It's interesting to see here in code how CUDA can be thought similarly to ISPC in n-dimensions. We also have another layer of complexity now with warps as a way of dividing up a large number of threads into "32-wide" execution contexts.
Please log in to leave a comment.
It's interesting to see here in code how CUDA can be thought similarly to ISPC in n-dimensions. We also have another layer of complexity now with warps as a way of dividing up a large number of threads into "32-wide" execution contexts.