I am learning CUDA and building a graphics engine.
When open GL rasterizes on an 8800+ are the computations occuring on the same hardware threads that I am using for my CUDA kernals or is there specialized hardware for doing rasterization? Specifically (starting with) rendering a big batch of lines I am considering doing my own rasterization but I don’t want to waste GPU cycles if the calculations are occuring on its own hardware.
thanks,
Paul
Rasterization is performed by dedicated hardware, which is not currently accessible from CUDA.
Vertex and pixel shaders are executed using the same processors as CUDA programs.
thanks
I take it this holds true for other “native” graphics card taskes such
as (potentially) texture lookups and anti-aliasing.
Are there any specs on these additional processing units in the 8800s?
Also, what are NVIDIA’s plans for 4 years from now? Should I start looking at writing Brzenhams in CUDA as the specialized hardware fades into where the specialized hardware has been fading to or am I wasting CUDA cycles no matter how nice my line drawing code is?
texture lookups are supported in CUDA. Did you read the programming guide?
I read through it.
It’s not whether or not CUDA supports the functionality.
It is what type of specialized hardware pipelines can we expect NVIDIA to maintain on future cards.
It is a waste to leave powerful rasterizers idle while I do my own rasterization.
If you can access this paper: [url=“http://forums.nvidia.com/index.php?showtopic=68096”]http://forums.nvidia.com/index.php?showtopic=68096[/url]
you will have a lot more information.