As computer architectures have become more parallel and more heterogeneous, we continue to primarily use conventional programming and execution models when developing applications. These entrenched CPU-centric models limit parallelism, incur redundant and unnecessary computation, create unnecessary data copies, and limit the potential of emerging architectural components. In this post-Moore's law era, where per-core performance improvements from generation to generation have slowed and architectures are limited by power limits rather than area limits, it is important for us to rethink the programming and execution models so as to avoid redundancies throughout the system.

To achieve “no-frills computing”, we need to eliminate redundant computation, avoid unnecessary data movement, and process data in the architectural component that incurs the least redundance. In this talk, I will share my research experiences in “no-frills computing”. I will introduce my research projects including: (1) data-triggered threads (DTT), an innovative programming model that initiates parallelism only when data is generated and changed in the system. (2) Gullfoss, a system that redesigns the software interface and runtime system to streamline data movement in data-intensive, heterogeneous computers. (3) Morpheus-SSD that utilizes computing resources inside storage devices to create more efficient applications

Hung-Wei Tseng is a postdoctoral researcher for the Non-volatile Systems Laboratory of the Department of Computer Science and Engineering at University of California, San Diego working with Professor Steven Swanson. He is now designing innovative computer architectures and storage systems for data-intensive applications. In addition to computer architecture and storage systems, he also has research experiences in many different areas including programming languages, compilers, embedded systems, computer networks and bioinformatics. He got his PhD in the Department of Computer Science and Engineering at University of California, San Diego, under the advising of Professor Dean Tullsen. His thesis work, “Data-triggered threads” was selected by IEEE Micro 'Top Picks from Computer Architecture' in 2012.