[P0] Verify the clang++ version used in the test #18

[yuqisun]

Test to print LLVM version of Github Actions.

[tancheng]

Hi @yuqisun, so the apt-installed method works on your local machine/docker?

[yuqisun]

Hi @yuqisun, so the apt-installed method works on your local machine/docker?

Nope :(, revert just now, looks MLIR is not included in apt-install. Have to build by source code.

[tancheng]

Hi @yuqisun, so the apt-installed method works on your local machine/docker?

Nope :(, revert just now, looks MLIR is not included in apt-install. Have to build by source code.

What about sudo apt install llvm-14 clang-14 libclang-14-dev libmlir-14-dev mlir-14-tools? Can try on your local first? (we prefer llvm-19 though)

[yuqisun]

Hi @yuqisun, so the apt-installed method works on your local machine/docker?

Nope :(, revert just now, looks MLIR is not included in apt-install. Have to build by source code.

What about sudo apt install llvm-14 clang-14 libclang-14-dev libmlir-14-dev mlir-14-tools? Can try on your local first? (we prefer llvm-19 though)

Still doesn't work. We are depending on MLIRArithDialect, MLIRFuncDialect target which are not installed by apt.
Your original manual step is also the recommendation of deepseek/copilot, compile and install from source code.

CMake Error at /usr/lib/llvm-14/lib/cmake/llvm/AddLLVM.cmake:518 (add_dependencies):
  The dependency target "MLIRArithDialect" of target "obj.NeuraArithToNeura"
  does not exist.
Call Stack (most recent call first):
  /usr/lib/llvm-14/lib/cmake/mlir/AddMLIR.cmake:148 (llvm_add_library)
  lib/Conversion/ArithToNeura/CMakeLists.txt:14 (add_mlir_library)


CMake Error at /usr/lib/llvm-14/lib/cmake/llvm/AddLLVM.cmake:518 (add_dependencies):
  The dependency target "MLIRFuncDialect" of target "obj.NeuraArithToNeura"
  does not exist.
Call Stack (most recent call first):
  /usr/lib/llvm-14/lib/cmake/mlir/AddMLIR.cmake:148 (llvm_add_library)
  lib/Conversion/ArithToNeura/CMakeLists.txt:14 (add_mlir_library)

[yuqisun]

Hi @tancheng ,

1. Looks MLIR has to be installed by source code
2. Upgraded to build from LLVM-19 branch
3. Added Dockerfile for container environment setup

Thanks,

[tancheng]

This is for?

[yuqisun]

Added this because Ninja cannot auto fix RPATH:

CMake Error at cmake/modules/AddLLVM.cmake:1029 (add_executable):
The install of the llvm-tblgen target requires changing an RPATH from the
build tree, but this is not supported with the Ninja generator unless on an
ELF-based or XCOFF-based platform. The CMAKE_BUILD_WITH_INSTALL_RPATH
variable may be set to avoid this relinking step.
Call Stack (most recent call first):
cmake/modules/TableGen.cmake:181 (add_llvm_executable)
utils/TableGen/CMakeLists.txt:32 (add_tablegen)

[tancheng]

Let's also have DCMAKE_EXPORT_COMPILE_COMMANDS=ON for VS Code users (to avoid warning in the IDE for include paths), I just set it up with copilot.

[yuqisun]

Added.

[tancheng]

Should we also mention we prefer the clang is also in llvm-19 version?

I am wondering, when we build LLVM-19.x from source, it is compiled using the built-in clang, which may not be llvm-19, right? Then if user install clang in llvm-19, should we update this

dataflow/test/c2llvm2mlir/test.mlir

Line 5 in ff914bc

// RUN: clang++ -S -emit-llvm -o %t-kernel.ll kernel.cpp

to distinguish the system clang and the user-installed clang?

[yuqisun]

How about now: Compile both clang and mlir and add note at the beginning of test.mlir to recommend user to add llvm-19 clang to PATH.

-DLLVM_ENABLE_PROJECTS="clang;mlir" --- To compile clang in llvm-19 but take more time.

[tancheng]

I may already ask, but what is this clang version?

[yuqisun]

Default install version is 14.

[yuqisun]

Image is pushing in progress:

The push refers to repository [docker.io/cgra/neura]
3b5d14681ffd: Pushing [==================================================>]  12.92GB
346f14bf17b9: Layer already exists

[yuqisun]

Uploaded to cgra/neura:v1.