Did we not need -DBUILD_SYCL_WITH_CUDA? Additionally, I'm running into issues with compute-benchmarks trying to compile l0: what commands are you using to run main.py to circumvent this (although, there is perhaps a cmake configure setting we're missing here)?

Did we not need -DBUILD_SYCL_WITH_CUDA?

I think we do as it it should be enabling cuda backend in sycl targets here

Additionally, I'm running into issues with compute-benchmarks trying to compile l0: what commands are you using to run main.py to circumvent this (although, there is perhaps a cmake configure setting we're missing here)?

python /llvm_path/devops/scripts/benchmarks/main.py /build_path/benchs/bench_cuda --sycl /build_path --output-html --verbose --adapter cuda --timeout 12000

One of the things that I ran into even with l0 is that I need to increase the timeout as compute-bench build takes some time locally.

@intel/llvm-gatekeepers please merge! failures should be unrelated to the PR. Thanks!