Tcl Source Code

*/config.cache
*/config.log
*/config.status
*/tclConfig.sh
*/tclsh*
*/tcltest*
*/versions.vc












unix/autoMkindex.tcl
unix/dltest.marker
unix/tcl.pc
unix/tclIndex
unix/pkgs/*
win/pkgs/*
win/tcl.hpj

*/config.cache
*/config.log
*/config.status
*/tclConfig.sh
*/tclsh*
*/tcltest*
*/versions.vc
libtommath/bn.ilg
libtommath/bn.ind
libtommath/*.pdf
libtommath/tombc/*
libtommath/pre_gen/*
libtommath/pics/*
libtommath/mtest/*
libtommath/logs/*
libtommath/etc/*
libtommath/demo/*
libtommath/*.out
libtommath/*.tex
unix/autoMkindex.tcl
unix/dltest.marker
unix/tcl.pc
unix/tclIndex
unix/pkgs/*
win/pkgs/*
win/tcl.hpj

This is makeindex, version 2.14 [02-Oct-2002] (kpathsea + Thai support).
Scanning input file bn.idx....done (79 entries accepted, 0 rejected).
Sorting entries....done (511 comparisons).
Generating output file bn.ind....done (82 lines written, 0 warnings).
Output written in bn.ind.
Transcript written in bn.ilg.

\begin{theindex}

  \item mp\_add, \hyperpage{29}
  \item mp\_add\_d, \hyperpage{52}
  \item mp\_and, \hyperpage{29}
  \item mp\_clear, \hyperpage{11}
  \item mp\_clear\_multi, \hyperpage{12}
  \item mp\_cmp, \hyperpage{24}
  \item mp\_cmp\_d, \hyperpage{25}
  \item mp\_cmp\_mag, \hyperpage{23}
  \item mp\_div, \hyperpage{30}
  \item mp\_div\_2, \hyperpage{26}
  \item mp\_div\_2d, \hyperpage{28}
  \item mp\_div\_d, \hyperpage{52}
  \item mp\_dr\_reduce, \hyperpage{40}
  \item mp\_dr\_setup, \hyperpage{40}
  \item MP\_EQ, \hyperpage{22}
  \item mp\_error\_to\_string, \hyperpage{10}
  \item mp\_expt\_d, \hyperpage{43}
  \item mp\_exptmod, \hyperpage{43}
  \item mp\_exteuclid, \hyperpage{51}
  \item mp\_gcd, \hyperpage{51}
  \item mp\_get\_int, \hyperpage{20}
  \item mp\_grow, \hyperpage{16}
  \item MP\_GT, \hyperpage{22}
  \item mp\_init, \hyperpage{11}
  \item mp\_init\_copy, \hyperpage{13}
  \item mp\_init\_multi, \hyperpage{12}
  \item mp\_init\_set, \hyperpage{21}
  \item mp\_init\_set\_int, \hyperpage{21}
  \item mp\_init\_size, \hyperpage{14}
  \item mp\_int, \hyperpage{10}
  \item mp\_invmod, \hyperpage{52}
  \item mp\_jacobi, \hyperpage{52}
  \item mp\_lcm, \hyperpage{51}
  \item mp\_lshd, \hyperpage{28}
  \item MP\_LT, \hyperpage{22}
  \item MP\_MEM, \hyperpage{9}
  \item mp\_mod, \hyperpage{35}
  \item mp\_mod\_d, \hyperpage{52}
  \item mp\_montgomery\_calc\_normalization, \hyperpage{38}
  \item mp\_montgomery\_reduce, \hyperpage{37}
  \item mp\_montgomery\_setup, \hyperpage{37}
  \item mp\_mul, \hyperpage{31}
  \item mp\_mul\_2, \hyperpage{26}
  \item mp\_mul\_2d, \hyperpage{28}
  \item mp\_mul\_d, \hyperpage{52}
  \item mp\_n\_root, \hyperpage{44}
  \item mp\_neg, \hyperpage{29}
  \item MP\_NO, \hyperpage{9}
  \item MP\_OKAY, \hyperpage{9}
  \item mp\_or, \hyperpage{29}
  \item mp\_prime\_fermat, \hyperpage{45}
  \item mp\_prime\_is\_divisible, \hyperpage{45}
  \item mp\_prime\_is\_prime, \hyperpage{46}
  \item mp\_prime\_miller\_rabin, \hyperpage{45}
  \item mp\_prime\_next\_prime, \hyperpage{46}
  \item mp\_prime\_rabin\_miller\_trials, \hyperpage{46}
  \item mp\_prime\_random, \hyperpage{47}
  \item mp\_prime\_random\_ex, \hyperpage{47}
  \item mp\_radix\_size, \hyperpage{49}
  \item mp\_read\_radix, \hyperpage{49}
  \item mp\_read\_unsigned\_bin, \hyperpage{50}
  \item mp\_reduce, \hyperpage{36}
  \item mp\_reduce\_2k, \hyperpage{41}
  \item mp\_reduce\_2k\_setup, \hyperpage{41}
  \item mp\_reduce\_setup, \hyperpage{36}
  \item mp\_rshd, \hyperpage{28}
  \item mp\_set, \hyperpage{19}
  \item mp\_set\_int, \hyperpage{20}
  \item mp\_shrink, \hyperpage{15}
  \item mp\_sqr, \hyperpage{33}
  \item mp\_sub, \hyperpage{29}
  \item mp\_sub\_d, \hyperpage{52}
  \item mp\_to\_unsigned\_bin, \hyperpage{50}
  \item mp\_toradix, \hyperpage{49}
  \item mp\_unsigned\_bin\_size, \hyperpage{50}
  \item MP\_VAL, \hyperpage{9}
  \item mp\_xor, \hyperpage{29}
  \item MP\_YES, \hyperpage{9}

\end{theindex}

#include <time.h>

#ifdef IOWNANATHLON
#include <unistd.h>
#define SLEEP sleep(4)
#else
#define SLEEP
#endif

#include "tommath.h"

void ndraw(mp_int * a, char *name)
{
   char buf[16000];

   printf("%s: ", name);
   mp_toradix(a, buf, 10);
   printf("%s\n", buf);
}

static void draw(mp_int * a)
{
   ndraw(a, "");
}


unsigned long lfsr = 0xAAAAAAAAUL;

int lbit(void)
{
   if (lfsr & 0x80000000UL) {
      lfsr = ((lfsr << 1) ^ 0x8000001BUL) & 0xFFFFFFFFUL;
      return 1;
   } else {
      lfsr <<= 1;
      return 0;
   }
}

int myrng(unsigned char *dst, int len, void *dat)
{
   int x;

   for (x = 0; x < len; x++)
      dst[x] = rand() & 0xFF;
   return len;
}



char cmd[4096], buf[4096];
int main(void)
{
   mp_int a, b, c, d, e, f;
   unsigned long expt_n, add_n, sub_n, mul_n, div_n, sqr_n, mul2d_n, div2d_n,
      gcd_n, lcm_n, inv_n, div2_n, mul2_n, add_d_n, sub_d_n, t;
   unsigned rr;
   int i, n, err, cnt, ix, old_kara_m, old_kara_s;
   mp_digit mp;


   mp_init(&a);
   mp_init(&b);
   mp_init(&c);
   mp_init(&d);
   mp_init(&e);
   mp_init(&f);

   srand(time(NULL));

