Building hipify-clang on Windows#

Building LLVM < 10.0.0#

Download older LLVM + Clang sources.

set ROOT_DIR=%cd%

mkdir build dist
cd build

cmake -G "Visual Studio 16 2019" -A x64 -Thost=x64 -DCMAKE_INSTALL_PREFIX=../dist -DLLVM_TARGETS_TO_BUILD="" -DLLVM_ENABLE_PROJECTS="clang" -DLLVM_INCLUDE_TESTS=OFF -DCMAKE_BUILD_TYPE=Release ../llvm-project/llvm

# Run Visual Studio 16 2019, open the generated "LLVM.sln", build all, and build the "INSTALL" project.

Building HIPIFY#

cd %ROOT_DIR%

git clone https://github.com/ROCm/HIPIFY.git

cd build

# To ensure LLVM is found, or in the case of multiple LLVM instances,
# specify the path to the root folder containing the LLVM distribution.
cmake -G "Visual Studio 17 2022" -A x64 -Thost=x64 -DCMAKE_PREFIX_PATH="../dist" -DCMAKE_INSTALL_PREFIX="../dist" -DCMAKE_BUILD_TYPE=Release ../hipify

# Run Visual Studio 17 2022, open the generated LLVM.sln, build all, and build project "INSTALL".
# Alternatively, you can build using "msbuild INSTALL.vcxproj /m" using the developer command prompt.

You can find the binary at ./dist/bin/hipify-clang or at the folder specified by the -DCMAKE_INSTALL_PREFIX option.

Testing hipify-clang#

hipify-clang is equipped with unit tests using LLVM lit or FileCheck.

We recommend that you build LLVM+Clang from sources, as prebuilt binaries are not exhaustive for testing.

  • Install CUDA version 7.0 or greater.

    In case of multiple CUDA installations, specify the particular version using DCUDA_TOOLKIT_ROOT_DIR option:

    -DCUDA_TOOLKIT_ROOT_DIR="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9"
    
    -DCUDA_SDK_ROOT_DIR="C:/ProgramData/NVIDIA Corporation/CUDA Samples/v12.9"
    
  • [Optional] Install cuTensor:

    To specify the path to cuTensor, use the CUDA_TENSOR_ROOT_DIR option:

    -DCUDA_TENSOR_ROOT_DIR=D:/CUDA/cuTensor/2.4.0.0
    
  • [Optional] Install cuDNN belonging to the version corresponding to the CUDA version:

    To specify the path to cuDNN, use the CUDA_DNN_ROOT_DIR option:

    -DCUDA_DNN_ROOT_DIR=D:/CUDA/cuDNN/9.15.0
    
  • [Optional] Install CUB 1.9.8 for CUDA < 11.0 only; for CUDA >= 11.0, the CUB shipped with CUDA will be used for testing.

    To specify the path to CUB, use the CUDA_CUB_ROOT_DIR option (only for CUDA < 11.0):

    -DCUDA_CUB_ROOT_DIR=D:/CUDA/CUB
    
  • Install Python version 3.0 or greater.

  • Install lit and FileCheck; these are distributed with LLVM.

    lit:

    python %ROOT_DIR%/llvm-project/llvm/utils/lit/setup.py install
    

    Starting with LLVM 6.0.1, specify the path to the llvm-lit Python script using the LLVM_EXTERNAL_LIT option:

    -DLLVM_EXTERNAL_LIT=%ROOT_DIR%/llvm-project/llvm/utils/lit/llvm-lit.py
    

    FileCheck:

    Copy from %ROOT_DIR%/llvm-project/llvm/utils/FileCheck to CMAKE_INSTALL_PREFIX/dist/bin.

  • To run OpenGL tests successfully, you need to install OpenGL headers and libraries.

    No installation required. All the required headers are shipped with the Windows SDK.

  • Set the HIPIFY_CLANG_TESTS option to ON: -DHIPIFY_CLANG_TESTS=ON.

  • Build and run tests. Run Visual Studio 17 2022, open the generated hipify-clang.sln, to build the project test-hipify.

Windows testing#

Tested configurations:

