Hi all,
Does anyone know how to emit LLVM IR for the kernel of a CUDA C++ code? clang++ -emit-llvm --cuda-gpu-arch=sm_75 [x.cu](http://x.cu/)
somehow only emits code for the host side code.
Thanks,
Navdeep
Hi all,
Does anyone know how to emit LLVM IR for the kernel of a CUDA C++ code? clang++ -emit-llvm --cuda-gpu-arch=sm_75 [x.cu](http://x.cu/)
somehow only emits code for the host side code.
Thanks,
Navdeep
Hi Navdeep,
you can emit the llvm IR code of the device side with clang -S -emit-llvm --cuda-gpu-arch=sm_75 --cuda-device-only code.cu -o code.ll
and with clang -S --cuda-gpu-arch=sm_75 --cuda-device-only code.cu -o code.ptx
you can emit the ptx code.
Cheers,
Simeon
Hi Simeon,
Thanks! that worked.
Cheers,
Navdeep