#if 0
   // test montgomery 
   printf("Testing montgomery...\n");
   for (i = 1; i < 10; i++) {
      printf("Testing digit size: %d\n", i);
      for (n = 0; n < 1000; n++) {
         mp_rand(&a, i);
         a.dp[0] |= 1;

         // let's see if R is right
         mp_montgomery_calc_normalization(&b, &a);
         mp_montgomery_setup(&a, &mp);

         // now test a random reduction 
         for (ix = 0; ix < 100; ix++) {
             mp_rand(&c, 1 + abs(rand()) % (2*i));
             mp_copy(&c, &d);
             mp_copy(&c, &e);

             mp_mod(&d, &a, &d);
             mp_montgomery_reduce(&c, &a, mp);
             mp_mulmod(&c, &b, &a, &c);

             if (mp_cmp(&c, &d) != MP_EQ) { 
printf("d = e mod a, c = e MOD a\n");
mp_todecimal(&a, buf); printf("a = %s\n", buf);
mp_todecimal(&e, buf); printf("e = %s\n", buf);
mp_todecimal(&d, buf); printf("d = %s\n", buf);
mp_todecimal(&c, buf); printf("c = %s\n", buf);
printf("compare no compare!\n"); exit(EXIT_FAILURE); }
         }
      }
   }
   printf("done\n");

   // test mp_get_int
   printf("Testing: mp_get_int\n");
   for (i = 0; i < 1000; ++i) {
      t = ((unsigned long) rand() * rand() + 1) & 0xFFFFFFFF;
      mp_set_int(&a, t);
      if (t != mp_get_int(&a)) {
	 printf("mp_get_int() bad result!\n");
	 return 1;
      }
   }
   mp_set_int(&a, 0);
   if (mp_get_int(&a) != 0) {
      printf("mp_get_int() bad result!\n");
      return 1;
   }
   mp_set_int(&a, 0xffffffff);
   if (mp_get_int(&a) != 0xffffffff) {
      printf("mp_get_int() bad result!\n");
      return 1;
   }
   // test mp_sqrt
   printf("Testing: mp_sqrt\n");
   for (i = 0; i < 1000; ++i) {
      printf("%6d\r", i);
      fflush(stdout);
      n = (rand() & 15) + 1;
      mp_rand(&a, n);
      if (mp_sqrt(&a, &b) != MP_OKAY) {
	 printf("mp_sqrt() error!\n");
	 return 1;
      }
      mp_n_root(&a, 2, &a);
      if (mp_cmp_mag(&b, &a) != MP_EQ) {
	 printf("mp_sqrt() bad result!\n");
	 return 1;
      }
   }

   printf("\nTesting: mp_is_square\n");
   for (i = 0; i < 1000; ++i) {
      printf("%6d\r", i);
      fflush(stdout);

      /* test mp_is_square false negatives */
      n = (rand() & 7) + 1;
      mp_rand(&a, n);
      mp_sqr(&a, &a);
      if (mp_is_square(&a, &n) != MP_OKAY) {
	 printf("fn:mp_is_square() error!\n");
	 return 1;
      }
      if (n == 0) {
	 printf("fn:mp_is_square() bad result!\n");
	 return 1;
      }

      /* test for false positives */
      mp_add_d(&a, 1, &a);
      if (mp_is_square(&a, &n) != MP_OKAY) {
	 printf("fp:mp_is_square() error!\n");
	 return 1;
      }
      if (n == 1) {
	 printf("fp:mp_is_square() bad result!\n");
	 return 1;
      }

   }
   printf("\n\n");

   /* test for size */
   for (ix = 10; ix < 128; ix++) {
      printf("Testing (not safe-prime): %9d bits    \r", ix);
      fflush(stdout);
      err =
	 mp_prime_random_ex(&a, 8, ix,
			    (rand() & 1) ? LTM_PRIME_2MSB_OFF :
			    LTM_PRIME_2MSB_ON, myrng, NULL);
      if (err != MP_OKAY) {
	 printf("failed with err code %d\n", err);
	 return EXIT_FAILURE;
      }
      if (mp_count_bits(&a) != ix) {
	 printf("Prime is %d not %d bits!!!\n", mp_count_bits(&a), ix);
	 return EXIT_FAILURE;
      }
   }

   for (ix = 16; ix < 128; ix++) {
      printf("Testing (   safe-prime): %9d bits    \r", ix);
      fflush(stdout);
      err =
	 mp_prime_random_ex(&a, 8, ix,
			    ((rand() & 1) ? LTM_PRIME_2MSB_OFF :
			     LTM_PRIME_2MSB_ON) | LTM_PRIME_SAFE, myrng,
			    NULL);
      if (err != MP_OKAY) {
	 printf("failed with err code %d\n", err);
	 return EXIT_FAILURE;
      }
      if (mp_count_bits(&a) != ix) {
	 printf("Prime is %d not %d bits!!!\n", mp_count_bits(&a), ix);
	 return EXIT_FAILURE;
      }
      /* let's see if it's really a safe prime */
      mp_sub_d(&a, 1, &a);
      mp_div_2(&a, &a);
      mp_prime_is_prime(&a, 8, &cnt);
      if (cnt != MP_YES) {
	 printf("sub is not prime!\n");
	 return EXIT_FAILURE;
      }
   }

   printf("\n\n");

   mp_read_radix(&a, "123456", 10);
   mp_toradix_n(&a, buf, 10, 3);
   printf("a == %s\n", buf);
   mp_toradix_n(&a, buf, 10, 4);
   printf("a == %s\n", buf);
   mp_toradix_n(&a, buf, 10, 30);
   printf("a == %s\n", buf);


#if 0
   for (;;) {
      fgets(buf, sizeof(buf), stdin);
      mp_read_radix(&a, buf, 10);
      mp_prime_next_prime(&a, 5, 1);
      mp_toradix(&a, buf, 10);
      printf("%s, %lu\n", buf, a.dp[0] & 3);
   }
#endif

   /* test mp_cnt_lsb */
   printf("testing mp_cnt_lsb...\n");
   mp_set(&a, 1);
   for (ix = 0; ix < 1024; ix++) {
      if (mp_cnt_lsb(&a) != ix) {
	 printf("Failed at %d, %d\n", ix, mp_cnt_lsb(&a));
	 return 0;
      }
      mp_mul_2(&a, &a);
   }

/* test mp_reduce_2k */
   printf("Testing mp_reduce_2k...\n");
   for (cnt = 3; cnt <= 128; ++cnt) {
      mp_digit tmp;

      mp_2expt(&a, cnt);
      mp_sub_d(&a, 2, &a);	/* a = 2**cnt - 2 */


      printf("\nTesting %4d bits", cnt);
      printf("(%d)", mp_reduce_is_2k(&a));
      mp_reduce_2k_setup(&a, &tmp);
      printf("(%d)", tmp);
      for (ix = 0; ix < 1000; ix++) {
	 if (!(ix & 127)) {
	    printf(".");
	    fflush(stdout);
	 }
	 mp_rand(&b, (cnt / DIGIT_BIT + 1) * 2);
	 mp_copy(&c, &b);
	 mp_mod(&c, &a, &c);
	 mp_reduce_2k(&b, &a, 2);
	 if (mp_cmp(&c, &b)) {
	    printf("FAILED\n");
	    exit(0);
	 }
      }
   }

