Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
International Journal of Engineering & Technology, 7 (2.8) (2018) 413-418
International Journal of Engineering & Technology
Website: www.sciencepubco.com/index.php/IJET
Research Paper
Design and comparative analysis of inexact speculative
adder and multiplier
K. Hari Kishore, B. K. V. Prasad, Y. Manoj Sai Teja*, D. Akhila, K. Nikhil Sai, P. Sravan Kumar
VLSI Research Group, Department of ECE, Koneru Lakshmaiah Education Foundation,
Vaddeswaram, Guntur, Andhra Pradesh, India 522502
*Corresponding Author Email: manojsaiteja@gmail.com
Abstract
A Carry look ahead adder is a sort of the summer used in the logic design of the digital systems. The CLA boost up the speed by
decreasing the measure of duration needed to calculate the carry bits. The CLA based outline of the inexact speculative adder is pipelined
architecture to incorporate couple of logic paths along its basic way and in this manner, improving the recurrence of operation. This paper
presents the comparative analysis of the pipelined inexact speculative adder and the general carry look ahead adder and showed that the
delay is reduced to 48.27% when compared to carry look ahead adder and also we have designed the pipelined multiplier using the
Inexact speculative adders and observed that the delay is reduced to 48.32% when compared to the normal multiplier. This entail Xilinx
ISE Design Suite 14.5 Tool.
Keywords: Inexact Speculative Adder, Multiplier, Carry Look Ahead Adder, Pipelining, Delay.
1. Introduction
The inexact speculative adder is designed in such a way that the
carry propagation chain is splitted into multiple paths which are
concurrently executed. Each path again contains a error
compensation block, sub-adder block and a carry speculator block.
The partial signal is generated by the speculator block and to give
the local sum the sub adder block uses carry in. Faulty sums are
corrected by the compensation block which corrects either local
sum or reduces the magnitude in the error. The carry for
speculation block is generated by using the bits in carry look ahead
sourced either by dynamic or static input. Errors are distributed
evenly by the latter in ETBA. Speculation faults are detected by
the COMP block by comparing carry out and generated carry from
the speculation block. The COMP block is implemented between
ADD blocks. The addition is implemented by following five steps
mainly. First for each sub-adder block , a carry-in is speculated
from short carry propagation chain and next based on that carry in
local sum is calculated by the sub adder. Faulty speculation is
detected by comparing the carry in and the carry out of the sub
adder. Correction of local sum is done when there is wrong
speculation. Error magnitude is minimized by the balance of
preceding sum bits if the correction is not possible. The
inconsistency between expected carry and speculated carry is
detected by the COMP block by using the XOR gate because of
which error flag is created that triggers activation either error
reduction or correction compensation techniques. The three main
advantages of this technique is that optimization of block size,
speculation and correction trade-off and error minimization and
failed correction [1].
In the speculative addition with the variable latency CSPA block
has carry and adder circuits. Carry predictor helps in prediction of
carry out bit. And adder has three internal components namely
multiplexer which consists of multi bit, the sum generator and also
an internal carry generator. Carry out bit of the adder block is
predicted by the carry predictor. Internal signals to be used in the
sum generator are produced by the internal carry signals.
Multiplexer takes the input from the carry predictor and selects on
of the carry signals. Sum generator calculates the partial sum bits
of adder block. The proposed method works as follows, internal
carry generators and predictors work simultaneously when input
patterns arrive. The carry out bit of the predictor is given to next
block adder and it is the selector for multiplexer and next internal
carry signal is determined and is sent to the sum generator. Error
recovery and detection circuit is also used. Error signal is 0 when
there is no error and valid signal becomes 1 when an error occurs
error and valid signal becomes vice versa during error input
registers becomes disable and no input is taken [2].
The carry speculative adder with modification of carry generators
that uses only less number of gates. In order to get data
continuously into circuit the data latching circuit is used. To get
results accurately the error recovery and detection circuit is
included. The CSPA is implemented in a way that in the modified
adder block sum and carry generators are separated by an logic
which results in increase of area and consumption of power. To
minimize the area two carry generators are for carry in=1 and
carry in=0 used instead of one. AND gate is replaced in the place
of 1bit carry generator for carry in=1 and carry in=0 it is replaced
with OR gate. Two types of modified carry generators are used in
place of two carry generators which has one gate delay with very
less area.
These generators produce carry without using Cin bit. In the
design of variable latency CSPA the circuits such as error
recovery, error detection, data latching are used [3]. When the
input is received the output sum is given by the circuit (VLCSPA).
Error block signals and error signals is given by the error detection
circuit. Multiplexer which is multibit takes the input from the
recovery circuit and cspa when it is 1 signal recovery is from
recovery circuit and when it is 0 result is selected from CSPA.
Input registers become disabled when an error occurs and no input
is taken and XOR gate is used in place of not gate and ex-or
operation is used between error signal and its compliment.