Practical Strategies for Power-Efficient Computing (original) (raw)

A Performance On-Demand Approach to Power-Efficient Computing

Complexity-Adaptive Processing (CAP) addresses increasing microprocessor power dissipation through on-the-fly, low-cost hardware adaptation and related circuit techniques so as to better match hardware complexity and speed to application demands. Specific results include adaptive processing elements and hardware/software control techniques, a Multiple Clock Domain processor that saves energy via fine-grain voltage scaling, power-efficient issue queue and register file techniques, a low-leakage dynamic logic circuit and associated control logic for functional units, multi-threaded power and noise reduction, efficient on-chip dc-dc conversion and clock control circuits, low power domino logic and interface circuits, and interconnect width optimization for low power. Overall, a several-fold reduction in power is demonstrated via the collective application of these various techniques.

Exploring the potential of architecture-level power optimizations

2005

This paper examines the limits of microprocessor energy reduction available via certain classes of architecture-level optimization. It focuses on three sources of waste that consume energy. The first is the execution of instructions that are unnecessary for correct program execution. The second source of wasted power is speculation waste–waste due to speculative execution of instructions that do not commit their results. The third source is architectural waste. This comes from suboptimal sizing of processor structures.

Changing computing paradigms towards power efficiency

2014

Power awareness is fast becoming immensely important in computing, ranging from the traditional high-performance computing applications to the new generation of data centric workloads. In this work, we describe our efforts towards a power-efficient computing paradigm that combines low- and high-precision arithmetic. We showcase our ideas for the widely used kernel of solving systems of linear equations that finds numerous applications in scientific and engineering disciplines as well as in large-scale data analytics, statistics and machine learning. Towards this goal, we developed tools for the seamless power profiling of applications at a fine-grain level. In addition, we verify here previous work on post-FLOPS/W metrics and show that these can shed much more light in the power/energy profile of important applications.