/* test mp_div_3  */
   printf("Testing mp_div_3...\n");
   mp_set(&d, 3);
   for (cnt = 0; cnt < 10000;) {
      mp_digit r1, r2;

      if (!(++cnt & 127))
	 printf("%9d\r", cnt);
      mp_rand(&a, abs(rand()) % 128 + 1);
      mp_div(&a, &d, &b, &e);
      mp_div_3(&a, &c, &r2);

      if (mp_cmp(&b, &c) || mp_cmp_d(&e, r2)) {
	 printf("\n\nmp_div_3 => Failure\n");
      }
   }
   printf("\n\nPassed div_3 testing\n");

/* test the DR reduction */
   printf("testing mp_dr_reduce...\n");
   for (cnt = 2; cnt < 32; cnt++) {
      printf("%d digit modulus\n", cnt);
      mp_grow(&a, cnt);
      mp_zero(&a);
      for (ix = 1; ix < cnt; ix++) {
	 a.dp[ix] = MP_MASK;
      }
      a.used = cnt;
      a.dp[0] = 3;

      mp_rand(&b, cnt - 1);
      mp_copy(&b, &c);

      rr = 0;
      do {
	 if (!(rr & 127)) {
	    printf("%9lu\r", rr);
	    fflush(stdout);
	 }
	 mp_sqr(&b, &b);
	 mp_add_d(&b, 1, &b);
	 mp_copy(&b, &c);

	 mp_mod(&b, &a, &b);
	 mp_dr_reduce(&c, &a, (((mp_digit) 1) << DIGIT_BIT) - a.dp[0]);

	 if (mp_cmp(&b, &c) != MP_EQ) {
	    printf("Failed on trial %lu\n", rr);
	    exit(-1);

	 }
      } while (++rr < 500);
      printf("Passed DR test for %d digits\n", cnt);
   }

#endif

/* test the mp_reduce_2k_l code */
#if 0
#if 0
/* first load P with 2^1024 - 0x2A434 B9FDEC95 D8F9D550 FFFFFFFF FFFFFFFF */
   mp_2expt(&a, 1024);
   mp_read_radix(&b, "2A434B9FDEC95D8F9D550FFFFFFFFFFFFFFFF", 16);
   mp_sub(&a, &b, &a);
#elif 1
/*  p = 2^2048 - 0x1 00000000 00000000 00000000 00000000 4945DDBF 8EA2A91D 5776399B B83E188F  */
   mp_2expt(&a, 2048);
   mp_read_radix(&b,
		 "1000000000000000000000000000000004945DDBF8EA2A91D5776399BB83E188F",
		 16);
   mp_sub(&a, &b, &a);
#endif

   mp_todecimal(&a, buf);
   printf("p==%s\n", buf);
/* now mp_reduce_is_2k_l() should return */
   if (mp_reduce_is_2k_l(&a) != 1) {
      printf("mp_reduce_is_2k_l() return 0, should be 1\n");
      return EXIT_FAILURE;
   }
   mp_reduce_2k_setup_l(&a, &d);
   /* now do a million square+1 to see if it varies */
   mp_rand(&b, 64);
   mp_mod(&b, &a, &b);
   mp_copy(&b, &c);
   printf("testing mp_reduce_2k_l...");
   fflush(stdout);
   for (cnt = 0; cnt < (1UL << 20); cnt++) {
      mp_sqr(&b, &b);
      mp_add_d(&b, 1, &b);
      mp_reduce_2k_l(&b, &a, &d);
      mp_sqr(&c, &c);
      mp_add_d(&c, 1, &c);
      mp_mod(&c, &a, &c);
      if (mp_cmp(&b, &c) != MP_EQ) {
	 printf("mp_reduce_2k_l() failed at step %lu\n", cnt);
	 mp_tohex(&b, buf);
	 printf("b == %s\n", buf);
	 mp_tohex(&c, buf);
	 printf("c == %s\n", buf);
	 return EXIT_FAILURE;
      }
   }
   printf("...Passed\n");
#endif

   div2_n = mul2_n = inv_n = expt_n = lcm_n = gcd_n = add_n =
      sub_n = mul_n = div_n = sqr_n = mul2d_n = div2d_n = cnt = add_d_n =
      sub_d_n = 0;

   /* force KARA and TOOM to enable despite cutoffs */
   KARATSUBA_SQR_CUTOFF = KARATSUBA_MUL_CUTOFF = 8;
   TOOM_SQR_CUTOFF = TOOM_MUL_CUTOFF = 16;

   for (;;) {
      /* randomly clear and re-init one variable, this has the affect of triming the alloc space */
      switch (abs(rand()) % 7) {
      case 0:
	 mp_clear(&a);
	 mp_init(&a);
	 break;
      case 1:
	 mp_clear(&b);
	 mp_init(&b);
	 break;
      case 2:
	 mp_clear(&c);
	 mp_init(&c);
	 break;
      case 3:
	 mp_clear(&d);
	 mp_init(&d);
	 break;
      case 4:
	 mp_clear(&e);
	 mp_init(&e);
	 break;
      case 5:
	 mp_clear(&f);
	 mp_init(&f);
	 break;
      case 6:
	 break;			/* don't clear any */
      }


      printf
	 ("%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu/%4lu ",
	  add_n, sub_n, mul_n, div_n, sqr_n, mul2d_n, div2d_n, gcd_n, lcm_n,
	  expt_n, inv_n, div2_n, mul2_n, add_d_n, sub_d_n);
      fgets(cmd, 4095, stdin);
      cmd[strlen(cmd) - 1] = 0;
      printf("%s  ]\r", cmd);
      fflush(stdout);
      if (!strcmp(cmd, "mul2d")) {
	 ++mul2d_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 sscanf(buf, "%d", &rr);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);

	 mp_mul_2d(&a, rr, &a);
	 a.sign = b.sign;
	 if (mp_cmp(&a, &b) != MP_EQ) {
	    printf("mul2d failed, rr == %d\n", rr);
	    draw(&a);
	    draw(&b);
	    return 0;
	 }
      } else if (!strcmp(cmd, "div2d")) {
	 ++div2d_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 sscanf(buf, "%d", &rr);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);

	 mp_div_2d(&a, rr, &a, &e);
	 a.sign = b.sign;
	 if (a.used == b.used && a.used == 0) {
	    a.sign = b.sign = MP_ZPOS;
	 }
	 if (mp_cmp(&a, &b) != MP_EQ) {
	    printf("div2d failed, rr == %d\n", rr);
	    draw(&a);
	    draw(&b);
	    return 0;
	 }
      } else if (!strcmp(cmd, "add")) {
	 ++add_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_copy(&a, &d);
	 mp_add(&d, &b, &d);
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("add %lu failure!\n", add_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    return 0;
	 }

