Written by JimStar:t_3 wrote:Zay wrote:to be precise, a limitation of CUDA, not Octane
AFAIK there are a max. number of 64 image and 32 float textures, as well as 4/4 HDRI rgba and greyscale textures possible.
128, "Maximum number of textures that can be bound to a kernel": http://en.wikipedia.org/wiki/CUDA
But why use CUDA textures in Octane kernels if CUDA has such a limitation? Why not regular array and just not get data from it in kernels something like "float c = texture_array[y*width+x]" instead of convenient but limited "float c = tex2D(tex, x, y)"? In this case this CUDA limitation does not matter (as long as there is enough GPU memory), only plain data will be loaded to GPU and only kernel will know what type of data it is... You can't limit "plain data" with number of textures in it if you don't know what that data contains and just get it loaded to GPU memory as bite sequence...
This is only my assumptions, I'm not CUDA programmer, so sorry if I wrote something wrong...

One more sorry if I'm wrong...

Anyway Octane is great, I respect this excellent work of RS team.