Practical Guides on Programming with Big Number Library in Scientific Researches

The International Journal Of Engineering And Science (IJES)
|| Volume || 6 || Issue || 2 || Pages || PP 06-11 || 2017 ||
ISSN (e): 2319 – 1813 ISSN (p): 2319 – 1805
www.theijes.com The IJES Page 6
Practical Guides on Programming with Big Number Library in
Scientific Researches
Jianhui LI1
, Xingbo WANG2
1
Department of Computer Science, Guangdong Neusoft Institute, FOSHAN CITY, GUANGDONG PROVINCE, PRC
2
Department of Mechatronic Engineering, FOSHAN University, FOSHAN CITY, GUANGDONG PROVINCE, PRC
--------------------------------------------------------ABSTRACT-----------------------------------------------------------
This article demonstrates some practical skills in programming with GMP and MPFR big number libraries
according to the traits of big numbers and C/C++ programming conventions. It points out that, the conditional
expressions, the incremental treatment, the loops together with their stops, and the pointer arguments are critical
issues in C/C++ programming with the big number libraries. By exhibiting certain guidance and demonstration,
the article summarizes a framework to the program that treats big number computations and presents a sample as
well as numerical experiments to factorize big odd composite number.
Keywords: Big number computation, GMP library, C/C++ programming, Integer factorization.
-------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: 25 January 2017 Date of Accepted: 09 February 2017
--------------------------------------------------------------------------------------------------------------------------------------
I. INTRODUCTION
Whenever a scientific project related big number computations, such as encryption and decryption, the GNU GMP
big number library is a mandatory tool for a programmer to choose, as stated in articles [1] and [2]. In point-of-view
of programming, big number computations form a special programming style distinguishing from the conventional
programming styles, as seen in articles [3] to [5]. Therefore, conventional programming habits need a lot of
modification to fit the big numbers’ traits, which exceed the conventional computers’ processing capabilities.
However, literatures of guidance to such a change are seldom seen in the public publications. This article presents
some key programming issues of using big number library so as to tell the beginners to program in an predictable
efficiency and quality.
II. TRAITS OF BIG NUMBERS
Big numbers means their representations exceed that of a conventional computer’s. For example, the biggest
integer that can be expressed by a conventional computer of 32 bits is 32
2 1 4294967295  while in some scientific
computation, an integer can be over 100 bits of decimal digits. For example, the number RSA100 is
15226050279225333605356183781326374297180681149613806886579084945801229632589528976540003
50692006139, which has 101 bits of decimal digits. Except for the big integers, scientific computations also often
meet very big real numbers, as introduced in article [6].
Hence big numbers have the following traits:
1. They are necessary in scientific computations;
2. They are too big to be represented in a conventional way;
3. There are special libraries to treat computations of big numbers;
4. Programming with big numbers requires special skills.
III. GMP AND GMFR BRIEFS
GMP is a free library for arbitrary precision arithmetic, operating on signed integers, rational numbers, and
floating-point numbers. GMP has a rich set of functions, and the functions have a regular interface.The main target
applications for GMP are cryptography applications and research, Internet security applications, algebra systems,
computational algebra research, etc. The first GMP release was made in 1991. Now it is version 6 and GMP is
distributed under the dual licenses, GNU LGPL v3 and GNU GPL v2, which make the library free to use, share,
and improve. GMP is said to be fastest both for small operands and for huge operands and is available for C/C++
programming, which can be referred to GMP library website at gmplib.org.
GNU MPFR is a GNU portable C library for arbitrary-precision binary floating-point computation based on GMP
Library. The computation is both efficient and has a well-defined semantics: the functions are completely specified
on all the possible operands and the results do not depend on the platform. MPFR provides each number a precision.
The floating-point results are correctly rounded to the precision of the target variable, in any of the four IEEE-754
rounding modes. MPFR implements all mathematical functions from C99 and other usual mathematical functions,