	 /* test the sign/unsigned storage functions */

	 rr = mp_signed_bin_size(&c);
	 mp_to_signed_bin(&c, (unsigned char *) cmd);
	 memset(cmd + rr, rand() & 255, sizeof(cmd) - rr);
	 mp_read_signed_bin(&d, (unsigned char *) cmd, rr);
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("mp_signed_bin failure!\n");
	    draw(&c);
	    draw(&d);
	    return 0;
	 }


	 rr = mp_unsigned_bin_size(&c);
	 mp_to_unsigned_bin(&c, (unsigned char *) cmd);
	 memset(cmd + rr, rand() & 255, sizeof(cmd) - rr);
	 mp_read_unsigned_bin(&d, (unsigned char *) cmd, rr);
	 if (mp_cmp_mag(&c, &d) != MP_EQ) {
	    printf("mp_unsigned_bin failure!\n");
	    draw(&c);
	    draw(&d);
	    return 0;
	 }

      } else if (!strcmp(cmd, "sub")) {
	 ++sub_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_copy(&a, &d);
	 mp_sub(&d, &b, &d);
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("sub %lu failure!\n", sub_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    return 0;
	 }
      } else if (!strcmp(cmd, "mul")) {
	 ++mul_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_copy(&a, &d);
	 mp_mul(&d, &b, &d);
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("mul %lu failure!\n", mul_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    return 0;
	 }
      } else if (!strcmp(cmd, "div")) {
	 ++div_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&d, buf, 64);

	 mp_div(&a, &b, &e, &f);
	 if (mp_cmp(&c, &e) != MP_EQ || mp_cmp(&d, &f) != MP_EQ) {
	    printf("div %lu %d, %d, failure!\n", div_n, mp_cmp(&c, &e),
		   mp_cmp(&d, &f));
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    draw(&e);
	    draw(&f);
	    return 0;
	 }

      } else if (!strcmp(cmd, "sqr")) {
	 ++sqr_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 mp_copy(&a, &c);
	 mp_sqr(&c, &c);
	 if (mp_cmp(&b, &c) != MP_EQ) {
	    printf("sqr %lu failure!\n", sqr_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    return 0;
	 }
      } else if (!strcmp(cmd, "gcd")) {
	 ++gcd_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_copy(&a, &d);
	 mp_gcd(&d, &b, &d);
	 d.sign = c.sign;
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("gcd %lu failure!\n", gcd_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    return 0;
	 }
      } else if (!strcmp(cmd, "lcm")) {
	 ++lcm_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_copy(&a, &d);
	 mp_lcm(&d, &b, &d);
	 d.sign = c.sign;
	 if (mp_cmp(&c, &d) != MP_EQ) {
	    printf("lcm %lu failure!\n", lcm_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    return 0;
	 }
      } else if (!strcmp(cmd, "expt")) {
	 ++expt_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&d, buf, 64);
	 mp_copy(&a, &e);
	 mp_exptmod(&e, &b, &c, &e);
	 if (mp_cmp(&d, &e) != MP_EQ) {
	    printf("expt %lu failure!\n", expt_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    draw(&e);
	    return 0;
	 }
      } else if (!strcmp(cmd, "invmod")) {
	 ++inv_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&c, buf, 64);
	 mp_invmod(&a, &b, &d);
	 mp_mulmod(&d, &a, &b, &e);
	 if (mp_cmp_d(&e, 1) != MP_EQ) {
	    printf("inv [wrong value from MPI?!] failure\n");
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    draw(&d);
	    mp_gcd(&a, &b, &e);
	    draw(&e);
	    return 0;
	 }

      } else if (!strcmp(cmd, "div2")) {
	 ++div2_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 mp_div_2(&a, &c);
	 if (mp_cmp(&c, &b) != MP_EQ) {
	    printf("div_2 %lu failure\n", div2_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    return 0;
	 }
      } else if (!strcmp(cmd, "mul2")) {
	 ++mul2_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 mp_mul_2(&a, &c);
	 if (mp_cmp(&c, &b) != MP_EQ) {
	    printf("mul_2 %lu failure\n", mul2_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    return 0;
	 }
      } else if (!strcmp(cmd, "add_d")) {
	 ++add_d_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 sscanf(buf, "%d", &ix);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 mp_add_d(&a, ix, &c);
	 if (mp_cmp(&b, &c) != MP_EQ) {
	    printf("add_d %lu failure\n", add_d_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    printf("d == %d\n", ix);
	    return 0;
	 }
      } else if (!strcmp(cmd, "sub_d")) {
	 ++sub_d_n;
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&a, buf, 64);
	 fgets(buf, 4095, stdin);
	 sscanf(buf, "%d", &ix);
	 fgets(buf, 4095, stdin);
	 mp_read_radix(&b, buf, 64);
	 mp_sub_d(&a, ix, &c);
	 if (mp_cmp(&b, &c) != MP_EQ) {
	    printf("sub_d %lu failure\n", sub_d_n);
	    draw(&a);
	    draw(&b);
	    draw(&c);
	    printf("d == %d\n", ix);
	    return 0;
	 }
      }
   }
   return 0;
}

#include <tommath.h>
#include <time.h>

ulong64 _tt;

#ifdef IOWNANATHLON
#include <unistd.h>
#define SLEEP sleep(4)
#else
#define SLEEP
#endif


void ndraw(mp_int * a, char *name)
{
   char buf[4096];

   printf("%s: ", name);
   mp_toradix(a, buf, 64);
   printf("%s\n", buf);
}

static void draw(mp_int * a)
{
   ndraw(a, "");
}


unsigned long lfsr = 0xAAAAAAAAUL;

int lbit(void)
{
   if (lfsr & 0x80000000UL) {
      lfsr = ((lfsr << 1) ^ 0x8000001BUL) & 0xFFFFFFFFUL;
      return 1;
   } else {
      lfsr <<= 1;
      return 0;
   }
}

/* RDTSC from Scott Duplichan */
static ulong64 TIMFUNC(void)
{
#if defined __GNUC__
#if defined(__i386__) || defined(__x86_64__)
   unsigned long long a;
   __asm__ __volatile__("rdtsc\nmovl %%eax,%0\nmovl %%edx,4+%0\n"::
			"m"(a):"%eax", "%edx");
   return a;
#else /* gcc-IA64 version */
   unsigned long result;
   __asm__ __volatile__("mov %0=ar.itc":"=r"(result)::"memory");

   while (__builtin_expect((int) result == -1, 0))
      __asm__ __volatile__("mov %0=ar.itc":"=r"(result)::"memory");

   return result;
#endif

   // Microsoft and Intel Windows compilers
#elif defined _M_IX86
   __asm rdtsc
#elif defined _M_AMD64
   return __rdtsc();
#elif defined _M_IA64
#if defined __INTEL_COMPILER
#include <ia64intrin.h>
#endif
   return __getReg(3116);
#else
#error need rdtsc function for this build
#endif
}

#define DO(x) x; x;
//#define DO4(x) DO2(x); DO2(x);
//#define DO8(x) DO4(x); DO4(x);
//#define DO(x)  DO8(x); DO8(x);

int main(void)
{
   ulong64 tt, gg, CLK_PER_SEC;
   FILE *log, *logb, *logc, *logd;
   mp_int a, b, c, d, e, f;
   int n, cnt, ix, old_kara_m, old_kara_s;
   unsigned rr;

   mp_init(&a);
   mp_init(&b);
   mp_init(&c);
   mp_init(&d);
   mp_init(&e);
   mp_init(&f);

   srand(time(NULL));


   /* temp. turn off TOOM */
   TOOM_MUL_CUTOFF = TOOM_SQR_CUTOFF = 100000;

