What's the difference between global memory and texture in CUDA? To speed up memory copying from host to device, which one is better? I am going to use them for Image Processing. I've seen the sample for bilateral filtering. It used texture instead of global memory.
I'd like some to explain about it. Thanks.