This is a template for a project applying some simple static analysis together with dynamic analysis to extract information about program behavior. It involves instrumenting an LLVM module to produce a new program that collects path frequency information for acyclic functions.

Building with CMake

1. Create a new directory for building.

    mkdir pathprofilerbuild
   

2. Change into the new directory.

    cd pathprofilerbuild
   

3. Run CMake with the path to the LLVM source.

    cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=True \
        -DLLVM_DIR=</path/to/LLVM/build>/lib/cmake/llvm/ ../path-profiler-template
   

4. Run make inside the build directory:

    make
   

This produces a path profiler tool called bin/pathprofiler and supporting libraries in lib/.

Note, building with a tool like ninja can be done by adding -G Ninja to the cmake invocation and running ninja instead of make.

Running

First suppose that you have a program compiled to bitcode:

clang -g -c -emit-llvm -S ../path-profiler-template/test/c/03-multipleprofiled.c -o example.ll

Running the path profiler:

bin/pathprofiler example.ll -o example
./example

When you have successfully completed the exercise, running an instrumented program like ./example in the above example should produce a file called path-profile-results in the current directory. Running the path profiler as follows then analyzes the results:

bin/pathprofiler example.ll -p path-profile-results

Again, when you have successfully completed the exercise, this produces another file in the current directory called top-five-paths.csv. This file contains the five most frequently executed paths, with each line formatted as follows:

<# Occurrences>, <Function Name> (, <File Name>, <Line #>)+