   CLK_PER_SEC = TIMFUNC();
   sleep(1);
   CLK_PER_SEC = TIMFUNC() - CLK_PER_SEC;

   printf("CLK_PER_SEC == %llu\n", CLK_PER_SEC);
   goto exptmod;
   log = fopen("logs/add.log", "w");
   for (cnt = 8; cnt <= 128; cnt += 8) {
      SLEEP;
      mp_rand(&a, cnt);
      mp_rand(&b, cnt);
      rr = 0;
      tt = -1;
      do {
	 gg = TIMFUNC();
	 DO(mp_add(&a, &b, &c));
	 gg = (TIMFUNC() - gg) >> 1;
	 if (tt > gg)
	    tt = gg;
      } while (++rr < 100000);
      printf("Adding\t\t%4d-bit => %9llu/sec, %9llu cycles\n",
	     mp_count_bits(&a), CLK_PER_SEC / tt, tt);
      fprintf(log, "%d %9llu\n", cnt * DIGIT_BIT, tt);
      fflush(log);
   }
   fclose(log);

   log = fopen("logs/sub.log", "w");
   for (cnt = 8; cnt <= 128; cnt += 8) {
      SLEEP;
      mp_rand(&a, cnt);
      mp_rand(&b, cnt);
      rr = 0;
      tt = -1;
      do {
	 gg = TIMFUNC();
	 DO(mp_sub(&a, &b, &c));
	 gg = (TIMFUNC() - gg) >> 1;
	 if (tt > gg)
	    tt = gg;
      } while (++rr < 100000);

      printf("Subtracting\t\t%4d-bit => %9llu/sec, %9llu cycles\n",
	     mp_count_bits(&a), CLK_PER_SEC / tt, tt);
      fprintf(log, "%d %9llu\n", cnt * DIGIT_BIT, tt);
      fflush(log);
   }
   fclose(log);

   /* do mult/square twice, first without karatsuba and second with */
 multtest:
   old_kara_m = KARATSUBA_MUL_CUTOFF;
   old_kara_s = KARATSUBA_SQR_CUTOFF;
   for (ix = 0; ix < 2; ix++) {
      printf("With%s Karatsuba\n", (ix == 0) ? "out" : "");

      KARATSUBA_MUL_CUTOFF = (ix == 0) ? 9999 : old_kara_m;
      KARATSUBA_SQR_CUTOFF = (ix == 0) ? 9999 : old_kara_s;

      log = fopen((ix == 0) ? "logs/mult.log" : "logs/mult_kara.log", "w");
      for (cnt = 4; cnt <= 10240 / DIGIT_BIT; cnt += 2) {
	 SLEEP;
	 mp_rand(&a, cnt);
	 mp_rand(&b, cnt);
	 rr = 0;
	 tt = -1;
	 do {
	    gg = TIMFUNC();
	    DO(mp_mul(&a, &b, &c));
	    gg = (TIMFUNC() - gg) >> 1;
	    if (tt > gg)
	       tt = gg;
	 } while (++rr < 100);
	 printf("Multiplying\t%4d-bit => %9llu/sec, %9llu cycles\n",
		mp_count_bits(&a), CLK_PER_SEC / tt, tt);
	 fprintf(log, "%d %9llu\n", mp_count_bits(&a), tt);
	 fflush(log);
      }
      fclose(log);

      log = fopen((ix == 0) ? "logs/sqr.log" : "logs/sqr_kara.log", "w");
      for (cnt = 4; cnt <= 10240 / DIGIT_BIT; cnt += 2) {
	 SLEEP;
	 mp_rand(&a, cnt);
	 rr = 0;
	 tt = -1;
	 do {
	    gg = TIMFUNC();
	    DO(mp_sqr(&a, &b));
	    gg = (TIMFUNC() - gg) >> 1;
	    if (tt > gg)
	       tt = gg;
	 } while (++rr < 100);
	 printf("Squaring\t%4d-bit => %9llu/sec, %9llu cycles\n",
		mp_count_bits(&a), CLK_PER_SEC / tt, tt);
	 fprintf(log, "%d %9llu\n", mp_count_bits(&a), tt);
	 fflush(log);
      }
      fclose(log);

   }
 exptmod:

   {
      char *primes[] = {
	 /* 2K large moduli */
	 "179769313486231590772930519078902473361797697894230657273430081157732675805500963132708477322407536021120113879871393357658789768814416622492847430639474124377767893424865485276302219601246094119453082952085005768838150682342462881473913110540827237163350510684586239334100047359817950870678242457666208137217",
	 "32317006071311007300714876688669951960444102669715484032130345427524655138867890893197201411522913463688717960921898019494119559150490921095088152386448283120630877367300996091750197750389652106796057638384067568276792218642619756161838094338476170470581645852036305042887575891541065808607552399123930385521914333389668342420684974786564569494856176035326322058077805659331026192708460314150258592864177116725943603718461857357598351152301645904403697613233287231227125684710820209725157101726931323469678542580656697935045997268352998638099733077152121140120031150424541696791951097529546801429027668869927491725169",
	 "1044388881413152506691752710716624382579964249047383780384233483283953907971557456848826811934997558340890106714439262837987573438185793607263236087851365277945956976543709998340361590134383718314428070011855946226376318839397712745672334684344586617496807908705803704071284048740118609114467977783598029006686938976881787785946905630190260940599579453432823469303026696443059025015972399867714215541693835559885291486318237914434496734087811872639496475100189041349008417061675093668333850551032972088269550769983616369411933015213796825837188091833656751221318492846368125550225998300412344784862595674492194617023806505913245610825731835380087608622102834270197698202313169017678006675195485079921636419370285375124784014907159135459982790513399611551794271106831134090584272884279791554849782954323534517065223269061394905987693002122963395687782878948440616007412945674919823050571642377154816321380631045902916136926708342856440730447899971901781465763473223850267253059899795996090799469201774624817718449867455659250178329070473119433165550807568221846571746373296884912819520317457002440926616910874148385078411929804522981857338977648103126085902995208257421855249796721729039744118165938433694823325696642096892124547425283",
	 /* 2K moduli mersenne primes */
	 "6864797660130609714981900799081393217269435300143305409394463459185543183397656052122559640661454554977296311391480858037121987999716643812574028291115057151",
	 "531137992816767098689588206552468627329593117727031923199444138200403559860852242739162502265229285668889329486246501015346579337652707239409519978766587351943831270835393219031728127",
	 "10407932194664399081925240327364085538615262247266704805319112350403608059673360298012239441732324184842421613954281007791383566248323464908139906605677320762924129509389220345773183349661583550472959420547689811211693677147548478866962501384438260291732348885311160828538416585028255604666224831890918801847068222203140521026698435488732958028878050869736186900714720710555703168729087",
	 "1475979915214180235084898622737381736312066145333169775147771216478570297878078949377407337049389289382748507531496480477281264838760259191814463365330269540496961201113430156902396093989090226259326935025281409614983499388222831448598601834318536230923772641390209490231836446899608210795482963763094236630945410832793769905399982457186322944729636418890623372171723742105636440368218459649632948538696905872650486914434637457507280441823676813517852099348660847172579408422316678097670224011990280170474894487426924742108823536808485072502240519452587542875349976558572670229633962575212637477897785501552646522609988869914013540483809865681250419497686697771007",
	 "259117086013202627776246767922441530941818887553125427303974923161874019266586362086201209516800483406550695241733194177441689509238807017410377709597512042313066624082916353517952311186154862265604547691127595848775610568757931191017711408826252153849035830401185072116424747461823031471398340229288074545677907941037288235820705892351068433882986888616658650280927692080339605869308790500409503709875902119018371991620994002568935113136548829739112656797303241986517250116412703509705427773477972349821676443446668383119322540099648994051790241624056519054483690809616061625743042361721863339415852426431208737266591962061753535748892894599629195183082621860853400937932839420261866586142503251450773096274235376822938649407127700846077124211823080804139298087057504713825264571448379371125032081826126566649084251699453951887789613650248405739378594599444335231188280123660406262468609212150349937584782292237144339628858485938215738821232393687046160677362909315071",
	 "190797007524439073807468042969529173669356994749940177394741882673528979787005053706368049835514900244303495954950709725762186311224148828811920216904542206960744666169364221195289538436845390250168663932838805192055137154390912666527533007309292687539092257043362517857366624699975402375462954490293259233303137330643531556539739921926201438606439020075174723029056838272505051571967594608350063404495977660656269020823960825567012344189908927956646011998057988548630107637380993519826582389781888135705408653045219655801758081251164080554609057468028203308718724654081055323215860189611391296030471108443146745671967766308925858547271507311563765171008318248647110097614890313562856541784154881743146033909602737947385055355960331855614540900081456378659068370317267696980001187750995491090350108417050917991562167972281070161305972518044872048331306383715094854938415738549894606070722584737978176686422134354526989443028353644037187375385397838259511833166416134323695660367676897722287918773420968982326089026150031515424165462111337527431154890666327374921446276833564519776797633875503548665093914556482031482248883127023777039667707976559857333357013727342079099064400455741830654320379350833236245819348824064783585692924881021978332974949906122664421376034687815350484991",

