Accuracy-configurable adder for approximate arithmetic designs
read more
Citations
Approximate computing: An emerging paradigm for energy-efficient design
A Survey of Techniques for Approximate Computing
A low latency generic accuracy configurable adder
Quality programmable vector processors for approximate computing
A Review, Classification, and Comparative Evaluation of Approximate Arithmetic Circuits
References
Trading Accuracy for Power with an Underdesigned Multiplier Architecture
Variable latency speculative addition: a new paradigm for arithmetic circuit design
Design of Low-Power High-Speed Truncation-Error-Tolerant Adder and Its Application in Digital Signal Processing
Enhanced low-power high-speed adder for error-tolerant application
Speeding up processing with approximation circuits
Related Papers (5)
Frequently Asked Questions (13)
Q2. What is the error signal in the pipelined approach?
The error signal holds the input pattern during the error correction and chooses the error-corrected value (SUMcorrect) as an output.
Q3. What is the effect of a metric on the accuracy of an approximate circuit?
The approximate designs produce almost-correct results at the given required accuracy, and obtain power reductions or performance improvements in return.
Q4. What accuracy metric is used to measure error significance?
The authors propose another accuracy metric, ACCinf , which measures error significance as Hamming distance, where Be is the number of error bits and Bw is the bit-width of the data.
Q5. How can the authors estimate the area of the critical path?
delay of the critical path can be expressed with Equation (2) and area can be estimated with Equation (3), where Cdelay and Carea are constants for delay and area, respectively.
Q6. What is the overhead for an error detection and correction system?
With these simple error detection and correction circuits, their proposed adder can be implemented to have variable latency like the previous VLSA adder [12], with a small overhead for an error detection and correction (EDC) system.
Q7. What is the PSNR of the ACA adder?
From the results, the ACA adder has PSNR of 24.5dB, and this suggests that image processing/filtering applications could employ their proposed adder with significant power savings and only small loss in image quality.
Q8. What is the probability of having a correct result in the ith sub-adder?
In the ith sub-adder, errors occur when (1) the LSB part of the result (SUMi[k − 1 : 0]) has all ‘1’ values (probability P = 12k ) and (2) the LSB part([k − 1 : 0]) of the (i + 1)th sub-adder produces a carry bit (probability P = 14 + 1 2 · 1 4 + 1 2 · 1 2 · 1 4+ ...).
Q9. How can the authors achieve 100% correct results when k is less than N/4?
when k is less than N/4, it is impossible to correct all errors and achieve 100% correct results within one clock cycle since the error-correction paths become critical.
Q10. What is the metric of accuracy for a DSP?
in communication systems that mainly handle information data, the number of incorrect bits(Hamming distance) is a more meaningful metric for accuracy – e.g. a (32,28) Reed-Solomon code can correct up to 2-byte errors.
Q11. How can the authors reduce the carry chain depth of sub-adders?
In the proposed adder implementation, to achieve higher performance or lower power consumption, the authors can reduce the carry chain depth (k) of sub-adders (see Table 1).
Q12. How many stages are required to add a ACA adder?
To show the benefit of accuracy configuration, the authors have implemented a 32-bit ACA adder (N = 32, k = 4) with 4-stage pipeline, and compared it with a conventional pipelined adder with an 8-bit CLA in each stage.
Q13. What mechanisms allow timing errors and manage design reliability dynamically?
To overcome consequences of overdesign, several recent mechanisms for variation-resilient design [4] allow timing errors and manage design reliability dynamically.