DPC++ LLVM NVIDIA compiler, unable to run programs

[tamara-laranga]

Describe the bug

I try to execute SYCL code on my NVIDIA GPU and for that I followed all the instructions on https://www.intel.com/content/www/us/en/developer/articles/technical/compiling-sycl-with-different-gpus.html, with the sample code vector_add.cpp it seems to compile the program using the -fsycl and -fsycl-targets=nvptx64-nvidia-cuda flags. The problem comes when trying to execute the program, because it generates this error:

$ ./vector_add-cuda
terminate called after throwing an instance of 'sycl::_V1::compile_program_error'
  what():  The program was built for 1 devices
Build program log for 'NVIDIA GeForce GTX 1650 Ti':
 -999 (Unknown PI error)
Abortado (`core' generado)

To reproduce

1. Install the DPC++ CUDA NVIDIA compiler following the instructions. https://www.intel.com/content/www/us/en/developer/articles/technical/compiling-sycl-with-different-gpus.html
2. Compile the vector_add.cpp file.

Environment

• OS: Ubuntu 24.04 LTS
• clang++ version: 19.0.0
• Target: NVIDIA GeFOrce GTX 1650 Ti

Additional context

No response

[JackAKirk]

Which version of the cuda toolkit are you using and does it support 24.04? Only very recent versions of the cuda toolkit support it.

[npmiller]

This is working fine with the latest compiler so closing this. Feel free to re-open if the issue persist in your environment.