	 /* DR moduli */
	 "14059105607947488696282932836518693308967803494693489478439861164411992439598399594747002144074658928593502845729752797260025831423419686528151609940203368612079",
	 "101745825697019260773923519755878567461315282017759829107608914364075275235254395622580447400994175578963163918967182013639660669771108475957692810857098847138903161308502419410142185759152435680068435915159402496058513611411688900243039",
	 "736335108039604595805923406147184530889923370574768772191969612422073040099331944991573923112581267542507986451953227192970402893063850485730703075899286013451337291468249027691733891486704001513279827771740183629161065194874727962517148100775228363421083691764065477590823919364012917984605619526140821797602431",
	 "38564998830736521417281865696453025806593491967131023221754800625044118265468851210705360385717536794615180260494208076605798671660719333199513807806252394423283413430106003596332513246682903994829528690198205120921557533726473585751382193953592127439965050261476810842071573684505878854588706623484573925925903505747545471088867712185004135201289273405614415899438276535626346098904241020877974002916168099951885406379295536200413493190419727789712076165162175783",
	 "542189391331696172661670440619180536749994166415993334151601745392193484590296600979602378676624808129613777993466242203025054573692562689251250471628358318743978285860720148446448885701001277560572526947619392551574490839286458454994488665744991822837769918095117129546414124448777033941223565831420390846864429504774477949153794689948747680362212954278693335653935890352619041936727463717926744868338358149568368643403037768649616778526013610493696186055899318268339432671541328195724261329606699831016666359440874843103020666106568222401047720269951530296879490444224546654729111504346660859907296364097126834834235287147",
	 "1487259134814709264092032648525971038895865645148901180585340454985524155135260217788758027400478312256339496385275012465661575576202252063145698732079880294664220579764848767704076761853197216563262660046602703973050798218246170835962005598561669706844469447435461092542265792444947706769615695252256130901271870341005768912974433684521436211263358097522726462083917939091760026658925757076733484173202927141441492573799914240222628795405623953109131594523623353044898339481494120112723445689647986475279242446083151413667587008191682564376412347964146113898565886683139407005941383669325997475076910488086663256335689181157957571445067490187939553165903773554290260531009121879044170766615232300936675369451260747671432073394867530820527479172464106442450727640226503746586340279816318821395210726268291535648506190714616083163403189943334431056876038286530365757187367147446004855912033137386225053275419626102417236133948503",
	 "1095121115716677802856811290392395128588168592409109494900178008967955253005183831872715423151551999734857184538199864469605657805519106717529655044054833197687459782636297255219742994736751541815269727940751860670268774903340296040006114013971309257028332849679096824800250742691718610670812374272414086863715763724622797509437062518082383056050144624962776302147890521249477060215148275163688301275847155316042279405557632639366066847442861422164832655874655824221577849928863023018366835675399949740429332468186340518172487073360822220449055340582568461568645259954873303616953776393853174845132081121976327462740354930744487429617202585015510744298530101547706821590188733515880733527449780963163909830077616357506845523215289297624086914545378511082534229620116563260168494523906566709418166011112754529766183554579321224940951177394088465596712620076240067370589036924024728375076210477267488679008016579588696191194060127319035195370137160936882402244399699172017835144537488486396906144217720028992863941288217185353914991583400421682751000603596655790990815525126154394344641336397793791497068253936771017031980867706707490224041075826337383538651825493679503771934836094655802776331664261631740148281763487765852746577808019633679",

