SOME HISTORICAL NOTES ON NUMBER THEORETIC TRANSFORM

Size: px
Start display at page:

Download "SOME HISTORICAL NOTES ON NUMBER THEORETIC TRANSFORM"

Transcription

1 SOME HISTORICAL NOTES ON NUMBER THEORETIC TRANSFORM M. Bhattacharya *, R.Creutzburg **, and J. Astola * * Institute of Signal Processing Tampere University of Technology P.O. Box 553, Tampere, FIN 33101, FINLAND. {mrinmoy, jta}@cs.tut.fi ** Fachhochschule Brandenburg University of Applied Sciences P. O. Box 2132 D Brandenburg an der Havel, GERMANY creutzburg@fh-brandenburg.de ABSTRACT Modulo arithmetic modulo a prime integer have many interesting properties. Such properties are found in standard books on number theory. Some properties are especially of interest to the signal processing application. It was observed analogy exists between some of them and that cyclic convolution of two sequences modulo a prime integer of two sequences could be computed in integer domain as can be done by Fast Fourier Transform using complex real numbers, leading to exactness of the final result (i.e., free of any roundoff errors). These methods, appropriately named as Number Theoretic Transform, are associated with both advantages and disadvantages. These developments in signal processing algorithms took place following the footsteps of developments of Fast Fourier Transform techniques. This paper traverses some of the developments of the Number Theoretic Transform techniques over time and discusses mostly the initial contributions and efforts made by various researchers. 1. INTRODUCTION Digital signal processing may be termed as arithmetic operations of real number sequences in some way as defined and stipulated by the algorithms in order to achieve the desired result. Also, we note that any real number appropriately scaled is equivalent to an integer and hence digital signal processing is arithmetic operations between numbers and this is where the applicability of number theory starts. Filtering is one of the basic process in signal processing that is represented by convolution of two sequences It is well known that direct computation of convolution of two sequences {x(n)} and {h(n} requires excessive amount of computational effort even for moderate lengths of the sequences. Transform domain methods are used to reduce this effort. The sequences are transformed to the transform domain using the transforms possessing the cyclic convolution property (CCP). Term by term (i.e., pointwise) multiplications are done; the sequence so obtained is inverse-transformed to get the output sequence. The primary method as on these days is Fast Fourier Transform (FFT). Number Theoretic Transform (NTT) has the necessary property of CCP and uses the same architecture as that of FFT and is attractive always due to its exactness i.e., free of round off errors compared to other methods and implementation by simple and real arithmetic for real sequences, in general. However, it is associated with the stringent relation of convolution length N with the choice of modulus M. Efforts to alleviate this problem is generally associated with increase in computational effort in terms of computational structure. The basic requirement of where, N W = 1 (1) j2π / N W = e (2) i.e., W is the N-th root of unity, in computing FFT, is similar to, N α =1 mod M, (3) where, M is the modulus, a prime integer, and α is an integer of order N or N-th root of unity that is required in computing Number Theoretic transform (NTT). It is opined by the authors that interest in NTT may have remained (if not in the lead vis-à-vis FFT), but for the facts that when the NTT schemes were being developed, hardware i.e., bits and bytes (those relates to word length of data sequences) were quite costly those days compared to nowadays, and unlike Fourier spectra, number theoretic spectra of a sequence carries no meaning (meaning that application of NTT would have remained confined generally to convolution; however, it is mentioned that algorithm associated with a basic principle exists wherein convolution can be used to compute FFT and that part of convolution could be computed by NTT as well [11, 47]), in spite of advantages of being free of any round off error due to arithmetic operations, as the

2 computations are done by modulo an integer, and data paths would remain real unlike those in FFT where each data path is actually comprised of two data paths i.e., real and complex parts. The major disadvantage is the stringent relation between the choice of modulus M and the sequence length N, especially in the cases where one wishes to employ simple arithmetic like bit shifts and adds in place of multiplications and chooses a simple modulus of type 2 n ±1 along with simple α like powers of 2. It may be noted that modulo arithmetic operations, modulo a general integer requiring many bits to represent, is more complex. We will see that the efforts of researchers generally mainly revolves around removing and/or reducing the stringent relation between the word length and the convolution length and a suitable value of α. While we will try to adhere to describe the developments chronologically as they appeared in the publications, we will not be able to follow that strictly, as we will be including certain discussions as well to highlight and elaboration (accompanied with illustration) of certain issues related to the development. At times we may appear to switch back and forth. Further, in this paper we will not try to explore or trace the history of the two of the most related topics. Those are: (a) Chinese Remainder Theorem (CRT) and (b) Fermat s theorem; these two are almost inseparable with theory and practices of NTT. CRT helps in two ways: (a) mapping indices of an unidimensional data vector into multidimensional data set and back (after the required operations), when the sequence length N is equal to i p, where i p i s are integer primes, and (b) increasing the dynamic range by using multiple moduli (mutually prime integers) of the type M = M 1 M 2..., where Mi s are mutually prime integers. Burrus [17] provides an excellent exposition of such mappings, while most of the authors generally provides the mapping of input and output indices those are quite similar. Fermat s too well-known theorem [1, 2, 29, 36, 42], p a = a mod p, a, p are integers with p a prime integer, along with its many ramifications are well documented in many books and publications; and this theorem is the basic foundation for the NTT. All the basic essentials and the related relevant portions of number theory and NTT are very well expositioned in [1] and we suggest that all uninitiated readers should read this. This will help them in understanding the cases of single modulus, the cases of multiple moduli, the mapping of indices, the wordlength requirement etc., and to sum up help them to understand the constraints and problems faced in developments of NTT. 2. Contributions of the Researchers In 1965, Cooley and Tukey [19] published their findings in the first paper on computation of Discrete Fourier Transform (DFT) that is later known to be the FFT. The method, shown to be far superior (in a sense that requires much less computation than the direct computation of multiplication of the DFT matrix and the data vector), depends primarily on two factors. (a) the sequence length must be highly composite, i.e., factorable into many factors of small length and a particular way unidimensional data is mapped into twodimensional data (which is again repeated for each of these dimensions till the smallest factor), that is known as the famous divide and conquer approach, that led to a gross reduction in computational effort (that depends again on the degree of compositeness of the sequence length N). N (b) The basic requirement of W = 1 where j2π / N W = e i.e., W is the N-th root of unity. Fairly large research activity followed in the wake of this celebrated and valuable publication. A plethora of activities on investigation in various aspects of computation of DFT and convolution was started by the researchers interested in this area. Good [28] in 1971 had shown the difference between the data mapping proposed based on CRT when the factors of N are mutually prime and those based on Cooley-Tukey approach where factors need not be mutually prime (rather, it is better to have small factors like, say two only with N as large powers of two). In the case of the former, while data tracking is more complex, there will not be any requirement of multiplication by inter-stage twiddle factor as in the case of latter case. This former case is generally referred to as prime factor algorithm (PFA) [31] of computing the DFT. Around the same time Pollard [45] discusses the fast Fourier transform in a finite field, where he also points out about the equivalence of convolution for multiplication of large integers and utility of transform method. 2.1 Mersenne Number Transform In 1972, Rader (one of the leading pioneers in the initial development of digital signal processing and the coauthor of the first book on digital signal processing [27]) proposed a scheme modulo Mersenne number primes defining Mersenne number transform (MNT) [46]. The p-th Mersenne numbers M p s are defined to be the prime integers of type 2 p 1, p also being a prime.we mention that so far there are only forty-one Mersenne primes found so far (the first few values of p are 2, 3, 5, 7, 13, 17, 19, 31, 61, 89, 107, 127, 521, 607, 1279, ). The 41 st prime is discovered only in this year

