OpenMP Loop Check

OpenMP Loop Check is a Clang utility for verifying if for loops in C/C++ files can be parallelized with OpenMP. Currently, it reports the following:

• Loops without init section.
• Loops without condition section.
• Loops without increment section.
• Loops with 'break' statements in the body

Build Instructions

Download the LLVM sources:

git clone https://github.com/llvm/llvm-project.git
cd llvm-project

Clone this repository into the clang-tools-extra folder of the LLVM sources:

cd clang-tools-extra
git clone https://github.com/American-Chip/omp-loop-check.git

Add omp-loop-check as a subdirectory into the CMakeLists.txt file of clang-tools-extra:

add_subdirectory(omp-loop-check)

Configure LLVM build with clang and clang-tools-extra projects enabled. E.g:

cd ..
cmake -S llvm -B build -G Ninja -DLLVM_ENABLE_PROJECTS="clang;clang-tools-extra" -DCMAKE_BUILD_TYPE=Release

Build omp-loop-check. E.g.:

cd build
ninja omp-loop-check

Usage

Now you can execute the omp-loop-check utility from the LLVM build folder.

bin/omp-loop-check <C/C++ file> --

You can also analyze more then one file:

bin/omp-loop-check <C/C++ file 1> <C/C++ file 2> <C/C++ file 3> --

You can add headers and macro definitions in two mutually exclusive ways:

1. Writing them explicitely using -- at the end of the command:

   bin/omp-loop-check <C/C++ file> -- [-I<header folder> -D<macro>]

2. Using the -p option to specify the containing folder of a project configuration stored on a compile_commands.json file:

   bin/omp-loop-check <C/C++ file> -p <compile_commands.json folder>