Automatic Parallelization of Sequential C Code

As the cost of hardware declines and the demand for computing power increases, it is becoming increasingly popular to turn to cluster computing. However, in order to gain the benefits of cluster computing, an existing software base must be converted to a parallel equivalent, or a new software base must be written. Both options require a developer skilled in both parallel programming, as well as the problem domain at hand. The ability to automate a conversion from sequential C code to a cluster-based equivalent offers a developer the power of parallel computing with a minimal learning curve. The following paper describes an ongoing project with the goal of automating the conversion from sequential C code to cluster-based parallel code. Currently the project relies on user input to guide the automation process, focusing on loop level parallelization. Long term goals center on using dependency analysis to automate the parallelization process.