How to build "mlir-cuda-runner"

MLIR
1

Hi,

Can anyone give me instructions on how to build mlir-cuda-runner. I use the following building commands but no mlir-cuda-runner generated:
$cmake -G Ninja …/llvm -DLLVM_BUILD_EXAMPLES=ON -DLLVM_ENABLE_PROJECTS=“mlir”
$cmake --build . --target check-mlir

Thank you so much for your help!

2

You need to pass -DMLIR_CUDA_RUNNER_ENABLED=ON.
It isn’t enabled by default as it’ll fail without Cuda available I believe.

3

Yes. It’s true. You are correct! Thank you so much!

4

I think this isn’t documented anywhere! Should there be a page for all -DMLIR_* options on mlir.llvm.org?

5

Ideally it’d be auto-generated from the code to avoid getting out of sync.

That said, all CMake options are in the top-level CMakeLists.txt

1 Like