CUDA clang running on GPU

After gaining a lot of insights from a CppCon 2016 talk Bringing Clang and C++ to GPUs I decided to try clang code to run on GPU using the CUDA Clang docs

  1. Use clang 17 (or 16 should work as well!)
  2. Installation of CUDA toolkit is assumed. Have drivers and a GPU
  3. file .cu -> .ptx -> .out
  4. clang -> ptxas -> sass -> exe

clang 14 ## did not work with Cuda 12.2 nor 11.8

clang-15 ## NOT WORKING // sudo apt install clang-15 

clang-17 ## OK GOOD. https://ubuntuhandbook.org/index.php/2023/09/how-to-install-clang-17-or-16-in-ubuntu-22-04-20-04/

## STIL did not work? Nope! It worked
## native
clang++-17 axpy.cu -o axpy.out --cuda-gpu-arch=sm_86 -L/usr/local/cuda-12.2/lib64 -lcudart_static -ldl -lrt -pthread 
## and from HPC tool kit also. Ignore the warining.
clang++-17 axpy.cu -o axpy.out --cuda-gpu-arch=sm_86 -L/opt/nvidia/hpc_sdk/Linux_x86_64/24.3/cuda/11.8/lib64 -lcudart_static -ldl -lrt -pthread

./axpy.out
y[0] = 2
y[1] = 4
y[2] = 6
y[3] = 8

Errors before fixing it!

/usr/lib/llvm-15/lib/clang/15.0.7/include/__clang_cuda_texture_intrinsics.h:696:13: error: no template named 'texture'
            texture<__DataT, __TexT, cudaReadModeNormalizedFloat> __handle,
            ^
/usr/lib/llvm-15/lib/clang/15.0.7/include/__clang_cuda_texture_intrinsics.h:709:13: error: no template named 'texture'
            texture<__DataT, __TexT, cudaReadModeElementType> __handle,
            ^
2 errors generated when compiling for sm_86.
★ 3 min read · Rajesh Pandian M · ★ cuda , clang , gpu