >> >> >>>> >> >> >>>> >> > >>>> >> > >>>> >> > _______________________________________________ >>>> >> > VtkEdge mailing list >>>> >> > [email protected] >>>> >> > http://public.kitware.com/cgi-bin/mailman/listinfo/vtkedge >>>> >> > >>>> >> > >>>> >> >>>> >> >>>> >> >>>> >> -- >>>> >> François Bertel, PhD | Kitware Inc. Suite 204 >>>> >> 1 (518) 371 3971 x113 | 28 Corporate Drive >>>> >> | Clifton Park NY 12065, USA >>>> >> _______________________________________________ >>>> >> VtkEdge mailing list >>>> >> [email protected] >>>> >> http://public.kitware.com/cgi-bin/mailman/listinfo/vtkedge >>>> > >>>> > >>>> >>>> >>>> >>>> -- >>>> François Bertel, PhD | Kitware Inc. Suite 204 >>>> 1 (518) 371 3971 x113 | 28 Corporate Drive >>>> | Clifton Park NY 12065, USA >>>> _______________________________________________ >>>> VtkEdge mailing list >>>> [email protected] >>>> http://public.kitware.com/cgi-bin/mailman/listinfo/vtkedge >>> >>> >>> _______________________________________________ >>> VtkEdge mailing list >>> [email protected] >>> http://public.kitware.com/cgi-bin/mailman/listinfo/vtkedge >>> >> >> > _______________________________________________ VtkEdge mailing list [email protected] http://public.kitware.com/cgi-bin/mailman/listinfo/vtkedge