Practical Guides on Programming with Big Number Library in Scientific Researches
including the logarithm and exponential in natural base, base 2 and base 10, the log(1 )X , the six trigonometric and
hyperbolic functions and their inverses, the gamma, zeta and error functions, the arithmetic geometric mean, the
power (xy) function. All those functions are correctly rounded over their complete range. Detail functions and
programming reference can be seen at www.mpfr.org
IV. PROGRAMMING SKILL WITH GMP OR MPFR LIBRARY
Since both MPFR and GMP are C-style programmable, this section only takes GMP as an example to demonstrate
some key skills in the programming on six aspects: new data type, conditional expressions, increment and
decrement, loops and their stops, pointer arguments and their returning values, and using string to exchange values
4.1 New Data Types
GMP library supports three types of data, integer as well as rational and floating-point number. It uses keyword
mpz_t , mpq_t and mpf_t to denote the three kinds of big numbers. For example, the following codes declare two
variables of big integers and two variables of big floating-point numbers, respectively.
mpz_t m,n; /* declare two variables of big integers: m, n*/
mpf_t s,t; /* declare two variables of big floating-point numbers: s, t*/
The types mpz_t, mpq_t and mpf_t are actually pointer type of struct in C/C++. Thus these kinds of
variables cannot be directly evaluated. For example, the following statement is wrong.
m=1234567890102020; /*Wrong statement! */
Variables of big number type must be first initialized and then set a value, as following examples.
mpz_init2(m,200); /*Initialize variable m by 200 bits of space to store it */
mpz_set_ui(m, 50); /* Set unsigned integer 50 to m*/
There are several methods to set value to a variable of big number; readers can refer to the GMP library
manual to find them, which are very easy to understand.
4.2 Conditional Expression
In conventional C/C++ programming, a conditional expression is usually very simple. For example, A<B, A==B or
A<=B always occur in conventional C/C++ programming. However, these expressions cannot be directly used in a
program related with big numbers unless they are treated normally in advance. Note that, either C or C++ requires
type-compatible in compiling process; hence one needs first to treat in conditional expression the two sides to be
the same type, especially, the types of big numbers. For example, when judging if two big integers are equal, it is
recommended to use the following statements.
unsigned cmp; /*Define an unsigned cmp*/
cmp=mpz_cmp(m, n); /*Calling mpz_cmp function to compare m and n*/
if(cmp<0) gmp_printf(“%Zd is smaller than %Zd n”, m, n);
/*Note: should use gmp_printf and %Zd format because m and n are big integers */
if(cmp=0) gmp_printf(“%Zd is equal to %Zd n”, m, n);
Generally speaking, using mpz_cmp function to compare two big numbers is a better custom in programming
related with big number operations.
4.3 Increment and Decrement
In conventional C/C++ programming, expressions such as i++, --i, a+=b and a-=b, are frequently seen in programs.
But these seldom occur in program of big numbers. It is recommended to use the following statements to treat
increment or decrement of a variable m of big numbers.
unsigned inc=2; /* Define an unsigned integer inc*/
mpz_add_ui(m, m, inc); /* Add inc to m*/
mpz_sub_ui(n, n, inc); /*Subtraction inc from n*/
If the increment or decrement is another big number, the following statements can be a reference.
mpz_add(m, m, n); /* Add n to m*/