3 and this is the largest prime found so far, at this point of writing (a tail-piece about the 41 st Mersenne number along with its some interesting features is being included at the end of the paper). As, then it follows that M p =2 p 1 (4) 2 p =1 mod M p (5) meaning that 2 is of order p in the finite ring of integers under operations of additions and multiplication modulo M p. Following our discussion in the earlier section, we conclude that 2 could be an α to compute NTT (and subsequently convolution) for a sequence length p. Addition modulo M p is simply one s complement addition in a p-bit word. Multiplication modulo M p is achieved by forming the 2p-bit product of two words and adding the least significant p-bits with the most significant p-bits, in one s complement fashion. Multiplication by powers of 2 modulo M p is done by rotation of the bits within the p-bit word; all the computation is done by additions and bit shifts Visualization of Constraints and problems in NTT In the earlier paragraph, while it appears to be quite attractive, realization of the deficiencies becomes selfevident. We illustrate them as follows: (a) With α as 2 the maximum length for convolution is p, a prime integer, while NTT follows the architecture of FFT that is efficient only when the length N is highly composite; with α as 2 the length can be doubled to 2p, but then even 2p is barely composite (the reasoning that 2 is a root of order 2p mod M p follows from the fact that (2, p)=1, meaning 2 and p are mutually prime, 4 i.e., 2 2 is also root of order p, so 2 = 4 1/2 will be a root of order 2p [1]. (b) Maximum length N max possible is M p 1=2 p 2 that will require α be a primitive root of order N max. This value of α will be represented by few bits unlike 2 and so will be its higher powers, leading to far more complexity in modular arithmetic especially in the case of multiplications. The advantage we gain is that N max or its submultiple length available may be composite (but may not be highly, as is desirable in the computationally efficient FFT structure). As an example we consider M 31 = = (that can be represented by 31-bits wordlength i.e., all ones in a 31-bit register noting powers of two runs from zero upwards), with α as 2 or 2, one can obtain a convolution length of 31 or 62 respectively. At the same time, N max = has factors (as 2, 3, 3, 7, 11, 31, 151, 331) that can be termed as barely composite, to gain any benefit from FFT structure; further α (for N max ) and its higher powers will not be simple like 2 or powers of 2 for simple multiplication by word shift or rotation operations. At this stage we see that there exist undesirable relations between choice of modulus M, i.e., the wordlength, N the convolution length, value of α the root of order N, and the computational complexity. Rader [46] suggested the use of Fermat numbers as the choice of modulus to alleviate some of the problems in use of MNT. 2.2 Fermat Number Transform Agarwal and Burrus [1 3] thoroughly investigated the use of Fermat numbers as the choice of modulus in the NTT that came to be known as Fermat number transfer (FNT). Fermat numbers denoted by F t where t-th Fermat number is defined as 2 2t +1, t an integer = 0, 1, 2,...., (it may be noted that most of the the Fermat numbers are not primes; only the first five numbers F 0 thru F 4 are primes, others so far searched are known to be composites (see page 554 in [1] )). As, F t = 2 2t +1 (6) it follows that for primes, i.e., t 4, the order of 2 is 2 t+1 as, leading to 2 2t = 1 mod F t (7) 2 (2t ).2 = 2 2t+1 = 1 mod F t. (8) Hence, with α as 2, N max would be 2 t+1. The concept of 2 was introduced to double the length to 2 t+2 with α s that required only two bit representation. Equating b, the number of bits, with 2 t, 2 may be termed as α 2b and 2 as α 4b. Actually the composite Fermat numbers F 5 (b=32) and F 6 (b=64) are attractive for the applications in signal processing. It was pointed out that while these numbers are composite and their factors are of the type K2 t+2 + 1, they can still be utilized for transforming with α 2b and α 4b with and α 4b = 2 b/4 (2 b/2 1) (9) α 4b 2 = 2 mod F t. (10) (Please refer to the relevant discussions in [1, 2]). It can be seen (from the Table 1 in [1]) that the undesirable and

4 restrictive relations between N, M, α, and computational complexity still continue in FNT as well. As suggested by Rader in [46], Agarwal and Burrus [3] proposed a two-dimensinal scheme for convolution of length N = LP being implemented as a two-dimensional cyclic convolution of length 2L by P that can be computed by using a two-dimensional FNT. Using this twodimensional scheme, the wordlength is proportional to the square root of the length of the sequences to be convolved which would give for a maximum length of 8b 2, rather than 4b (refer Table II in [1]). Because of modular arithmetic, in the ring of integers mod M, unambiguous representation of integers is possible only if their absolute value is less than M/2. So, either the wordlength M will have to be chosen accordingly, or the input sequences will have to be suitably scaled. This is akin to the overflow constraint in fixed-point digital filtering. Some schemes are proposed in [1] as alternatives; arithmetic operations in FNT are described in [2]. They had also suggested and described use of multiple moduli to increase the dynamic range and complex NTT (CNTT) that was mainly theorized and initiated by Reed and Truong [48 51]. McClellan [35] demonstrated hardware for FNT. He used a number representation scheme based on a general class of representation evolved by Leibowitz [32] that is known to be Diminished 1 number representation. This leads to modulo 2 b +1 arithmetic to be implemented in a manner similar to one s complement arithmetic i.e., modulo 2 b 1 arithmetic. 2.4 Some Related Contemporary Developments In this subsection, we briefly describe a few contemporary developments, while not being number theoretic exactly in a sense that we have discussed till this point, will be generally related to some of the developments those we will describe in the latter subsections. Firstly, in 1968 Rader [47] made a very interesting observation that when number of data samples is a prime, DFT can be expressed as convolution plus addition of a few simple terms. Illustrations would be found in [31, 55, 60] where one can see that by appropriate permutation of indices, the DFT matrix (excluding the first row and the first column) is a convolution matrix. Secondly, Agarwal and Cooley [4] developed direct algorithms for convolutions for short lengths of 2, 3, 5, 6, 7, 8 and 9 (hereafter will be referred to as the AC algorithm). When the factors of convolution length are mutually primes, these short length algorithms can be made use of. Using CRT type of mapping the unidimensional data are formulated as multidimensional data. Nested algorithm is made use to combine the individual length algorithms and remap the final result similar to input mapping. The individual short length algorithms may be termed as rectangular transform. The speciality is that elements of the rectangular matrices are quite simple that leads to the matrix-vector multiplication by additions (and/or subtractions). The disadvantage is that the length of the data vector expands in the intermediate stages and then contracts back to its original length that leads to increase in complexity in data tracking unlike in FFT structure where the length of the data vector remains the same through all the intermediate stages. It was seen that for convolution lengths up to 210 this algorithm requires the least amount of computational effort compared to other methods. Thirdly, based on [47] Winograd [55, 60] developed small length DFT algorithm of lengths 2, 3, 4, 5, 7, 8, 9, and 16 in a manner that is somewhat akin to the AC algorithm. This algorithm is referred to as Winograd Fourier transform algorithm (WFTA). Composite lengths comprising of mutually prime factors is employed using a similar nested algorithm. Lucid illustrations and relevant explanations are found in [55, 60]. As in the case of the AC algorithm, the length of the data vector expands in the intermediate stages and then contracts back to its original length that leads to increase in complexity in data tracking unlike in typical Cooley- Tukey type FFT structure where the length of the data vector remains the same through all the intermediate stages. Baily [6] developed a number theoretic version of the WFTA; independently of this development Bhattacharya and Agarwal [7] also developed the number theoretic version that they termed as Winograd Number Theoretic transform algorithm (WNTA). Kolba and Parks [31] demonstrated that WFTA can be used in prime factor algorithm (PFA), unlike in nested algorithm, with an advantage of reduced additions per point at the cost of some marginal increase of multiplications. We mention that using nested algorithm in both AC and WFTA the number of additions increases at a much rapid rate with increase in length Complex Number Theoretic Transform (CNT) Reed and Truong [48 50] initiated the concept of complex number theoretic transform (CNT) from the theoretical view point of Galois field theory (We mention that I. S. Reed is the co-inventor of the famous Reed- Solomon codes). They defined the transform in the Galois field of q 2 elements GF(q 2 ), a finite field analogous to the field of complex numbers, when q is a prime such that ( 1) is not a quadratic residue, i.e., x 2 = 1 mod q (11) is not solvable (meaning there is no integer x that satisfies above relation). It can be shown that it holds good for M p s the Mersenne primes, i.e., q could be a Mersenne

