This paper presents an empirical approach to inferring low power code generation techniques for VLIW processors. Architectural usage variables are used to generate equations for power prediction which are in turn used to infer new code generation techniques for low power. Unlike previous techniques, the methodology empirically derives a power prediction equation and then based upon the coefficients of the architectural-usage variables identifies new VLIW code generation techniques for low power. The approach is illustrated using functional unit usage within a VLIW architecture and identifies a new operation rebinding technique for low power which improved power dissipation up to 18%. The approach is general and results are verified with real power measurements. This result is important for developing a general methodology for power minimization of embedded DSP software since low power is critical to complex DSP applications in many cost sensitive markets.

This paper presents a comparison of statistically-derived power prediction models at the algorithmic, instruction, and architectural levels for embedded high performance DSP processors. The approach is general enough to be applied to any embedded DSP processor. Results from 168 power measurements of DSP code show that power can be predicted at instruction and architecture levels with less than 2% error. This result is important for developing a general methodology for power characterization of embedded DSP software since low power is critical to complex DSP applications in many cost sensitive markets.