	 /* generic unrestricted moduli */
	 "17933601194860113372237070562165128350027320072176844226673287945873370751245439587792371960615073855669274087805055507977323024886880985062002853331424203",
	 "2893527720709661239493896562339544088620375736490408468011883030469939904368086092336458298221245707898933583190713188177399401852627749210994595974791782790253946539043962213027074922559572312141181787434278708783207966459019479487",
	 "347743159439876626079252796797422223177535447388206607607181663903045907591201940478223621722118173270898487582987137708656414344685816179420855160986340457973820182883508387588163122354089264395604796675278966117567294812714812796820596564876450716066283126720010859041484786529056457896367683122960411136319",
	 "47266428956356393164697365098120418976400602706072312735924071745438532218237979333351774907308168340693326687317443721193266215155735814510792148768576498491199122744351399489453533553203833318691678263241941706256996197460424029012419012634671862283532342656309677173602509498417976091509154360039893165037637034737020327399910409885798185771003505320583967737293415979917317338985837385734747478364242020380416892056650841470869294527543597349250299539682430605173321029026555546832473048600327036845781970289288898317888427517364945316709081173840186150794397479045034008257793436817683392375274635794835245695887",
	 "436463808505957768574894870394349739623346440601945961161254440072143298152040105676491048248110146278752857839930515766167441407021501229924721335644557342265864606569000117714935185566842453630868849121480179691838399545644365571106757731317371758557990781880691336695584799313313687287468894148823761785582982549586183756806449017542622267874275103877481475534991201849912222670102069951687572917937634467778042874315463238062009202992087620963771759666448266532858079402669920025224220613419441069718482837399612644978839925207109870840278194042158748845445131729137117098529028886770063736487420613144045836803985635654192482395882603511950547826439092832800532152534003936926017612446606135655146445620623395788978726744728503058670046885876251527122350275750995227",
	 "11424167473351836398078306042624362277956429440521137061889702611766348760692206243140413411077394583180726863277012016602279290144126785129569474909173584789822341986742719230331946072730319555984484911716797058875905400999504305877245849119687509023232790273637466821052576859232452982061831009770786031785669030271542286603956118755585683996118896215213488875253101894663403069677745948305893849505434201763745232895780711972432011344857521691017896316861403206449421332243658855453435784006517202894181640562433575390821384210960117518650374602256601091379644034244332285065935413233557998331562749140202965844219336298970011513882564935538704289446968322281451907487362046511461221329799897350993370560697505809686438782036235372137015731304779072430260986460269894522159103008260495503005267165927542949439526272736586626709581721032189532726389643625590680105784844246152702670169304203783072275089194754889511973916207",
	 "1214855636816562637502584060163403830270705000634713483015101384881871978446801224798536155406895823305035467591632531067547890948695117172076954220727075688048751022421198712032848890056357845974246560748347918630050853933697792254955890439720297560693579400297062396904306270145886830719309296352765295712183040773146419022875165382778007040109957609739589875590885701126197906063620133954893216612678838507540777138437797705602453719559017633986486649523611975865005712371194067612263330335590526176087004421363598470302731349138773205901447704682181517904064735636518462452242791676541725292378925568296858010151852326316777511935037531017413910506921922450666933202278489024521263798482237150056835746454842662048692127173834433089016107854491097456725016327709663199738238442164843147132789153725513257167915555162094970853584447993125488607696008169807374736711297007473812256272245489405898470297178738029484459690836250560495461579533254473316340608217876781986188705928270735695752830825527963838355419762516246028680280988020401914551825487349990306976304093109384451438813251211051597392127491464898797406789175453067960072008590614886532333015881171367104445044718144312416815712216611576221546455968770801413440778423979",
	 NULL
      };
      log = fopen("logs/expt.log", "w");
      logb = fopen("logs/expt_dr.log", "w");
      logc = fopen("logs/expt_2k.log", "w");
      logd = fopen("logs/expt_2kl.log", "w");
      for (n = 0; primes[n]; n++) {
	 SLEEP;
	 mp_read_radix(&a, primes[n], 10);
	 mp_zero(&b);
	 for (rr = 0; rr < (unsigned) mp_count_bits(&a); rr++) {
	    mp_mul_2(&b, &b);
	    b.dp[0] |= lbit();
	    b.used += 1;
	 }
	 mp_sub_d(&a, 1, &c);
	 mp_mod(&b, &c, &b);
	 mp_set(&c, 3);
	 rr = 0;
	 tt = -1;
	 do {
	    gg = TIMFUNC();
	    DO(mp_exptmod(&c, &b, &a, &d));
	    gg = (TIMFUNC() - gg) >> 1;
	    if (tt > gg)
	       tt = gg;
	 } while (++rr < 10);
	 mp_sub_d(&a, 1, &e);
	 mp_sub(&e, &b, &b);
	 mp_exptmod(&c, &b, &a, &e);	/* c^(p-1-b) mod a */
	 mp_mulmod(&e, &d, &a, &d);	/* c^b * c^(p-1-b) == c^p-1 == 1 */
	 if (mp_cmp_d(&d, 1)) {
	    printf("Different (%d)!!!\n", mp_count_bits(&a));
	    draw(&d);
	    exit(0);
	 }
	 printf("Exponentiating\t%4d-bit => %9llu/sec, %9llu cycles\n",
		mp_count_bits(&a), CLK_PER_SEC / tt, tt);
	 fprintf(n < 4 ? logd : (n < 9) ? logc : (n < 16) ? logb : log,
		 "%d %9llu\n", mp_count_bits(&a), tt);
      }
   }
   fclose(log);
   fclose(logb);
   fclose(logc);
   fclose(logd);

   log = fopen("logs/invmod.log", "w");
   for (cnt = 4; cnt <= 128; cnt += 4) {
      SLEEP;
      mp_rand(&a, cnt);
      mp_rand(&b, cnt);

      do {
	 mp_add_d(&b, 1, &b);
	 mp_gcd(&a, &b, &c);
      } while (mp_cmp_d(&c, 1) != MP_EQ);

      rr = 0;
      tt = -1;
      do {
	 gg = TIMFUNC();
	 DO(mp_invmod(&b, &a, &c));
	 gg = (TIMFUNC() - gg) >> 1;
	 if (tt > gg)
	    tt = gg;
      } while (++rr < 1000);
      mp_mulmod(&b, &c, &a, &d);
      if (mp_cmp_d(&d, 1) != MP_EQ) {
	 printf("Failed to invert\n");
	 return 0;
      }
      printf("Inverting mod\t%4d-bit => %9llu/sec, %9llu cycles\n",
	     mp_count_bits(&a), CLK_PER_SEC / tt, tt);
      fprintf(log, "%d %9llu\n", cnt * DIGIT_BIT, tt);
   }
   fclose(log);

   return 0;
}

256-bits (k = 36113) = 115792089237316195423570985008687907853269984665640564039457584007913129603823
512-bits (k = 38117) = 13407807929942597099574024998205846127479365820592393377723561443721764030073546976801874298166903427690031858186486050853753882811946569946433649006045979

/* Makes safe primes of a 2k nature */
#include <tommath.h>
#include <time.h>

int sizes[] = {256, 512, 768, 1024, 1536, 2048, 3072, 4096};

int main(void)
{
   char buf[2000];
   int x, y;
   mp_int q, p;
   FILE *out;
   clock_t t1;
   mp_digit z;
   
   mp_init_multi(&q, &p, NULL);
   
   out = fopen("2kprime.1", "w");
   for (x = 0; x < (int)(sizeof(sizes) / sizeof(sizes[0])); x++) {
   top:
       mp_2expt(&q, sizes[x]);
       mp_add_d(&q, 3, &q);
       z = -3;
       
       t1 = clock();
       for(;;) {
         mp_sub_d(&q, 4, &q);
         z += 4;

         if (z > MP_MASK) {
            printf("No primes of size %d found\n", sizes[x]);
            break;
         }
         
         if (clock() - t1 > CLOCKS_PER_SEC) { 
            printf("."); fflush(stdout);
//            sleep((clock() - t1 + CLOCKS_PER_SEC/2)/CLOCKS_PER_SEC);
            t1 = clock();
         }
         
         /* quick test on q */
         mp_prime_is_prime(&q, 1, &y);
         if (y == 0) {
            continue;
         }

         /* find (q-1)/2 */
         mp_sub_d(&q, 1, &p);
         mp_div_2(&p, &p);
         mp_prime_is_prime(&p, 3, &y);
         if (y == 0) {
            continue;
         }

         /* test on q */
         mp_prime_is_prime(&q, 3, &y);
         if (y == 0) {
            continue;
         }

         break;
       }
       
       if (y == 0) {
          ++sizes[x];
          goto top;
       }
       
       mp_toradix(&q, buf, 10);
       printf("\n\n%d-bits (k = %lu) = %s\n", sizes[x], z, buf);
       fprintf(out, "%d-bits (k = %lu) = %s\n", sizes[x], z, buf); fflush(out);
   }
   
   return 0;
}   

/* Makes safe primes of a DR nature */
#include <tommath.h>

int sizes[] = { 1+256/DIGIT_BIT, 1+512/DIGIT_BIT, 1+768/DIGIT_BIT, 1+1024/DIGIT_BIT, 1+2048/DIGIT_BIT, 1+4096/DIGIT_BIT };
int main(void)
{
   int res, x, y;
   char buf[4096];
   FILE *out;
   mp_int a, b;
   
   mp_init(&a);
   mp_init(&b);
   
   out = fopen("drprimes.txt", "w");
   for (x = 0; x < (int)(sizeof(sizes)/sizeof(sizes[0])); x++) {
   top:
       printf("Seeking a %d-bit safe prime\n", sizes[x] * DIGIT_BIT);
       mp_grow(&a, sizes[x]);
       mp_zero(&a);
       for (y = 1; y < sizes[x]; y++) {
           a.dp[y] = MP_MASK;
       }
       