5 prime where the modulo arithmetic is quite simple as mentioned earlier. They also outlined a procedure for finding a primitive element of order 2 p+1, of the type a + ib, a, b, integers, (and i is analogous to 1 of complex number field) in G 2 p+1 of GF(q 2 ). It is seen that that this transform has certain disadvantage due to the fact that multiplication by powers of primitive element is not simple like bit shifts and/or word rotation, because the element is like an integer comprising of few bits (to many bits). While lengths obtainable are quite high, the modulo operation is simple due to use of Mersenne primes. In [49] they had shown that by using CRT the transforms can be computed (i.e., leading to complex integer convolution like earlier) over direct sum of Galois fields GF(q 2 ) in order to obtain larger dynamic range. It is akin to our earlier statement of using multiple moduli, along with CRT method of combining, to obtain increased dynamic range. Further extending their work [50], for utilizing the Fermat numbers, they had defined transforms over a ring of quadratic integers, modulo a prime number q in the quadratic field R( m), where m is a square free integer. When q is a Fermat prime, one can use FFT algorithm over the resulting finite fields to yield fast convolutions of quadratic integer sequences in R( m). This was extended to to direct sum of such finite fields; as a result, such transforms can be utilized with nonprime Fermat numbers F 5 and F Pseudo- Number Theoretic Transforms Noting the scarcity of primes of the type of primes of the type 2 b ± 1 along with simple α like powers of 2, so that we have wide choices of wordlength along with wide choices of convolution lengths Nussbaumer [37, 38, 39] suggested using large prime factors of the composite numbers such as M =M b /x, where M b = (2 b ± 1) and x is small part of the factors that may or may not be composite. As a result there will be some loss in the effective wordlength compared to b the working wordlength corresponding to M. The complete operation is done modulo M b that is equal to 2 b ± 1 to have simple modulo arithmetic except the the final output is taken after reduction modulo M =M b /x that gives the correct output. These types of M b s are known as Pseudo-Fermat or Pseudo-Mersenne numbers depending upon whether M b = 2 b +1 or 2 b 1, respectively. Accordingly, the transform is known as Pseudo-FNT or Pseudo-MNT. He also introduced the use of roots i.e., like 2(j 1) and (j+1) (like the concept of 2 to increase the length permissible with the exception that while 2 can be represented in general by two bits or that element whose square is 2 modulo M, the aforesaid roots are actually complex and complex Pseudo NTT s are performed; readers may note that the operation as such, and the computation by such roots and its higher power are quite simple and no multiplication is involved at all). The logic of such roots is illustrated by the example in next paragraph. In the case of M = ( ) / (3. 11), with α as 2 one can compute a transform a length of 50, then 2 4 will be of order 25. Now as (j +1) 8 = 2 4 and (8, 25) = 1, i.e., mutually prime, the order of (j +1) will be 200 (eight times twenty-five). Here the effective wordlength is reduced to 20. Examples of such moduli will be found in [7, 37 39]. 2.7 Hybrid Transform methods It is normally the case, that there exists many methods for doing the same thing, with each method having its own advantages and disadvantages. Hence it may be possible that one may be able to combine advantages of some of them and devise methods to gain some advantage. We briefly outline some of the schemes. Agarwal and Cooley [3] had shown that FNT can be combined with the AC algorithm i.e., the short length convolution algorithms. The length of the first dimension is taken as 128 (using F 5 ) and the lengths of the other dimension are taken as mutually prime odd numbers 3, 5, 7, and 9. From the results it is amply clear that the length is increased with a very small computational effort. Reed and Truong [51] had similarly shown that the short length AC convolution algorithms can be combined with the CNT s (refer to subsection 2.5) they had developed in GF(q 2 ) using the Mersenne primes as q. In [7], Bhattacharya and Agarwal demonstrated quite in detail the efficacy of using the hybrid type of techniques by decomposing the convolution length N in three dimensions (mutually prime factors) and employing different methods in each dimension like WNTA in nested mode or PFA mode (the number theoretic variant of WFTA), the short length AC algorithm, and the NTT of the type (that we had described here these requires no multiplication). The requirement of computational effort had confirmed the utility of such schemes in computing convolution. 2.8 New Mersenne Number Transform (NMNT) This type of transform was developed by Boussakta and Holt [12 15] and is considered an intelligent and excellent derivative where by the CNT approach as mentioned in subsection 2.5 is suitably converted to a real transform. The transform is defined as

6 Here X ( k) = N 1 n= 0 x( n)( β ( nk) + β ( nk)) mod (12) M p is a Mersenne prime and 1 2 M p β 1( nk) = Re (( α1 + jα 2) ) and (13) nk β nk) = Im (( α + jα ) ) 2( 1 2 Also, α 1 = 2 q, α 2 = ( 3) q, q = 2 p 2 mod nk M p, and Re(.) and Im(.) denotes the real and imaginary parts. In (13) α 1 and α 2 are of order N = 2 p + 1 as given in [48] and mentioned in the subsection on CNT. For lengths N/d, β 1( nk) = Re ((( α1 + jα2) ) ) and (14) d 2( 1 2 β nk) = Im ((( α + jα ) The inverse transform is the same as in (12) except for the multiplying factor of 1/N. The transform length is an integer powers of two and can be up to 2 p. It is mentioned that arithmetic operations and residue reduction modulo Mersenne primes are simpler, all practical lengths of powers of two will be available, and the multiplications are normal type as (β 1 (.) + β 2 (.)) will be an integer needing few bits representation not like powers of 2. They have also proved that this transform has the necessary circular convolution property (CCP); also, they have shown that this can be combined with FNT to increase the dynamic range as like earlier cases of multiple moduli. 2.9 Miscellaneous Effort and Schemes In this subsection, we outline very briefly the efforts and schemes made by some researchers. Creutzburg [20 23] along with his co-researchers investigated on NTT of prescribed length, pseudo- NTT s, and CNT s using cyclotomic integers and polynomials in order to determine the relevant parameters, i.e. N, M, and α. in a general sense. Lu and Lee [34] made an effort to solve the sequence length constraint problem. They proposed generalized modulo primes (GMP) of the type given by q t ± d ) nk nk M = p a (15) such that M is a prime for some p, q, t, and a. One can see that for some values we can get a Fermat number and for some other set we can obtain Mersenne numbers. Parker and Benaissa [43] proposed application of Rader s algorithm [47] that states that when number of ) samples is a prime, DFT of the sequence can be represented by convolution that again can be computed by NTT. This holds good for NTT also. They presented a scheme by which one can obtain large lengths by recursively applying a small length NTT algorithm module. Dimitrov, Cooklev, and Donevsky [24, 25] presented the scheme for generalized Fermat-Mersenne number theoretic transform; they also defined NTT over the golden section quadratic field. Lastly, we mention that Golomb, Reed, and Truong [26] investigated the prime numbers of the type 3*2 n +1, where there are many values of n. Under such choice of modulus N max is equal to 3*2 n that is highly composite. They had developed certain arithmetic operations scheme modulo such M s, and the procedure to find a root of given order. Bhattacharya and Astola [9] investigated the set of primes K*2 n +1, a much larger sets of primes [52] for their utilization. They noted that modulo such numbers one can see that the sequence length constraint vis-à-vis the word length almost vanishes and proposed certain schemes for their utilization. Further, noting, that N max contains many factors those are squares, they developed some scheme for implementing Bluestein [11] algorithm for linear filtering. 3. Conclusions It is opined by the authors that interest in NTT may have remained and would not have waned as one can notice this days (if not in the lead vis-à-vis FFT), but for the facts that when the NTT schemes were being developed, hardware i.e., bits and bytes (those relates to word length of data sequences) were quite costly those days compared to nowadays. Unlike Fourier spectra, number theoretic spectra of a sequence carries no meaning indicating that application of NTT would have remained confined generally to convolution. However, it is mentioned in the paper that algorithm associated with a basic principle exists [11, 47] wherein convolution can be used to compute FFT and that part of convolution could be computed by NTT as well. We would reiterate the advantages of being free of any round off error due to arithmetic operations, as the computations are done by modulo an integer; and data paths would remain real unlike those in FFT where each data path is actually comprised of two data paths i.e., real and complex parts. The major disadvantage is the stringent relation between the choice of modulus M and the sequence length N, especially in the cases where one wishes to employ simple arithmetic like bit shifts and adds in place of multiplications and chooses a simple modulus of type 2 n ±1. It may be noted that modulo arithmetic operations modulo a general integer is more complex.

7 However, one would also point out that for real sequences DFT is symmetric; there is a well established way that one can form a complex sequence of length N from two consecutive real sequence of data and then compute one N length FFT. Subsequently, it is possible to separate the DFT s of the sequence by a few adds and subtracts per point. As number theoretic spectra has no meaning, this advantage is not available for NTT; for comparison purpose one may have to compare 2N length sequence NTT with N length FFT. However for convolution of complex sequences such disadvantage will not apply for using NTT especially with roots as 2(j 1) and (j + 1) using pseudo-ntt s. This paper was also intended for reviving some interest in this topic for further investigation against the backdrop of present day signal processing research. It is with this intention that there are many references included in the reference list those are not referred to in the main body of the paper. A fully exhaustive and all inclusive investigation and comparison at present days hardware and software techniques along with cost for practical application is warranted, to appreciate the full benefits. 4. Some References. [1] R. C. Agarwal and C. S. Burrus, "Number Theoretic Transforms to Implement Fast Digital Convolution," Proc. IEEE, vol. 63, pp , Apr [2]..", "Fast Convolution using Fermat number transforms with applications to digital filtering," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-22, pp , Apr [3], Fast one-dimensional digital convolution by multi-dimensional techniques, IEEE Trans. Acoust., Speech,and Signal Processing, vol. ASSP- 22, pp. 1-10, Feb [4] R. C. Agarwal and J. W. Cooley, "New Algorithms for Digital Convolution," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-25, pp , Oct [5].., New algorithms for digital convolution, Proc. IEEE ICASSP, May 1977, pp [6] D. Bailey, "Winograd s algorithm applied to number theoretic transform". Electron. Lett. vol. 13, pp , Sep [7] M. Bhattacharya and R. C. Agarwal, "Number Theoretic Techniques for Computation of Digital Convolution," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP- 32, pp , Jun [8]., Comments on A fast computation of complex convolution using a hybrid transform, IEEE Trans. Acoust., Speech, Signal Processing, vol. ASSP-29, pp , Oct [9] M. Bhattacharya and J. Astola, Recursive structure for linear filtering using number theoretic transform, Proc. 8 th IEEE Intl. Conf. Electronics, Circuits, Systs., ICECS 2001, vol. 1, pp , Sep [10] R.E. Blahut, Fast Algorithms for Digital Signal Processing. Addison-Wesley,1987. [11] L.I. Bluestein, "A linear Filtering Approach to the Computation of Discrete Fourier Transform". IEEE Trans. Audio. Electroacoust. vol. AU-18, pp , Dec [12] S. Boussakta and A. G. J. Holt, New Number Theoretic Transform. Signal Processing, vol. 28, pp , Aug [13].."., New transform using the Mersenne numbers, IEE Proc. Vis. Image Signal Process., vol. 142, pp , Dec [14]., A novel combination of NTTs using the MRC, Signal Processing, vol. 54(1), pp , [15].."., New Two Dimensional Transform, Electron. Lett. Vol. 29, pp , May [16] C. S. Burrus and I. W. Selesnick, "On programs for prime length FFTs and Circular Convolution," IEEE Proc. ICASSP-95, vol. 2, pp , [17] C. S. Burrus, "Index Mapping for Multidimensional formulation of the DFT and the Convolution," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-25, pp , Jun [18] " "Notes on the FFT". Rice University, Houston, TX 77005, 29 Sep [19] J. W. Cooley and J. W. Tukey, An algorithm for the machine computation of complex Fourier series, Math. Comp., vol. 19, pp , Apr [20] R. Creutzberg, Constructive Parameterization of Number- Theoretic Transforms A Short Overview, T1CSP Workshop on Transforms and Filter Banks, [21] R. Creutzberg and M. Tasche, Number-theoretic transforms of prescribed length, Math. Comp., vol. 47, pp , [22].., parameter determination for complex number.theoretic transforms using cyclotomic polynomials, Math Comp., vol. 52, pp , [23] R. Creutzberg and G. Steidl, Construction of parametersfor number-theoretic transforms in rings of cyclotomic integers, ournal Inform. Process.,Cybernetics, vol. EIK- 24, pp , [24] V. S. Dimitrov, T. V. Cooklev, and B. D. Donevsky, Generalized Fermat-Mersenne number theoretic transform, IEEE Trans. CAS-II, vol.41, pp , Feb [25].., Number Theoretic Transforms Over the Golden Section Quadratic Field, IEEE Trans. Signal Processing, vol. SP-43, pp , Aug [26] S. W. Golomb, I. S. Reed, and T. K.Truong, "Integer Convolution over Finite Field GF(3.2 n +1),". SIAM. J. Appl. Math., vol. 32, pp , Mar [27] B. Gold and C. M. Rader, Digital Processing of Signals. New York : McGraw-Hill, [28] I. J. Good, The relation between the two fast Fourier transforms, IEEE Trans. Comput., vol. C-20, pp , Mar [29] G. H. Hardy and E. M. Wright, The Theory of Numbers. Oxford, England: Oxford Univ. Press, [30] D. Kibler, R. C. Agarwal, C. S. Burrus, Necessary and sufficient conditions for the existence of the modular Fourier transform: Comments on Number theoretic