LLVM

CUDA

cuDNN

Visual Studio

CMake

Python

4.0.0 - 5.0.2

7.0 - 8.0

5.1.10 - 7.1.4

2015.14.0, 2017.15.5.2

3.5.1  - 3.18.0

3.6.4 - 3.8.5

6.0.0 - 6.0.1

7.0 - 9.0

7.0.5  - 7.6.5

2015.14.0, 2017.15.5.5

3.6.0  - 3.18.0

3.7.2 - 3.8.5

7.0.0 - 7.1.0

7.0 - 9.2

7.0.5  - 7.6.5

2017.15.9.11

3.13.3 - 3.18.0

3.7.3 - 3.8.5

8.0.0 - 8.0.1

7.0 - 10.0

7.6.5

2017.15.9.15

3.14.2 - 3.18.0

3.7.4 - 3.8.5

9.0.0 - 9.0.1

7.0 - 10.1

7.6.5

2017.15.9.20, 2019.16.4.5

3.16.4 - 3.18.0

3.8.0 - 3.8.5

10.0.0 - 11.0.0

7.0 - 11.1

7.6.5  - 8.0.5

2017.15.9.30, 2019.16.8.3

3.19.2

3.9.1

11.0.1 - 11.1.0

7.0 - 11.2.2

7.6.5  - 8.0.5

2017.15.9.31, 2019.16.8.4

3.19.3

3.9.2

12.0.0 - 13.0.1

7.0 - 11.5.1

7.6.5  - 8.3.2

2017.15.9.43, 2019.16.11.9

3.22.2

3.10.2

14.0.0 - 14.0.6

7.0 - 11.7.1

8.0.5  - 8.4.1

2017.15.9.57, 5 2019.16.11.17, 2022.17.2.6

3.24.0

3.10.6

15.0.0 - 15.0.7

7.0 - 11.8.0

8.0.5  - 8.8.1

2019.16.11.25, 2022.17.5.2

3.26.0

3.11.2

16.0.0 - 16.0.6

7.0 - 12.2.2

8.0.5  - 8.9.5

2019.16.11.29, 2022.17.7.1

3.27.3

3.11.4

17.0.1 6 - 18.1.8 7

7.0 - 12.3.2

8.0.5  - 9.13.1

2019.16.11.50, 2022.17.14.12

4.1.1

3.13.6

19.1.0 - 20.1.8

7.0 - 12.8.1

8.0.5  - 9.15.0

2019.16.11.51, 2022.17.14.14

4.1.1

3.13.7

21.1.0 - 21.1.7

7.0 - 12.9.1

8.0.5  - 9.15.0

2019.16.11.51, 2022.17.14.14, 2026.18.0.2

4.2.0

3.14.0

5 LLVM 14.x.x is the latest major release supporting Visual Studio 2017.

To build LLVM 14.x.x correctly using Visual Studio 2017, add -DLLVM_FORCE_USE_OLD_TOOLCHAIN=ON to corresponding CMake command line.

You can also build LLVM < 14.x.x correctly using Visual Studio 2017 without the LLVM_FORCE_USE_OLD_TOOLCHAIN option.

6 Note that LLVM 17.0.0 was withdrawn due to an issue; use 17.0.1 or newer instead.

7 Note that LLVM 18.0.0 has never been released; use 18.1.0 or newer instead.

Building with testing support using Visual Studio 17 2022 on Windows 11:

cmake \
-G "Visual Studio 17 2022" \
-A x64 \
-Thost=x64 \
-DHIPIFY_CLANG_TESTS=ON \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_PREFIX=../dist \
-DCMAKE_PREFIX_PATH=%ROOT_DIR%/dist \
-DCUDA_TOOLKIT_ROOT_DIR="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9" \
-DCUDA_SDK_ROOT_DIR="C:/ProgramData/NVIDIA Corporation/CUDA Samples/v12.9" \
-DCUDA_DNN_ROOT_DIR=D:/CUDA/cuDNN/9.15.0 \
-DCUDA_TENSOR_ROOT_DIR=D:/CUDA/cuTensor/2.4.0.0 \
-DLLVM_EXTERNAL_LIT=%ROOT_DIR%/build/Release/bin/llvm-lit.py \
../hipify

The corresponding successful output is (assuming %ROOT_DIR% is D:/LLVM/21.1.7):

-- The C compiler identification is MSVC 19.44.35216.0
-- The CXX compiler identification is MSVC 19.44.35216.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.44.35207/bin/Hostx64/x64/cl.exe - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.44.35207/bin/Hostx64/x64/cl.exe - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- HIPIFY config:
--    - Build hipify-clang    : ON
--    - Test hipify-clang     : ON
--    - Is part of HIP SDK    : OFF
--    - Install clang headers : ON
-- Found LLVM 21.1.7:
--    - CMake module path     : D:/LLVM/21.1.7/dist/lib/cmake/llvm
--    - Clang include path    : D:/LLVM/21.1.7/dist/include
--    - LLVM Include path     : D:/LLVM/21.1.7/dist/include
--    - Binary path           : D:/LLVM/21.1.7/dist/bin
-- ---- The below configuring for hipify-clang testing only ----
-- Found Python: C:/Users/TT/AppData/Local/Programs/Python/Python314/python.exe (found suitable version "3.14.0", required range is "3.0...3.15") found components: Interpreter
-- Found lit: C:/Users/TT/AppData/Local/Programs/Python/Python314/Scripts/lit.exe
-- Found FileCheck: D:/LLVM/21.1.7/dist/bin/FileCheck.exe
-- Initial CUDA to configure:
--    - CUDA Toolkit path     : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9
--    - CUDA Samples path     : C:/ProgramData/NVIDIA Corporation/CUDA Samples/v12.9
--    - cuDNN path            : D:/CUDA/cuDNN/9.15.0
--    - cuTENSOR path         : D:/CUDA/cuTensor/2.4.0.0
--    - CUB path              :
-- Found CUDAToolkit: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9/include (found version "12.9.86")
-- Found CUDA config:
--    - CUDA Toolkit path     : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9
--    - CUDA Samples path     : C:/ProgramData/NVIDIA Corporation/CUDA Samples/v12.9
--    - cuDNN path            : D:/CUDA/cuDNN/9.15.0/include
--    - cuTENSOR path         : D:/CUDA/cuTensor/2.4.0.0/include
--    - CUB path              : C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.9/include/cub
-- Configuring done (4.4s)
-- Generating done (0.1s)
-- Build files have been written to: D:/HIPIFY/build