Dependency analysis and loop transformation characteristics of auto-parallelizers

Automatic Parallelization play a key role in parallelizing legacy codes and manually written single-threaded codes. Auto-parallelizers like Cetus, Par4all, Pluto and Parallware help to parallelize sequential code without any manual intervention. Programmers working in the domain of parallel programming meet with challenges in maintaining the semantics of the code by preserving the dependencies of their original code. Parallelization in loops is significant since it takes maximum time for execution. It is to be noted that all loops are not parallelizable when the instructions in loops are dependent. Dependency analysis has become absolutely essential to retain the order in which the original code flows so that the code could run in parallel. Loop optimizations are significant in automatic parallelization for maximizing parallelism. This paper gives an outline of parallelization mechanisms mainly dependency analysis and loop transformation techniques. A case study has been carried out with auto-parallelizer tools to study and analyze their ability in locating and handling the dependencies by applying code transformation techniques in serial code while converting to parallel codes.