8 transforms to implement fast digital convolution [and reply], Proc. IEEE, vol. 65, pages , Feb [31] D. P. Kolba and T. W. Parks, "A Prime Factor FFT Algorithm using High Speed Convolution," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-25, pp , Aug [32] L. M. Leibowitz, A binary arithmetic for the Fermat number transform, Naval Research Lab, Report 7971, [33].., A Simplified Binary Arithmetic for the Fermat Number Transform, IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-24, pp , Oct [34] H. Lu and S. C. Lee, "A new approach to Solve the Sequence-Length Constraint Problem in Circular Convolution Using Number Theoretic Transform," IEEE Trans. Signal Processing, vol. 39, pp , Jun [35] J. H. McClellan, Hardware Realization of a Fermat Number Transform, IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-24, pp , Jun [36] J. H. McClellan and C.M. Rader, Number Theory in Digital Signal Processing. Prentice-Hall Inc. Englewood Cliffs, NJ., [37] H.J. Nussbaumer "Complex convolutions via Fermat Number Transforms," IBM Journal Research and Development, pp , May [38]. ", "Digital Filtering using complex Mersenne Transform," IBM Journal Research and Development, pp , Sep [39]. " "Digital Filtering using Pseudo Fermat Number Transforms," IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-25, pp , Oct [40]. ", "Linear Filtering Technique for Computing Mersenne and Fermat number Transform," IBM Journal Research and Development, pp , Jul [41] H. J. Nussbaumer, Fast Fourier Transform and Convolution Algorithm. Springer-Verlag, 1982 [42] O. Ore, Number Theory and Its History. New York: McGraw-Hill, [43] M. G. Parker and M. Benaissa, Unusual-length numbertheoretic transforms using recursive extensions of Rader s algorithm, IEE Proc. Vis. Image Signal Process., vol. 142, pages 31-34, Feb [44] A. Peled and B. Liu, A new Hardware realization of Digital Filters, IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-22, pp , Dec [45] J. M. Pollard, The fast Fourier transform in a finite field, Math. Comp., vol. 25, pp , Apr [46] C. M. Rader, Discrete convolutions via Mersenne Transform. IEEE Trans. Comput., vol. C-21, pp , Dec [47]..., Discrete Fourier transform when the number of data samples is prime, Proc. IEEE, vol. 56, pp , Jun [48] I. S. Reed and T.K. Truong, The Use of Finite Fields to Compute Convolutions, IEEE Trans. Inform. Theory, vol IT-21, pp , Mar [49]...., Complex Integer Convolution over a Direct Sum of Galois Fields, IEEE Trans. Inform. Theory, vol IT-21, pp , Nov [50].., Convolutions over residue Classes of Quadratic Integers, IEEE Trans. Inform. Theory, vol. IT-22, pp , Jul [51]....., A Fast Computation of complex convolution using a Hybrid Transform, IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP- 26, pp , Dec [52] R.M. Robinson, A report on Primes of the form k.2 n +1 and on factors of Fermat Numbers, Proc. Amer. Math. Soc. 9(1958), pp , Oct [53] I. W. Selesnick and C. S. Burrus, Extending Winograd s small convolution algorithm to longer lengths. IEEE Proc. ISCAS-94, vol. 2, pp , [54]., Automatic Generation of Prime Length FFT Programs, IEEE Trans. Signal Processing, vol. 44, pp , Jan [55] H. F. Silverman An Introduction to Programming the Winograd Fourier Transform Algorithm (WFTA), IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP- 25, pp , Apr [56] N. S. Szabo and R. I. Tanaka, Residue Arithmetic and Its Applications to Computer Technology. New York: Mcgraw- Hill, [57] H. Tamori, N. Aoiki, and T. Yamamoto, A Fragile Watermarking Technique by Number Theoretic Transform, IEICE Trans. Fundamentals, vol. E85A, pp , Aug [58] E. Vegh and L. M. Leibowitz, Fast Complex Convolution in Finite Rings, IEEE Trans. Acoust., Speech, and Signal Processing, vol. ASSP-24, pp , Aug [59]...., Discrete convolution of complex integer sequences, Proc. ICASSP 76, vol. 1, pp , Apr [60] S. Winograd, On Computing the Discrete Fourier Transform, Math. Comp., vol. 32, pp , Jan 1978 (also in IBM Research report RC-6291 of 1976, IBM Thomas J. Watson Research Center, Yorktown Heights, New York). [61]., Some bilinear forms whose multiplicative complexity depends on the field of constants, Math, Syst, Theory, 10(2), pp , *********** Some Interesting Notes:- (a) Pierre Fermat ( ) was a lawyer and government official most remembered for his work in number theory, in particular for Fermat s last theorem. (b) Marin Mersenne ( ) was a monk and is best known as clearinghouse for correspondence between eminent philosophers and scientists and for his work in number theory. (c) The 41 st Mersenne prime was found in May 2004 by Josh Findley. He used a 2.4 GHz Pentium 4 Windows XP computer running for 14 days to prove the number was prime. It has 7,235,733 decimal digits represented by Written out, it would stretch for 25 kilometres (only!!). The 7,235,733 digits, would take someone the best part of six weeks to write out longhand. The authors have the digits in a text file (Any volunteers to try it out are welcome to approach the authors or download them from GIMPS (Great Internet Mersenne Prime Search) website).

On the computation of discrete Fourier transform using Fermat number transform

On the computation of discrete Fourier transform using Fermat number transform On the computation of discrete Fourier transform using Fermat number transform Wan-Chi Siu, AP(HK), M.Phil., C.Eng., M.I.E.R.E., Mem.I.E.E.E., A.G. Constantinides, B.Sc.(Eng.), Ph.D., C.Eng., M.I.E.E.,

More information

The DFT as Convolution or Filtering

The DFT as Convolution or Filtering Connexions module: m16328 1 The DFT as Convolution or Filtering C. Sidney Burrus This work is produced by The Connexions Project and licensed under the Creative Commons Attribution License A major application

More information

Frequency Domain Finite Field Arithmetic for Elliptic Curve Cryptography

Frequency Domain Finite Field Arithmetic for Elliptic Curve Cryptography Frequency Domain Finite Field Arithmetic for Elliptic Curve Cryptography Selçuk Baktır, Berk Sunar {selcuk,sunar}@wpi.edu Department of Electrical & Computer Engineering Worcester Polytechnic Institute

More information

Radix-4 Factorizations for the FFT with Ordered Input and Output

Radix-4 Factorizations for the FFT with Ordered Input and Output Radix-4 Factorizations for the FFT with Ordered Input and Output Vikrant 1, Ritesh Vyas 2, Sandeep Goyat 3, Jitender Kumar 4, Sandeep Kaushal 5 YMCA University of Science & Technology, Faridabad (Haryana),

