MLIR Build

[jeromeku]

Learning MLIR currently and trying to build Triton from source to understand entire compilation pipeline.

I'm getting a cmake generation error:

CMake Error at /notebooks/CUDA-CPP/MLIR/llvm-project/build_mlir/lib/cmake/llvm/AddLLVM.cmake:556 (add_dependencies):
  The dependency target "MLIRGPUOps" of target "obj.TritonGPUIR" does not
  exist.
Call Stack (most recent call first):
  /notebooks/CUDA-CPP/MLIR/llvm-project/build_mlir/lib/cmake/mlir/AddMLIR.cmake:347 (llvm_add_library)
  /notebooks/CUDA-CPP/MLIR/llvm-project/build_mlir/lib/cmake/mlir/AddMLIR.cmake:626 (add_mlir_library)
  lib/Dialect/TritonGPU/IR/CMakeLists.txt:1 (add_mlir_dialect_library)

Note: I've set MLIR_DIR and LLVM_ROOT_DIR to the respective MLIR and LLVM build dirs.

Are there specific flags I need to set when building MLIR such that MLIRGPUOps is built? Is this referring to the built-in gpu dialect?

I built MLIR with the following config:

cmake  -G Ninja  \
     -DMLIR_ENABLE_BINDINGS_PYTHON=ON \
     -DPYTHON3_EXECUTABLE=$PYTHON_EXECUTABLE \
    -DLLVM_ENABLE_PROJECTS=mlir \
    -DLLVM_BUILD_EXAMPLES=OFF \
    -DLLVM_TARGETS_TO_BUILD="Native;NVPTX;AMDGPU" \
    -DCMAKE_BUILD_TYPE=Release

[jeromeku]

FYI: I'd cloned a more recent build of MLIR, which changed GPUOps to GPUDialect. See llvm/llvm-project@61223c4.