Computing systems in the next decade will continue to expand the landscape from pervasive wearable and embedded devices to exascale high performance computing systems and data centers. However, significant challenges must be addressed in order to meet the requirements and expectations of future computing systems. In this talk, I will first discuss several important technology and application trends and briefly summarize our past research in improving power, performance and resource efficiency for various system components. I will then present two recent works on enabling energy-proportional adaptive on-chip networks, which serve as the backbone to connect general-purpose processing cores, IP blocks and various hardware accelerators in MPSoCs and chip multiprocessors. The two novel designs, namely Node-Router Decoupling (NoRD) and Minimal Performance Penalty Power-gating (MP3), allow on-chip routers to be powered on/off dynamically based on application characteristics while maintaining network connectivity and incurring negligible performance overhead. These works greatly extend the achievable operating range of many-core processors (e.g., minimal and maximum design points), thus facilitating their use not only in high performance computing but also in embedded and mobile devices. Finally, I will conclude the talk by mentioning several promising research lines that are worthy of pursuit in the near future.