Power-aware Programming with GPU Accelerators

On-chip parallelism with GPU accelerators is now ubiquitous and has received significant attention in the past few years. GPU is becoming an integral part of mainstream computing systems with highly parallel, multithreaded, many-core processors of great computational power and high memory bandwidth. Finding the best tradeoff between performance and power efficiency is more challenging than mere performance tuning. To find the principles of power-aware programming with GPU accelerators, we abstract a set of primitives from program statements. These power consumption values of primitives are helpful for power estimation during high-level program development.