mpz_sub_ui(n, n, m); /*Subtraction m from n*/
4.4 Loops and Their stops
Loops might occur anywhere in program if they are required. In conventional program, there are three kinds of
loops, the for-loop, the while-loop and the do-while-loop. Programming with big numbers also applies these loops.
The difference between the big number programs and the conventional programs are their stops. A loop in program
with big numbers seldom has explicit conditions to stop the loop because many computations with big numbers are
highly related algorithms of searching, divide-and-conquer, etc., which performed by iteration or recursion.
Consequently, controlling the stop of a loop is a key issue in programs of big number. In practical computations,
while-loop and do-while-loop are appreciate and a stop activator is always placed in the loop, as the following
codes demonstrate.
/*===Search from S to find common divisor d between R and N====
======== until searching RcylNumber steps or find a d =========
=====where S, R, d, N and RcylNumber are all big numbers======*/
unsigned found=0; /* found: stop activator . If found =1, stop loop*/
unsigned cmp=1;
mpz_set_ui(tmp,0); /* Big number tmp is used to be a recycle variable*/
while(1) /* A never end loop */
{ mpz_add_ui(tmp,tmp,1); /* Recycle variable increases by 1*/
mpz_gcd(d , R, N); /* Compute common divisor d between R and N*/
cmp=mpz_cmp_ui(d,1); /* Judge if d=1?*/
if(cmp>0) /*d>1 means a common divisor is found*/
{ found=1; /*Activate stop of the loop*/
break; /* break the loop*/
}
mpz_add_ui(N,N,2); /*Otherwise, prepare next computation by changing N*/
cmp=mpz_cmp(tmp, RcylNumber); /*Judge if reaching the maximal computational steps*/
if(cmp==0) break; /*If it reaches, stop the loop*/
}
4.5 Pointer Arguments and Their Returning Values
As stated previously, the types mpz_t and mpf_t are actually pointer types of struct in C/C++; hence the arguments
of mpz_t or mpf_t in a function can by default return values of computations in the body of the function. This
behavior occurs both in gmp-defined function and in user-defined function. For example, the following
user-defined function SetMersenne (See in Section 5.2) can return a big number by its argument M.
4.6 Using String to Exchange Values
Sometimes, it is necessary to change part of a big number into a conventional number. Then the string-type can play
an important role in exchanging the data. Note that, gmp library provides function to change unsigned type into
mpz_t type, but does not provide a function to change the mpz_t type into the unsigned type. Hence using sscanf
function can easily solve the problem, as the following codes demonstrate.
unsigned getTail(mpz_t T)
{unsigned tail; /* An unsigned variable tail*/
char s[10]; /* An string s*/
mpz_t r; /* mpz_t type r*/
mpz_init2(r,10); /*Initialize r*/
mpz_mod_ui(r,T,10); /*r=T mod 10*/
mpz_get_str(s,10,r); /* Save the mpz_t type r to the string s*/
sscanf(s,"%d",&tail); /* Obtain tail from s*/
mpz_clear(r); /*Release r*/
return tail; /*Return tail*/
}
V. INSTANCE OF BIG NUMBER COMPUTATION
As stated above, programming with big number computations has its own traits based on two considerations. One
is that the program should meet the needs of the traits of big number libraries, and the other is that the program

should match to the algorithms of big number computations. This section presents a complete example that
factorizes a big odd number, as introduced in article [7].
5.1 Lemma and Algorithm
Lemma 1 Suppose N is an odd composite number; then the following statements are true.
(1) If 6 1N n  and there exists a
1
2
n
k

 such that (6 1) | ( )k n k  , then (6 1) |k N .
(2) If 6 1N n  and there exists a
1
2
n
k

 such that (6 1) | ( )k n k  , then (6 1) |k N
According to the lemma, algorithm to factorize a big odd integer can be simply designed by dealing with
the following issues.
========Algorithm To Factorize an Odd Number of 6n+1========
(1) Determine the form of an big odd number in terms of its form 6 1n  ;
(2) Extract n from N;
(3) Set a maximal computational limit m ax
1
2
n
k
 
  
 