More information

Convolution Algorithms

Convolution Algorithms Connexions module: m16339 1 Convolution Algorithms C Sidney Burrus This work is produced by The Connexions Project and licensed under the Creative Commons Attribution License 1 Fast Convolution by the

More information

Generalized Fermat-Mersenne Number Theoretic Transform Vassil S. Dimitrov, Todor V. Cooklev, and Borislav D. Donevsky

Generalized Fermat-Mersenne Number Theoretic Transform Vassil S. Dimitrov, Todor V. Cooklev, and Borislav D. Donevsky IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS-11: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 41, NO., FEBRUARY 1994 133 Generalized Fermat-Mersenne Number Theoretic Transform Vassil S. Dimitrov, Todor V. Cooklev,

More information

Number-Theoretic Transforms of Prescribed Length

Number-Theoretic Transforms of Prescribed Length MATHEMATICS OF COMPUTATION VOLUME 47, NUMBER 176 OCTOBER 1986. PAGES 693-701 Number-Theoretic Transforms of Prescribed Length By R. Creutzburg and M. Tasche Abstract. A new constructive method for finding

More information

Low Power, High Speed Parallel Architecture For Cyclic Convolution Based On Fermat Number Transform (FNT)

Low Power, High Speed Parallel Architecture For Cyclic Convolution Based On Fermat Number Transform (FNT) RESEARCH ARTICLE OPEN ACCESS Low Power, High Speed Parallel Architecture For Cyclic Convolution Based On Fermat Number Transform (FNT) T.Jyothsna 1 M.Tech, M.Pradeep 2 M.Tech 1 E.C.E department, shri Vishnu

More information

Large Integer Multiplication on Hypercubes. Barry S. Fagin Thayer School of Engineering Dartmouth College Hanover, NH

Large Integer Multiplication on Hypercubes. Barry S. Fagin Thayer School of Engineering Dartmouth College Hanover, NH Large Integer Multiplication on Hypercubes Barry S. Fagin Thayer School of Engineering Dartmouth College Hanover, NH 03755 barry.fagin@dartmouth.edu Large Integer Multiplication 1 B. Fagin ABSTRACT Previous

More information

Chapter 9 Mathematics of Cryptography Part III: Primes and Related Congruence Equations

Chapter 9 Mathematics of Cryptography Part III: Primes and Related Congruence Equations Chapter 9 Mathematics of Cryptography Part III: Primes and Related Congruence Equations Copyright The McGraw-Hill Companies, Inc. Permission required for reproduction or display. 9.1 Chapter 9 Objectives

More information

An Effective New CRT Based Reverse Converter for a Novel Moduli Set { 2 2n+1 1, 2 2n+1, 2 2n 1 }

An Effective New CRT Based Reverse Converter for a Novel Moduli Set { 2 2n+1 1, 2 2n+1, 2 2n 1 } An Effective New CRT Based Reverse Converter for a Novel Moduli Set +1 1, +1, 1 } Edem Kwedzo Bankas, Kazeem Alagbe Gbolagade Department of Computer Science, Faculty of Mathematical Sciences, University

More information

THE discrete sine transform (DST) and the discrete cosine

THE discrete sine transform (DST) and the discrete cosine IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS-II: EXPRESS BIREFS 1 New Systolic Algorithm and Array Architecture for Prime-Length Discrete Sine Transform Pramod K. Meher Senior Member, IEEE and M. N. S. Swamy

More information

Subquadratic Computational Complexity Schemes for Extended Binary Field Multiplication Using Optimal Normal Bases

Subquadratic Computational Complexity Schemes for Extended Binary Field Multiplication Using Optimal Normal Bases 1 Subquadratic Computational Complexity Schemes for Extended Binary Field Multiplication Using Optimal Normal Bases H. Fan and M. A. Hasan March 31, 2007 Abstract Based on a recently proposed Toeplitz

More information

Multiplicative Complexity, Convolution, and the DFT

Multiplicative Complexity, Convolution, and the DFT Michael T. Heideman Multiplicative Complexity, Convolution, and the DFT C.S. Bunus, Consulting Editor Springer-Verlag New York Berlin Heidelberg London Paris Tokyo Michael T. Heideman Etak, Incorporated

More information

Trace Representation of Legendre Sequences

Trace Representation of Legendre Sequences C Designs, Codes and Cryptography, 24, 343 348, 2001 2001 Kluwer Academic Publishers. Manufactured in The Netherlands. Trace Representation of Legendre Sequences JEONG-HEON KIM School of Electrical and

More information

A High-Speed Realization of Chinese Remainder Theorem

A High-Speed Realization of Chinese Remainder Theorem Proceedings of the 2007 WSEAS Int. Conference on Circuits, Systems, Signal and Telecommunications, Gold Coast, Australia, January 17-19, 2007 97 A High-Speed Realization of Chinese Remainder Theorem Shuangching

More information

Design of Low Power, High Speed Parallel Architecture of Cyclic Convolution Based on Fermat Number Transform (FNT)

Design of Low Power, High Speed Parallel Architecture of Cyclic Convolution Based on Fermat Number Transform (FNT) Advance in Electronic and Electric Engineering. ISSN 2231-1297, Volume 3, Number 6 (2013), pp. 641-650 Research India Publications http://www.ripublication.com/aeee.htm Design of Low Power, High Speed

More information

Volume 3, No. 1, January 2012 Journal of Global Research in Computer Science RESEARCH PAPER Available Online at

Volume 3, No. 1, January 2012 Journal of Global Research in Computer Science RESEARCH PAPER Available Online at Volume 3, No 1, January 2012 Journal of Global Research in Computer Science RESEARCH PAPER Available Online at wwwjgrcsinfo A NOVEL HIGH DYNAMIC RANGE 5-MODULUS SET WHIT EFFICIENT REVERSE CONVERTER AND

More information

Infinite Sequences, Series Convergence and the Discrete Time Fourier Transform over Finite Fields

Infinite Sequences, Series Convergence and the Discrete Time Fourier Transform over Finite Fields Infinite Sequences, Series Convergence and the Discrete Time Fourier Transform over Finite Fields R M Campello de Souza M M Campello de Souza H M de Oliveira M M Vasconcelos Depto de Eletrônica e Sistemas,

More information

GENERALIZED ARYABHATA REMAINDER THEOREM

GENERALIZED ARYABHATA REMAINDER THEOREM International Journal of Innovative Computing, Information and Control ICIC International c 2010 ISSN 1349-4198 Volume 6, Number 4, April 2010 pp. 1865 1871 GENERALIZED ARYABHATA REMAINDER THEOREM Chin-Chen

More information

RADIX-2 FAST HARTLEY TRANSFORM REVISITED. de Oliveira, H.M.; Viviane L. Sousa, Silva, Helfarne A.N. and Campello de Souza, R.M.

RADIX-2 FAST HARTLEY TRANSFORM REVISITED. de Oliveira, H.M.; Viviane L. Sousa, Silva, Helfarne A.N. and Campello de Souza, R.M. RADIX- FAST HARTLY TRASFRM RVISITD de liveira, H.M.; Viviane L. Sousa, Silva, Helfarne A.. and Campello de Souza, R.M. Federal University of Pernambuco Departamento de letrônica e Sistemas - Recife - P,

More information

AREA EFFICIENT MODULAR ADDER/SUBTRACTOR FOR RESIDUE MODULI

AREA EFFICIENT MODULAR ADDER/SUBTRACTOR FOR RESIDUE MODULI AREA EFFICIENT MODULAR ADDER/SUBTRACTOR FOR RESIDUE MODULI G.CHANDANA 1 (M.TECH),chandana.g89@gmail.com P.RAJINI 2 (M.TECH),paddam.rajani@gmail.com Abstract Efficient modular adders and subtractors for

More information

A Generalized Reverse Jacket Transform

A Generalized Reverse Jacket Transform 684 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 48, NO. 7, JULY 2001 A Generalized Reverse Jacket Transform Moon Ho Lee, Senior Member, IEEE, B. Sundar Rajan,

More information

A Suggestion for a Fast Residue Multiplier for a Family of Moduli of the Form (2 n (2 p ± 1))

A Suggestion for a Fast Residue Multiplier for a Family of Moduli of the Form (2 n (2 p ± 1)) The Computer Journal, 47(1), The British Computer Society; all rights reserved A Suggestion for a Fast Residue Multiplier for a Family of Moduli of the Form ( n ( p ± 1)) Ahmad A. Hiasat Electronics Engineering

More information

Summary of the FFT *

Summary of the FFT * OpenStax-CNX module: m42270 1 Summary of the FFT * C. Sidney Burrus This work is produced by OpenStax-CNX and licensed under the Creative Commons Attribution License 3.0 Abstract Summary of research at

More information

Time Series Analysis

Time Series Analysis Time Series Analysis A time series is a sequence of observations made: 1) over a continuous time interval, 2) of successive measurements across that interval, 3) using equal spacing between consecutive

More information

Optimization of new Chinese Remainder theorems using special moduli sets

Optimization of new Chinese Remainder theorems using special moduli sets Louisiana State University LSU Digital Commons LSU Master's Theses Graduate School 2010 Optimization of new Chinese Remainder theorems using special moduli sets Narendran Narayanaswamy Louisiana State

