Hi Hui, I was referring to the possibility of avoiding nvcc altogether rather than setting the --ccbin (essentially eschewing nvcc to do this instead : ` https://llvm.org/docs/CompileCudaWithLLVM.html`). As Ken noted previously, it looks like building yaksa (with CUDA support) requires the use of nvcc. Thank You, Sajid Ali (he/him) | Research Associate Scientific Computing Division Fermi National Accelerator Laboratory s-sajid-ali.github.io