CONFERENCE / ICICMCT'23

In image-related applications like image processing, image recognition, machine learning, and digital signal processing, approximate computing effectively lowers computational costs. With this approach, there is a trade-off between computational accuracy and the circuit's area, delay, and power requirements. For various applications, the accuracy requirements could change, nevertheless. In some circumstances, precise results are necessary, whereas, in others, tolerable errors are permitted for reduced power and speed. A dynamically reconfigurable hybrid adder built on the radix-4 adder (RD4A) with Brent-Kung (BK) is included in the proposed work. The Accuracy configurable Radix 4 adder (ACRA) is a hybrid adder that can be dynamically reconfigured and is based on the radix-4 adder. The power gating approach is used by the ACRA or Radix-4 adder in the LSB section to dynamically turn off the partial logic gates of an adder element in order to compute accurate or approximative results. The Brent-Kung adder is a parallel prefix adder (PPA) that improves performance and requires less chip area because it has less wiring congestion and increases the regularity of the adder structure. It is used in the hybrid adder's MSB section. The partial sum of one adder element is changed while the adder is operating in the approximate mode in order to decrease the error gap between the inaccurate and accurate results. Additionally, the proposed adder's performance can be assessed when used in image processing applications like image smoothening.