More information

Fast, Parallel Algorithm for Multiplying Polynomials with Integer Coefficients

Fast, Parallel Algorithm for Multiplying Polynomials with Integer Coefficients , July 4-6, 01, London, UK Fast, Parallel Algorithm for Multiplying Polynomials with Integer Coefficients Andrzej Chmielowiec Abstract This paper aims to develop and analyze an effective parallel algorithm

More information

Fast evaluation of iterated multiplication of very large polynomials: An application to chinese remainder theory

Fast evaluation of iterated multiplication of very large polynomials: An application to chinese remainder theory ANZIAM J. 48 (CTAC2006) pp.c709 C724, 2007 C709 Fast evaluation of iterated multiplication of very large polynomials: An application to chinese remainder theory D. Laing 1 B. Litow 2 (Received 30 August

More information

International Association of Scientific Innovation and Research (IASIR) (An Association Unifying the Sciences, Engineering, and Applied Research)

International Association of Scientific Innovation and Research (IASIR) (An Association Unifying the Sciences, Engineering, and Applied Research) International Association of Scientific Innovation and Research (IASIR) (An Association Unifying the Sciences, Engineering, and Applied Research) International Journal of Emerging Technologies in Computational

More information

Robust Polynomial Reconstruction via Chinese Remainder Theorem in the Presence of Small Degree Residue Errors

Robust Polynomial Reconstruction via Chinese Remainder Theorem in the Presence of Small Degree Residue Errors 1 Robust Polynomial Reconstruction via Chinese Remainder Theorem in the Presence of Small Degree Residue Errors Li Xiao and Xiang-Gen Xia arxiv:1703.07907v1 [cs.it] 23 Mar 2017 Abstract Based on unique

More information

Technical Note

Technical Note Technical Note 1975-50 The Use of Finite Fields and Rings to Compute Convolutions I. S. Reed 6 June 1975 undi P n Lincoln Laboratory MASSACHUSETTS INSTITUTE OF TECH LEXINGTON, MASSACHI Approved for pn

More information

DESIGN OF QUANTIZED FIR FILTER USING COMPENSATING ZEROS

DESIGN OF QUANTIZED FIR FILTER USING COMPENSATING ZEROS DESIGN OF QUANTIZED FIR FILTER USING COMPENSATING ZEROS Nivedita Yadav, O.P. Singh, Ashish Dixit Department of Electronics and Communication Engineering, Amity University, Lucknow Campus, Lucknow, (India)

More information

KEYWORDS: Multiple Valued Logic (MVL), Residue Number System (RNS), Quinary Logic (Q uin), Quinary Full Adder, QFA, Quinary Half Adder, QHA.

KEYWORDS: Multiple Valued Logic (MVL), Residue Number System (RNS), Quinary Logic (Q uin), Quinary Full Adder, QFA, Quinary Half Adder, QHA. GLOBAL JOURNAL OF ADVANCED ENGINEERING TECHNOLOGIES AND SCIENCES DESIGN OF A QUINARY TO RESIDUE NUMBER SYSTEM CONVERTER USING MULTI-LEVELS OF CONVERSION Hassan Amin Osseily Electrical and Electronics Department,

More information

Subquadratic Space Complexity Multiplication over Binary Fields with Dickson Polynomial Representation

Subquadratic Space Complexity Multiplication over Binary Fields with Dickson Polynomial Representation Subquadratic Space Complexity Multiplication over Binary Fields with Dickson Polynomial Representation M A Hasan and C Negre Abstract We study Dickson bases for binary field representation Such representation

More information

Optimal Extension Field Inversion in the Frequency Domain

Optimal Extension Field Inversion in the Frequency Domain Optimal Extension Field Inversion in the Frequency Domain Selçuk Baktır, Berk Sunar WPI, Cryptography & Information Security Laboratory, Worcester, MA, USA Abstract. In this paper, we propose an adaptation

More information

10 Public Key Cryptography : RSA

10 Public Key Cryptography : RSA 10 Public Key Cryptography : RSA 10.1 Introduction The idea behind a public-key system is that it might be possible to find a cryptosystem where it is computationally infeasible to determine d K even if

More information

Security Level of Cryptography Integer Factoring Problem (Factoring N = p 2 q) December Summary 2

Security Level of Cryptography Integer Factoring Problem (Factoring N = p 2 q) December Summary 2 Security Level of Cryptography Integer Factoring Problem (Factoring N = p 2 ) December 2001 Contents Summary 2 Detailed Evaluation 3 1 The Elliptic Curve Method 3 1.1 The ECM applied to N = p d............................

More information

Algebra. Modular arithmetic can be handled mathematically by introducing a congruence relation on the integers described in the above example.

Algebra. Modular arithmetic can be handled mathematically by introducing a congruence relation on the integers described in the above example. Coding Theory Massoud Malek Algebra Congruence Relation The definition of a congruence depends on the type of algebraic structure under consideration Particular definitions of congruence can be made for

More information

Reverse Berlekamp-Massey Decoding

Reverse Berlekamp-Massey Decoding Reverse Berlekamp-Massey Decoding Jiun-Hung Yu and Hans-Andrea Loeliger Department of Information Technology and Electrical Engineering ETH Zurich, Switzerland Email: {yu, loeliger}@isi.ee.ethz.ch arxiv:1301.736v

More information

Maximally Flat Lowpass Digital Differentiators

Maximally Flat Lowpass Digital Differentiators Maximally Flat Lowpass Digital Differentiators Ivan W. Selesnick August 3, 00 Electrical Engineering, Polytechnic University 6 Metrotech Center, Brooklyn, NY 0 selesi@taco.poly.edu tel: 78 60-36 fax: 78

More information

AN IMPROVED LOW LATENCY SYSTOLIC STRUCTURED GALOIS FIELD MULTIPLIER

AN IMPROVED LOW LATENCY SYSTOLIC STRUCTURED GALOIS FIELD MULTIPLIER Indian Journal of Electronics and Electrical Engineering (IJEEE) Vol.2.No.1 2014pp1-6 available at: www.goniv.com Paper Received :05-03-2014 Paper Published:28-03-2014 Paper Reviewed by: 1. John Arhter

More information

Fast convolution Implementation using FNT

Fast convolution Implementation using FNT VLSI Signal Processing Final Project Fast convolution Implementation using FT R89921145 林秉勳 Contents I. Introduction II. The round-off and truncation issues III. Fast convolution IV. The structure of transforms

More information

REED-SOLOMON CODE SYMBOL AVOIDANCE

REED-SOLOMON CODE SYMBOL AVOIDANCE Vol105(1) March 2014 SOUTH AFRICAN INSTITUTE OF ELECTRICAL ENGINEERS 13 REED-SOLOMON CODE SYMBOL AVOIDANCE T Shongwe and A J Han Vinck Department of Electrical and Electronic Engineering Science, University

More information

Residue Number Systems Ivor Page 1

Residue Number Systems Ivor Page 1 Residue Number Systems 1 Residue Number Systems Ivor Page 1 7.1 Arithmetic in a modulus system The great speed of arithmetic in Residue Number Systems (RNS) comes from a simple theorem from number theory:

More information

CPSC 518 Introduction to Computer Algebra Schönhage and Strassen s Algorithm for Integer Multiplication

CPSC 518 Introduction to Computer Algebra Schönhage and Strassen s Algorithm for Integer Multiplication CPSC 518 Introduction to Computer Algebra Schönhage and Strassen s Algorithm for Integer Multiplication March, 2006 1 Introduction We have now seen that the Fast Fourier Transform can be applied to perform

More information

A Low-Error Statistical Fixed-Width Multiplier and Its Applications

A Low-Error Statistical Fixed-Width Multiplier and Its Applications A Low-Error Statistical Fixed-Width Multiplier and Its Applications Yuan-Ho Chen 1, Chih-Wen Lu 1, Hsin-Chen Chiang, Tsin-Yuan Chang, and Chin Hsia 3 1 Department of Engineering and System Science, National

More information

Continuing discussion of CRC s, especially looking at two-bit errors

Continuing discussion of CRC s, especially looking at two-bit errors Continuing discussion of CRC s, especially looking at two-bit errors The definition of primitive binary polynomials Brute force checking for primitivity A theorem giving a better test for primitivity Fast

More information

Design and Implementation of Efficient Modulo 2 n +1 Adder

Design and Implementation of Efficient Modulo 2 n +1 Adder www..org 18 Design and Implementation of Efficient Modulo 2 n +1 Adder V. Jagadheesh 1, Y. Swetha 2 1,2 Research Scholar(INDIA) Abstract In this brief, we proposed an efficient weighted modulo (2 n +1)

More information

Speedy Maths. David McQuillan

Speedy Maths. David McQuillan Speedy Maths David McQuillan Basic Arithmetic What one needs to be able to do Addition and Subtraction Multiplication and Division Comparison For a number of order 2 n n ~ 100 is general multi precision

More information

Two -Dimensional Digital Signal Processing II

Two -Dimensional Digital Signal Processing II Two -Dimensional Digital Signal Processing II Transforms and Median Filters Edited by T. S. Huang With Contributions by J.-O. Eklundh T.S. Huang B.I. Justusson H. J. Nussbaumer S.G. Tyan S. Zohar With

More information

GAMINGRE 8/1/ of 7

GAMINGRE 8/1/ of 7 FYE 09/30/92 JULY 92 0.00 254,550.00 0.00 0 0 0 0 0 0 0 0 0 254,550.00 0.00 0.00 0.00 0.00 254,550.00 AUG 10,616,710.31 5,299.95 845,656.83 84,565.68 61,084.86 23,480.82 339,734.73 135,893.89 67,946.95

More information

Chapter 4 Mathematics of Cryptography

Chapter 4 Mathematics of Cryptography Chapter 4 Mathematics of Cryptography Part II: Algebraic Structures Copyright The McGraw-Hill Companies, Inc. Permission required for reproduction or display. 4.1 Chapter 4 Objectives To review the concept

More information

Cyclic convolution of long sequences using number theoretic transform

Cyclic convolution of long sequences using number theoretic transform Cyclic convolution of long sequences using number theoretic transform Wan-Chi Siu, AP(HK). M.Phil.. C.Eng., M.I.ERE., Mem.I.E.E.E.. A.G. Constantinides, B.Sc.(Eng.), Ph.D., C.Eng., M.I.E.E., Sen.Mem.I.E.E.E.

