I imagine we'll also need some changes to the d3d-asm output in order to distinguish between e.g. ADD on VKD3D_DATA_FLOAT sources and ADD on VKD3D_DATA_DOUBLE sources.
I was thinking of doing this where double is used with non-double instructions. If we don't normalise TPF it would only change the disassembly of DXIL shaders.
Right. I guess the question is what that would look like. One approach would be to output ADD with VKD3D_DATA_DOUBLE as DADD. At that point we could arguably just as well just emit a DADD in the first place from the DXIL parser though. Another approach would be to just include type information in the output for shader model 6+, much like DXIL assembly does. There would be some choices to make there as well; e.g., we could always switch this based on the shader model, but we could also make this a formatting option, which would allow displaying these for e.g. TPF shaders as well.