;
(4) Search and find a k such that (6 1) | ( )k n k  or (6 1) | ( )k n k  .
===============End of Algorithm Descript=================
5.2 C Program to Realize the Algorithm
The previous algorithm can be realized by using GMP big number library as follows.
void SetMersenne(mpz_t M, unsigned int p)
/* Produce Mersennne Number M*/
{ mpz_set_ui(M,1);
mpz_mul_2exp (M, M, p);//M=M*2^p
mpz_sub_ui(M,M,1);
printf("n FactorizeMersenneNumber 2^(%u) - 1n",p);}
void get_n_from_N(mpz_t n, unsigned *tail, mpz_t N)
{/* Note: argument n is a pointer */
unsigned whatN;
char s[10]; /*Use string s*/
mpz_mod_ui(n,N,6); /* n=N mod 6, resulting 1,3,5*/
mpz_get_str(s,10,n); /* Change n to s*/
sscanf(s,"%d", &whatN); /* Obtain whatN from s*/
switch(whatN)
{ case 3: break; /* N is 3’s multiple*/
case 1: mpz_sub_ui(n,N,1); /* When N=6n+1*/
*tail=1; break;
case 5: mpz_add_ui(n,N,1); /* When N=6n-1*/
*tail=5; break; }
mpz_divexact_ui(n,n,6); /*Extracting n from N = 6n+1 or N= 6n -1*/
return ;}
void GetMaxLoopNumber(mpz_t lmax, mpz_t n)
{ mpz_sqrt(lmax,n); /*lmax=sqrt(N)*/
mpz_add_ui(lmax,lmax,1) ; /*lmax=lmax+1*/
mpz_fdiv_q_ui(lmax,lmax,2) ; /*lmax=lmax/2*/
}
void SetD_1(mpz_t d, mpz_t k)
{mpz_set(d,k); /*Thus d=k*/
mpz_mul_ui(d,d,6); /* Thus d=6*k*/
mpz_add_ui(d,d,1); /*Thus d=6*k+1*/
}
void SetD_2(mpz_t d, mpz_t k)

{mpz_set(d,k); /*Thus d=k*/
mpz_mul_ui(d,d,6); /* Thus d=6*k*/
mpz_sub_ui(d,d,1); /*Thus d=6*k+1*/
}
int DivTest1(mpz_t res, mpz_t d, mpz_t k, mpz_t n)
{mpz_sub(res,n,k); /*res=n-k*/
if(!mpz_divisible_p(res,d)) /* d not div res*/
return 0;
mpz_set(res,d); /*Set d to res */
return 1;}
int DivTest2(mpz_t res, mpz_t d, mpz_t k, mpz_t n)
{ mpz_add(res,n,k); /*res=n+k*/
if(!mpz_divisible_p(res,d)) /* d not div res)*/
return 0;
mpz_set(res,d); /*Set d to res */
return 1;}
void Factorize(mpz_t res, mpz_t k, mpz_t s, mpz_t d, mpz_t n, mpz_t lmax)
{ int cmp=1;
mpz_set_ui(k,0); /* Set initial value of tmp by 0 */
cmp=mpz_cmp(k, lmax); /* Compare k with the maximal loop-number lmax*/
while(cmp<0) /* cmp<0 means k<lmax */
{mpz_add_ui(k,k,1); /* k increases by 1*/
SetD_1(d, k); /* Set d=6*k+1 */
if(DivTest1(res,d,k,n)) break; /*When d|(n-k) then d|N, return res and stop loop*/
/*Otherwise test the other case*/
SetD_2(d, k); /* Set d=6*k-1 */
if(DivTest2 (res, d,k, n))
break; /*When d|(n-k) then d|N, return res and stop loop*/
cmp=mpz_cmp(k, lmax); /* Compare k with the maximal loop-number lmax*/
}
return ;}
void InitBigData(mpz_t &d1, mpz_t &d2, mpz_t &d3, mpz_t &d4, mpz_t &d5,
mpz_t &d6, mpz_t &d7, unsigned nBits)
{mpz_init2(d1,nBits); mpz_init2(d2,nBits);mpz_init2(d3,nBits);mpz_init2(d4,nBits);
mpz_init2(d5,nBits);mpz_init2(d6,nBits); mpz_init2(d7,nBits);}
void ClearMpzData(mpz_t d1, mpz_t d2, mpz_t d3, mpz_t d4,
mpz_t d5, mpz_t d6 , mpz_t d7)
{ mpz_clear(d1); mpz_clear(d2); mpz_clear(d3); mpz_clear(d4);
mpz_clear(d5); mpz_clear(d6); mpz_clear(d7); }
int main()
{ unsigned tail, Bits=200,p=113;
mpz_t N, factor,s,d,n,l,lmax;
InitBigData(factor, s, d, n, l, lmax, N,Bits);
SetMersenne(N, p);
get_n_from_N(n, &tail, N); /*Get tail of N and extract n from N*/
GetMaxLoopNumber(lmax, n);
if(tail==1)
{Factorize(factor, l, s, d, n, lmax);
gmp_printf("l=%Zd Factor=%Zdn", l,factor);
ClearMpzData(factor, s, d, n, l, lmax, N);
return;}
gmp_printf("Not fit the form 6n+1n");

