Hey I've been try to build a compiler that ingests glsl and dumps out nvptx code. I am trying to do so using LunarG's LunarGLASS compiler stack. I am getting an nvptx backend ready using the dummy and glsl backends provided in the github as reference, or would it be better to try to use the top and or bottom irs described in the stack as an input to an LLVM->ptx conversion pass.
Hoping you guys can help.