High-Level Support for Pipeline Parallelism on Many-Core Architectures