Today, many programs are memory-bound rather than compute-bound. The reason for this is that processor technology has advanced faster than memory technology.
Given that the Z-Buffer algorithm operates on a large amount of data (in particular, all samples for all triangles), how have engineers minimized the amount of slow data movement between the processor and the GPU?
Today, many programs are memory-bound rather than compute-bound. The reason for this is that processor technology has advanced faster than memory technology.
Given that the Z-Buffer algorithm operates on a large amount of data (in particular, all samples for all triangles), how have engineers minimized the amount of slow data movement between the processor and the GPU?