ClearMpzData(factor, s, d, n, l, lmax, N);
Return 0;}
5.3 Numerical Tests
Applying the previous codes to test factorizing some of the Mersenne numbers reveals expected results. Table1 1
shows the experimental results, which are made on a PC with an Intel Xeon E5450 CPU and 4GB memory.
Table 1 Experiments on Mersenne and Fermat Numbers
N N’s factor Searching number k
M67=267
-1 193707721 32284620
M71=271
-1 228479 38080
M83=283
-1 167 28
M97=297
-1 11447 1908
M103=2103
-1 2550183799 435030633
M109=2109
-1 745988807 124331468
M113=2113
-1 3391 565
VI. CONCLUSION
GMP and MPFR big number libraries are frequently used in scientific computations. Programming with these
libraries requires certain skills, especially the expressions of conditions, increment, decrement, loops and their
stops. Only plentiful experiences can help a programmer to program in an efficient and highly skillful way. The
contents in this article summarize the authors’ consideration on the program from practical point of view. The
authors hope it a valuable reference to the related developers.
ACKNOWLEDGEMENTS
The research work is supported by Department of Guangdong Science and Technology under projects
2015A030401105 and 2015A010104011, Foshan Bureau of Science and Technology under projects
2016AG100311, 2016AG100652, 2016AG100792 and 2016AG100382 Special Innovative Projects
2014KTSCX156, 2014SFKC30 and 2014QTLXXM42 from Guangdong Education Department. The authors
sincerely present thanks to them all.
REFERENCES
[1]. L. Rutten and M.V. Eekelen. Efficient and formally proven reduction of large integers by small moduli, Acm Transactions on
Mathematical Software, 2008, 37(2):163-172.
[2]. Isa, Mohd Anuar Mat, et al. A Series of Secret Keys in a Key Distribution Protocol. Transactions on Engineering Technologies.
Springer Netherlands, 2015:615-628.
[3]. W.B. Hart. Fast Library for Number Theory: An Introduction (Mathematical Software – ICMS 2010. Springer Berlin Heidelberg,
2010)
[4]. L. M. Surhone, M. T.Tennoe and S. F. Henssonow. Fast Library for Number Theory (Betascript Publishing, 2010)
[5]. T. Granlund and G. D. Team. GNU MP 6.0 Multiple Precision Arithmetic Library (Samurai Media Limited, 2015).
[6]. L Fousse,G Hanrot, V Lefevre, et al. MPFR: A multiple-precision binary floating-point library with correct rounding, ACM Trans.
Math. Softw. 2007:00000818.
[7]. Xingbo WANG. Seed and Sieve of Odd Composite Numbers with Applications In Factorization of Integers, OSR Journal of
Mathematics (IOSR-JM), 2016, 12(5, Ver. VIII): 01-07
Author’s Biography
Jianhui Li is an assistant professor of Guangdong Neusoft Institute. He obtained his Ph D degree in
Hunan Agriculture University in 2013, and since then has been a staff in charge of affairs of
scientific researches in the university. He is skill at computer programming, wireless-sensor
network development and computer image process.

Practical Guides on Programming with Big Number Library in Scientific Researches

More Related Content

Viewers also liked (20)

Similar to Practical Guides on Programming with Big Number Library in Scientific Researches (20)

Recently uploaded (20)

Practical Guides on Programming with Big Number Library in Scientific Researches