More information

CHAPTER 3. Congruences. Congruence: definitions and properties

CHAPTER 3. Congruences. Congruence: definitions and properties CHAPTER 3 Congruences Part V of PJE Congruence: definitions and properties Definition. (PJE definition 19.1.1) Let m > 0 be an integer. Integers a and b are congruent modulo m if m divides a b. We write

More information

Closed-Form Design of Maximally Flat IIR Half-Band Filters

Closed-Form Design of Maximally Flat IIR Half-Band Filters IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 49, NO. 6, JUNE 2002 409 Closed-Form Design of Maximally Flat IIR Half-B Filters Xi Zhang, Senior Member, IEEE,

More information

Mathematics of Cryptography

Mathematics of Cryptography UNIT - III Mathematics of Cryptography Part III: Primes and Related Congruence Equations 1 Objectives To introduce prime numbers and their applications in cryptography. To discuss some primality test algorithms

More information

On the Construction and Decoding of Cyclic LDPC Codes

On the Construction and Decoding of Cyclic LDPC Codes On the Construction and Decoding of Cyclic LDPC Codes Chao Chen Joint work with Prof. Baoming Bai from Xidian University April 30, 2014 Outline 1. Introduction 2. Construction based on Idempotents and

More information

Computer Architecture 10. Residue Number Systems

Computer Architecture 10. Residue Number Systems Computer Architecture 10 Residue Number Systems Ma d e wi t h Op e n Of f i c e. o r g 1 A Puzzle What number has the reminders 2, 3 and 2 when divided by the numbers 7, 5 and 3? x mod 7 = 2 x mod 5 =

More information

A Search for Large Twin Prime Pairs. By R. E. Crandall and M. A. Penk. Abstract. Two methods are discussed for finding large integers m such that m I

A Search for Large Twin Prime Pairs. By R. E. Crandall and M. A. Penk. Abstract. Two methods are discussed for finding large integers m such that m I MATHEMATICS OF COMPUTATION, VOLUME 33, NUMBER 145 JANUARY 1979, PAGES 383-388 A Search for Large Twin Prime Pairs By R. E. Crandall and M. A. Penk Abstract. Two methods are discussed for finding large

More information

RSA Implementation. Oregon State University

RSA Implementation. Oregon State University RSA Implementation Çetin Kaya Koç Oregon State University 1 Contents: Exponentiation heuristics Multiplication algorithms Computation of GCD and Inverse Chinese remainder algorithm Primality testing 2

More information

Existence of Cyclic Hadamard Difference Sets and its Relation to Binary Sequences with Ideal Autocorrelation

Existence of Cyclic Hadamard Difference Sets and its Relation to Binary Sequences with Ideal Autocorrelation 14 JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL.1, NO.1, MARCH 1999 Existence of Cyclic Hadamard Difference Sets and its Relation to Binary Sequences with Ideal Autocorrelation Jeong-Heon Kim and Hong-Yeop

More information

Constructions of Nonbinary Quasi-Cyclic LDPC Codes: A Finite Field Approach

Constructions of Nonbinary Quasi-Cyclic LDPC Codes: A Finite Field Approach Constructions of Nonbinary Quasi-Cyclic LDPC Codes: A Finite Field Approach Shu Lin, Shumei Song, Lan Lan, Lingqi Zeng and Ying Y Tai Department of Electrical & Computer Engineering University of California,

More information

A Generalization of Wilson s Theorem

A Generalization of Wilson s Theorem A Generalization of Wilson s Theorem R. Andrew Ohana June 3, 2009 Contents 1 Introduction 2 2 Background Algebra 2 2.1 Groups................................. 2 2.2 Rings.................................

More information

Computer Vision, Convolutions, Complexity and Algebraic Geometry

Computer Vision, Convolutions, Complexity and Algebraic Geometry Computer Vision, Convolutions, Complexity and Algebraic Geometry D. V. Chudnovsky, G.V. Chudnovsky IMAS Polytechnic Institute of NYU 6 MetroTech Center Brooklyn, NY 11201 December 6, 2012 Fast Multiplication:

More information

Some Results on the Arithmetic Correlation of Sequences

Some Results on the Arithmetic Correlation of Sequences Some Results on the Arithmetic Correlation of Sequences Mark Goresky Andrew Klapper Abstract In this paper we study various properties of arithmetic correlations of sequences. Arithmetic correlations are

More information

Novel Modulo 2 n +1Multipliers

Novel Modulo 2 n +1Multipliers Novel Modulo Multipliers H. T. Vergos Computer Engineering and Informatics Dept., University of Patras, 26500 Patras, Greece. vergos@ceid.upatras.gr C. Efstathiou Informatics Dept.,TEI of Athens, 12210

More information

Subquadratic space complexity multiplier for a class of binary fields using Toeplitz matrix approach

Subquadratic space complexity multiplier for a class of binary fields using Toeplitz matrix approach Subquadratic space complexity multiplier for a class of binary fields using Toeplitz matrix approach M A Hasan 1 and C Negre 2 1 ECE Department and CACR, University of Waterloo, Ontario, Canada 2 Team

More information

CPSC 518 Introduction to Computer Algebra Asymptotically Fast Integer Multiplication

CPSC 518 Introduction to Computer Algebra Asymptotically Fast Integer Multiplication CPSC 518 Introduction to Computer Algebra Asymptotically Fast Integer Multiplication 1 Introduction We have now seen that the Fast Fourier Transform can be applied to perform polynomial multiplication

More information

Old and new algorithms for computing Bernoulli numbers

Old and new algorithms for computing Bernoulli numbers Old and new algorithms for computing Bernoulli numbers University of New South Wales 25th September 2012, University of Ballarat Bernoulli numbers Rational numbers B 0, B 1,... defined by: x e x 1 = n

More information

A COMBINED 16-BIT BINARY AND DUAL GALOIS FIELD MULTIPLIER. Jesus Garcia and Michael J. Schulte

A COMBINED 16-BIT BINARY AND DUAL GALOIS FIELD MULTIPLIER. Jesus Garcia and Michael J. Schulte A COMBINED 16-BIT BINARY AND DUAL GALOIS FIELD MULTIPLIER Jesus Garcia and Michael J. Schulte Lehigh University Department of Computer Science and Engineering Bethlehem, PA 15 ABSTRACT Galois field arithmetic

More information

The Karhunen-Loeve, Discrete Cosine, and Related Transforms Obtained via the Hadamard Transform

The Karhunen-Loeve, Discrete Cosine, and Related Transforms Obtained via the Hadamard Transform The Karhunen-Loeve, Discrete Cosine, and Related Transforms Obtained via the Hadamard Transform Item Type text; Proceedings Authors Jones, H. W.; Hein, D. N.; Knauer, S. C. Publisher International Foundation

More information

Three Ways to Test Irreducibility

Three Ways to Test Irreducibility Three Ways to Test Irreducibility Richard P. Brent Australian National University joint work with Paul Zimmermann INRIA, Nancy France 12 Feb 2009 Outline Polynomials over finite fields Irreducibility criteria

More information

Low complexity bit-parallel GF (2 m ) multiplier for all-one polynomials

Low complexity bit-parallel GF (2 m ) multiplier for all-one polynomials Low complexity bit-parallel GF (2 m ) multiplier for all-one polynomials Yin Li 1, Gong-liang Chen 2, and Xiao-ning Xie 1 Xinyang local taxation bureau, Henan, China. Email:yunfeiyangli@gmail.com, 2 School

More information

On Walsh transform and matrix factorization 1

On Walsh transform and matrix factorization 1 Eighth International Workshop on Optimal Codes and Related Topics July 10-14, 2017, Sofia, Bulgaria pp. 55-60 On Walsh transform and matrix factorization 1 Iliya Bouyukliev iliyab@math.bas.bg Paskal Piperkov

