[vtkusers] Passing output of a render to CUDA without going through the Host

[Paris Flood]

After calling a Render() on a vtkRenderWindow, I would like to pass the rendered image to a CUDA function on the device (GPU) without going through the host (CPU). Currently I have been using a vtkWindowToImageFilter to get a pointer to the image data ( windowFilter->GetOutput()->GetScalarPointer() ), however it appears this pointer is to memory on the CPU and I therefore have to upload back to the GPU to begin my CUDA computations. Of course, this is a massive bottleneck and it would greatly improve the speed of my computation to bypass this whole download/upload process.
Any advice is greatly appreciated, thank you!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20160622/21ba1961/attachment.html>