Manuscript Title:




DOI Number:


Published : 2021-11-10

About the author(s)

1. C.V. KRISHNA REDDY​ - Nalla Narasimha Reddy Education Society Group of Institutions, Hyderabad

Full Text : PDF


For low power application, power is a major design constraint. Its consumption during software program execution is a critical consideration for constructing low power embedded systems. This software power can be diminished in a variety of ways. Software related power could be lowered by changing the instruction of a code. When it comes to building embedded applications, power is becoming a key limitation. To predict the to estimate the present power consumption of an embedded digital signal processor, a new power analysis model was developed in this research. Power is becoming a major constraint when it comes to developing embedded applications. Power analysis methods that use circuitlevel or architectural-level modeling to forecast the power consumption of a particular piece of software are either ineffective or incorrect. This paper develops an embedded DSP instruction-level power analysis model using physical current measurements. There are major discrepancies between the software power model for this unique DSP processor and previous power models for different general purpose commercial microprocessors. The circuit condition affects an instruction stream's power consumption more noticeably in this DSP processor. Dual memory access and instruction pair packing are also possible because of the processor's architecture, which includes these two characteristics. The amount of energy saved as a result of implementing these features is being studied. The processor's on-chip Booth multiplier consumes a significant amount of energy while running DSP programs. A microarchitectural power model for the multiplier is developed and evaluated for further power minimization. Novel instruction-level power model-driven scheduling approach is offered to harness all of the above impacts. To demonstrate the success of this method, several example programmes are presented. There have been energy savings ranging from 26 percent to 73 percent. These energy savings are real, as evidenced by tangible measurements. It's worth noting that the energy savings are basically free. It is obtained by modifying software and so does not require any software. Instead of the traditional analysis the power consumption of an algorithm as a whole is evaluated at the behavioral level, which is done at the instruction level. Our energy models for a Texas Instruments DSP provide some findings, but not all of them. To provide an overview of embedded DSP power consumption models and energy-saving methods at the software level, this article was written. The literature-based software level power consumption models are described, together with their benefits and drawbacks. For the processor, an improved software level power consumption model is proposed, which is shown to attain an accuracy of 96.8% or higher across a variety of benchmarks.


DPS, methodology and analysis, levels, use of DPS, advantages, Low power, high power, techniques.