Implementing a CUDA based renderer

[bilgili]

• Implement a renderer which simply uploads all data to memory and renders
• Compute rendering only in the 2D region of frame buffer where volume is projected.
• Try different memory models ( Pinned, Managed memory ). ( Add/Use -thrust- an allocator class for allocating, cache objects for cuda )
• Try different data orderings ( Morton, Hilbert, etc ) for spatial locality
• Try different partitioning of cuda kernels
• Implement GI