Texture cache efficiency using cudaBindTexture2D

When cudaArray is bound to a texture reference, I understand the access is cached based on 2D spacial proximity (3D for 3D textures.)

How does the texture cache work when raw memory is bound to a texture reference? Does it optimize based on linear address proximity?


Yes, as far as I can tell from browsing around these forums, 1D textures bound to linear memory and 2D textures bound using cudaBindTexture2D will be cached based on linear proximity.