I’d like to create a R32G32 image and write to it in a compute pass first, accessing it through 64bit atomic operations.
In a second pass, which is a render pass, I’d like to use the same image for both, a color attachment and a depth attachment—e.g., the R32 part for depth and the G32 part for color output.
Is that even posible?
In terms of use case, think of Nanite: First software rasterizing via compute using 64bit atomics, then hardware rasterizing into the same depth/color attachments.