Fine-grain Parallelism using Multi-core, Cell/BE, and GPU Systems