This is a thread error detector for detecting synchronization errors in C, C++, and Fortran programs that include POSIX threads.
Helgrind can detect three classes of error. Firstly, it can detect the incorrect use of the API. For example, it can unlock a mutex that is already unlocked, unlock a mutex that was locked by a different thread, not checking the return value of certain Pthread functions. Secondly, it monitors the order in which threads acquire locks and thus detects potential deadlocks which could arise from the formation of cycles of locks. Finally, it detects data races which can happen when two threads access a shared memory location without using suitable locks or other synchronization to ensure single-threaded access.
Using Helgrind is simple, you just need this command:
# valgrind --tool=helgrind <program>
It prints problems and potential problems as it finds them. You can direct these messages to a file by adding --log-file=<filename>.