More information

Implementation Of Digital Fir Filter Using Improved Table Look Up Scheme For Residue Number System

Implementation Of Digital Fir Filter Using Improved Table Look Up Scheme For Residue Number System Implementation Of Digital Fir Filter Using Improved Table Look Up Scheme For Residue Number System G.Suresh, G.Indira Devi, P.Pavankumar Abstract The use of the improved table look up Residue Number System

More information

Multiplicative Order of Gauss Periods

Multiplicative Order of Gauss Periods Multiplicative Order of Gauss Periods Omran Ahmadi Department of Electrical and Computer Engineering University of Toronto Toronto, Ontario, M5S 3G4, Canada oahmadid@comm.utoronto.ca Igor E. Shparlinski

More information

Integer multiplication with generalized Fermat primes

Integer multiplication with generalized Fermat primes Integer multiplication with generalized Fermat primes CARAMEL Team, LORIA, University of Lorraine Supervised by: Emmanuel Thomé and Jérémie Detrey Journées nationales du Calcul Formel 2015 (Cluny) November

More information

EULER S THEOREM KEITH CONRAD

EULER S THEOREM KEITH CONRAD EULER S THEOREM KEITH CONRAD. Introduction Fermat s little theorem is an important property of integers to a prime modulus. Theorem. (Fermat). For prime p and any a Z such that a 0 mod p, a p mod p. If

More information

HARDWARE IMPLEMENTATION OF FIR/IIR DIGITAL FILTERS USING INTEGRAL STOCHASTIC COMPUTATION. Arash Ardakani, François Leduc-Primeau and Warren J.

HARDWARE IMPLEMENTATION OF FIR/IIR DIGITAL FILTERS USING INTEGRAL STOCHASTIC COMPUTATION. Arash Ardakani, François Leduc-Primeau and Warren J. HARWARE IMPLEMENTATION OF FIR/IIR IGITAL FILTERS USING INTEGRAL STOCHASTIC COMPUTATION Arash Ardakani, François Leduc-Primeau and Warren J. Gross epartment of Electrical and Computer Engineering McGill

More information

GF(2 m ) arithmetic: summary

GF(2 m ) arithmetic: summary GF(2 m ) arithmetic: summary EE 387, Notes 18, Handout #32 Addition/subtraction: bitwise XOR (m gates/ops) Multiplication: bit serial (shift and add) bit parallel (combinational) subfield representation

More information

E The Fast Fourier Transform

E The Fast Fourier Transform Fourier Transform Methods in Finance By Umberto Cherubini Giovanni Della Lunga Sabrina Mulinacci Pietro Rossi Copyright 2010 John Wiley & Sons Ltd E The Fast Fourier Transform E.1 DISCRETE FOURIER TRASFORM

More information

Three Ways to Test Irreducibility

Three Ways to Test Irreducibility Outline Three Ways to Test Irreducibility Richard P. Brent Australian National University joint work with Paul Zimmermann INRIA, Nancy France 8 Dec 2008 Polynomials over finite fields Irreducibility criteria

More information

Statistical Properties of the Arithmetic Correlation of Sequences. Mark Goresky School of Mathematics Institute for Advanced Study

Statistical Properties of the Arithmetic Correlation of Sequences. Mark Goresky School of Mathematics Institute for Advanced Study International Journal of Foundations of Computer Science c World Scientific Publishing Company Statistical Properties of the Arithmetic Correlation of Sequences Mark Goresky School of Mathematics Institute

More information

Lucas Lehmer primality test - Wikipedia, the free encyclopedia

Lucas Lehmer primality test - Wikipedia, the free encyclopedia Lucas Lehmer primality test From Wikipedia, the free encyclopedia In mathematics, the Lucas Lehmer test (LLT) is a primality test for Mersenne numbers. The test was originally developed by Edouard Lucas

More information

Faster integer multiplication using short lattice vectors

Faster integer multiplication using short lattice vectors Faster integer multiplication using short lattice vectors David Harvey and Joris van der Hoeven ANTS XIII, University of Wisconsin, Madison, July 2018 University of New South Wales / CNRS, École Polytechnique

More information

New algebraic decoding method for the (41, 21,9) quadratic residue code

New algebraic decoding method for the (41, 21,9) quadratic residue code New algebraic decoding method for the (41, 21,9) quadratic residue code Mohammed M. Al-Ashker a, Ramez Al.Shorbassi b a Department of Mathematics Islamic University of Gaza, Palestine b Ministry of education,

More information

4214 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 11, NOVEMBER 2006

4214 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 11, NOVEMBER 2006 4214 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 11, NOVEMBER 2006 Closed-Form Design of Generalized Maxflat R-Regular FIR M th-band Filters Using Waveform Moments Xi Zhang, Senior Member, IEEE,

More information

Lifting Parameterisation of the 9/7 Wavelet Filter Bank and its Application in Lossless Image Compression

Lifting Parameterisation of the 9/7 Wavelet Filter Bank and its Application in Lossless Image Compression Lifting Parameterisation of the 9/7 Wavelet Filter Bank and its Application in Lossless Image Compression TILO STRUTZ Deutsche Telekom AG, Hochschule für Telekommunikation Institute of Communications Engineering

More information

Numbering Systems. Contents: Binary & Decimal. Converting From: B D, D B. Arithmetic operation on Binary.

Numbering Systems. Contents: Binary & Decimal. Converting From: B D, D B. Arithmetic operation on Binary. Numbering Systems Contents: Binary & Decimal. Converting From: B D, D B. Arithmetic operation on Binary. Addition & Subtraction using Octal & Hexadecimal 2 s Complement, Subtraction Using 2 s Complement.

More information

SQUARE PATTERNS AND INFINITUDE OF PRIMES

SQUARE PATTERNS AND INFINITUDE OF PRIMES SQUARE PATTERNS AND INFINITUDE OF PRIMES KEITH CONRAD 1. Introduction Numerical data suggest the following patterns for prime numbers p: 1 mod p p = 2 or p 1 mod 4, 2 mod p p = 2 or p 1, 7 mod 8, 2 mod

More information

COMPUTER ARITHMETIC. 13/05/2010 cryptography - math background pp. 1 / 162

COMPUTER ARITHMETIC. 13/05/2010 cryptography - math background pp. 1 / 162 COMPUTER ARITHMETIC 13/05/2010 cryptography - math background pp. 1 / 162 RECALL OF COMPUTER ARITHMETIC computers implement some types of arithmetic for instance, addition, subtratction, multiplication

More information

The New Largest Known Prime is 2 p 1 With p = Who Cares? Sam Wagstaff Computer Sciences and Mathematics.

The New Largest Known Prime is 2 p 1 With p = Who Cares? Sam Wagstaff Computer Sciences and Mathematics. The New Largest Known Prime is 2 p 1 With p = 74207281. Who Cares? Sam Wagstaff Computer Sciences and Mathematics November 10, 2016 Earlier in 2016, Cooper, Woltman, Kurowski, Blosser and GIMPS found this

More information

How does the computer generate observations from various distributions specified after input analysis?

How does the computer generate observations from various distributions specified after input analysis? 1 How does the computer generate observations from various distributions specified after input analysis? There are two main components to the generation of observations from probability distributions.

More information

International Journal of Advanced Research in Computer Science and Software Engineering

International Journal of Advanced Research in Computer Science and Software Engineering Volume 2, Issue 8, August 2012 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com An Efficient

More information

Decomposing Bent Functions

Decomposing Bent Functions 2004 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 49, NO. 8, AUGUST 2003 Decomposing Bent Functions Anne Canteaut and Pascale Charpin Abstract In a recent paper [1], it is shown that the restrictions

More information

How to Write Fast Numerical Code

How to Write Fast Numerical Code How to Write Fast Numerical Code Lecture: Discrete Fourier transform, fast Fourier transforms Instructor: Markus Püschel TA: Georg Ofenbeck & Daniele Spampinato Rest of Semester Today Lecture Project meetings

More information

Counting Two-State Transition-Tour Sequences

Counting Two-State Transition-Tour Sequences Counting Two-State Transition-Tour Sequences Nirmal R. Saxena & Edward J. McCluskey Center for Reliable Computing, ERL 460 Department of Electrical Engineering, Stanford University, Stanford, CA 94305

More information

DSP Design Lecture 2. Fredrik Edman.

DSP Design Lecture 2. Fredrik Edman. DSP Design Lecture Number representation, scaling, quantization and round-off Noise Fredrik Edman fredrik.edman@eit.lth.se Representation of Numbers Numbers is a way to use symbols to describe and model

More information

Performance Evaluation of Signed-Digit Architecture for Weighted-to-Residue and Residue-to-Weighted Number Converters with Moduli Set (2 n 1, 2 n,

Performance Evaluation of Signed-Digit Architecture for Weighted-to-Residue and Residue-to-Weighted Number Converters with Moduli Set (2 n 1, 2 n, Regular Paper Performance Evaluation of Signed-Digit Architecture for Weighted-to-Residue and Residue-to-Weighted Number Converters with Moduli Set (2 n 1, 2 n, 2 n +1) Shuangching Chen and Shugang Wei

More information