I'm currently using an image in order to take advantage of the GPU caching system.
My algorithm iterates the image from top to bottom for each thread ( but skipping some random texels differently for each thread ).
I was wondering if I could improve the caching efficiency storing the image using Morton codes/Z-order but... how is the texture swizzled/tiled internally in VRAM?
There is no general answer because texture layout is hardware dependent.
Here you could find texture format used by old AMD/ATI chips: