This presentation introduces a new DSEL, StreamPU, for efficient execution of streaming applications on multicore and heterogeneous CPUs. First, the DSEL is detailed: the basic components, the expressiveness, some optimizations and a comparison with State-of-the-art steaming DSLs are discussed. Then, the multi-threaded runtime is introduced with mainly two levels of parallelism: pipeline and replication. Finally, the DSEL and the runtime are evaluated over two applications representative of the real-world.