Design and Analysis of Majority Logic-Based Approximate Adders and Multipliers
read more
Citations
Non-Volatile Approximate Arithmetic Circuits Using Scalable Hybrid Spin-CMOS Majority Gates
A high-speed 4-bit Carry Look-Ahead architecture as a building block for wide word-length Carry-Select Adder
Design of Majority Logic-Based Approximate Booth Multipliers for Error-Tolerant Applications
Atomic Silicon Quantum Dot: A New Designing Paradigm of an Atomic Logic Circuit
Low-power and high-speed approximate 4:2 compressors for image multiplication applications in CNFETs
References
A device architecture for computing with quantum dots
Approximate computing: An emerging paradigm for energy-efficient design
A Survey of Techniques for Approximate Computing
Low-Power Digital Signal Processing Using Approximate Adders
New Metrics for the Reliability of Approximate and Probabilistic Adders
Related Papers (5)
Frequently Asked Questions (11)
Q2. How many times does MLAC4 reduce the number of majority gates?
The MLAC4 based 8×8 multiplier using 2×2 multipliers (p=10) reduces the number of majority gates by up to 48%, the number of inverters by up to 67%, and the delay by up to 47% compared to the exact counterpart, achieving a significant decrease of hardware without incurring in large errors.
Q3. What are the parameters used for the coherence vector simulation engine?
Approximate PPR circuitry design for 8 × 8 MLAMs: (a) MLAC (2 outputs) based (p=10), and (b) MLAC (3 outputs) based (p=10).ing parameters have been used for the coherence vector simulation engine: Number of Samples: 128000; Convergence Tolerance: 0.00001; Radius of Effect: 55 nm.
Q4. What is the effect of the weight of the compensation bits on the NMED?
The increase of weight will ultimately increase the NMED; so, when the weight of the compensation bits remains unchanged, the NMED will increase linearly.
Q5. How many lines can be generated using MLAMs?
the PPs can be compressed by utilizing eight 1-bit full adders to generate 2 lines requiring an additional 11-bit full adder.
Q6. What is the effect of a complement bit on the final NMED?
To measure the importance of each complement bit C2i x as well as to determine the ignored items, an influence factor denoted by PC2i xn is defined; this is required to show the impact of a complement bit on the final NMED.
Q7. What is the way to compare MLAMs to other similar designs?
By combining multiple approximate techniques (such as the proposed MLACs and approximate PPR circuitry) with the so-called complement bits, ML based multi-bit AMs have been proposed: an influence factor has been defined to measure the importance of different complement bits; selection of the complement bits has also been pursued by an in-depth analysis depending on the size of multipliers; multiple MLACs has been proposed based on MLAFAs or K-Map simplification, and has been employed in the approximate PPR circuitry design for 8 × 8 MLAMs.
Q8. What is the way to calculate the PP of a row?
This depends on the distribution of the generated partial products (PPs) and the compensation bits, such that the PP of two rows (or a carry in the lowest order) can be obtained.
Q9. What is the first method used to design the 2-bit MLAFA?
The first designs merge the proposed MLAFA2 and MLAFA1 (hence the hybrid nature); the second method designs the 2-bit MLAFA using a truth table reduction process.
Q10. How many errors are there in MLAC3?
Fig. 3(b) employs two 1-bit MLAFA1s [12] to substitute the two 1-bit EFAs, so resulting in 12 errors out of 32 cases; to improve the accuracy, the 1-bit MLAFA1 replaces one of the two EFAs in MLAC3.
Q11. What is the performance of the proposed MLAMs?
Compared with other designs, the utilization of a 2 × 2 approximate multiplier as a module and the proposed approximate compression result in the best performance.