       /* make a DR modulus */
       a.dp[0] = -1;
       a.used = sizes[x];
       
       /* now loop */
       res = 0;
       for (;;) { 
          a.dp[0] += 4;
          if (a.dp[0] >= MP_MASK) break;
          mp_prime_is_prime(&a, 1, &res);
          if (res == 0) continue;
          printf("."); fflush(stdout);
          mp_sub_d(&a, 1, &b);
          mp_div_2(&b, &b);
          mp_prime_is_prime(&b, 3, &res);  
          if (res == 0) continue;
          mp_prime_is_prime(&a, 3, &res);
          if (res == 1) break;
	}
        
        if (res != 1) {
           printf("Error not DR modulus\n"); sizes[x] += 1; goto top;
        } else {
           mp_toradix(&a, buf, 10);
           printf("\n\np == %s\n\n", buf);
           fprintf(out, "%d-bit prime:\np == %s\n\n", mp_count_bits(&a), buf); fflush(out);
        }           
   }
   fclose(out);
   
   mp_clear(&a);
   mp_clear(&b);
   
   return 0;
}

DR safe primes for 28-bit digits.

224-bit prime:
p == 26959946667150639794667015087019630673637144422540572481103341844143

532-bit prime:
p == 14059105607947488696282932836518693308967803494693489478439861164411992439598399594747002144074658928593502845729752797260025831423419686528151609940203368691747

784-bit prime:
p == 101745825697019260773923519755878567461315282017759829107608914364075275235254395622580447400994175578963163918967182013639660669771108475957692810857098847138903161308502419410142185759152435680068435915159402496058513611411688900243039

1036-bit prime:
p == 736335108039604595805923406147184530889923370574768772191969612422073040099331944991573923112581267542507986451953227192970402893063850485730703075899286013451337291468249027691733891486704001513279827771740183629161065194874727962517148100775228363421083691764065477590823919364012917984605619526140821798437127

1540-bit prime:
p == 38564998830736521417281865696453025806593491967131023221754800625044118265468851210705360385717536794615180260494208076605798671660719333199513807806252394423283413430106003596332513246682903994829528690198205120921557533726473585751382193953592127439965050261476810842071573684505878854588706623484573925925903505747545471088867712185004135201289273405614415899438276535626346098904241020877974002916168099951885406379295536200413493190419727789712076165162175783

2072-bit prime:
p == 542189391331696172661670440619180536749994166415993334151601745392193484590296600979602378676624808129613777993466242203025054573692562689251250471628358318743978285860720148446448885701001277560572526947619392551574490839286458454994488665744991822837769918095117129546414124448777033941223565831420390846864429504774477949153794689948747680362212954278693335653935890352619041936727463717926744868338358149568368643403037768649616778526013610493696186055899318268339432671541328195724261329606699831016666359440874843103020666106568222401047720269951530296879490444224546654729111504346660859907296364097126834834235287147

3080-bit prime:
p == 1487259134814709264092032648525971038895865645148901180585340454985524155135260217788758027400478312256339496385275012465661575576202252063145698732079880294664220579764848767704076761853197216563262660046602703973050798218246170835962005598561669706844469447435461092542265792444947706769615695252256130901271870341005768912974433684521436211263358097522726462083917939091760026658925757076733484173202927141441492573799914240222628795405623953109131594523623353044898339481494120112723445689647986475279242446083151413667587008191682564376412347964146113898565886683139407005941383669325997475076910488086663256335689181157957571445067490187939553165903773554290260531009121879044170766615232300936675369451260747671432073394867530820527479172464106442450727640226503746586340279816318821395210726268291535648506190714616083163403189943334431056876038286530365757187367147446004855912033137386225053275419626102417236133948503

4116-bit prime:
p == 1095121115716677802856811290392395128588168592409109494900178008967955253005183831872715423151551999734857184538199864469605657805519106717529655044054833197687459782636297255219742994736751541815269727940751860670268774903340296040006114013971309257028332849679096824800250742691718610670812374272414086863715763724622797509437062518082383056050144624962776302147890521249477060215148275163688301275847155316042279405557632639366066847442861422164832655874655824221577849928863023018366835675399949740429332468186340518172487073360822220449055340582568461568645259954873303616953776393853174845132081121976327462740354930744487429617202585015510744298530101547706821590188733515880733527449780963163909830077616357506845523215289297624086914545378511082534229620116563260168494523906566709418166011112754529766183554579321224940951177394088465596712620076240067370589036924024728375076210477267488679008016579588696191194060127319035195370137160936882402244399699172017835144537488486396906144217720028992863941288217185353914991583400421682751000603596655790990815525126154394344641336397793791497068253936771017031980867706707490224041075826337383538651825493679503771934836094655802776331664261631740148281763487765852746577808019633679

300-bit prime:
p == 2037035976334486086268445688409378161051468393665936250636140449354381298610415201576637819

540-bit prime:
p == 3599131035634557106248430806148785487095757694641533306480604458089470064537190296255232548883112685719936728506816716098566612844395439751206810991770626477344739

780-bit prime:
p == 6359114106063703798370219984742410466332205126109989319225557147754704702203399726411277962562135973685197744935448875852478791860694279747355800678568677946181447581781401213133886609947027230004277244697462656003655947791725966271167

CFLAGS += -Wall -W -Wshadow -O3 -fomit-frame-pointer -funroll-loops -I../

# default lib name (requires install with root)
# LIBNAME=-ltommath

# libname when you can't install the lib with install
LIBNAME=../libtommath.a

#provable primes
pprime: pprime.o
	$(CC) pprime.o $(LIBNAME) -o pprime

# portable [well requires clock()] tuning app
tune: tune.o
	$(CC) tune.o $(LIBNAME) -o tune
	
# same app but using RDTSC for higher precision [requires 80586+], coff based gcc installs [e.g. ming, cygwin, djgpp]
tune86: tune.c
	nasm -f coff timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86
	
# for cygwin
tune86c: tune.c
	nasm -f gnuwin32 timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86

#make tune86 for linux or any ELF format
tune86l: tune.c
	nasm -f elf -DUSE_ELF timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o $(LIBNAME) -o tune86l
        
# spits out mersenne primes
mersenne: mersenne.o
	$(CC) mersenne.o $(LIBNAME) -o mersenne

# fines DR safe primes for the given config
drprime: drprime.o
	$(CC) drprime.o $(LIBNAME) -o drprime
	
# fines 2k safe primes for the given config
2kprime: 2kprime.o
	$(CC) 2kprime.o $(LIBNAME) -o 2kprime

mont: mont.o
	$(CC) mont.o $(LIBNAME) -o mont

        
clean:
	rm -f *.log *.o *.obj *.exe pprime tune mersenne drprime tune86 tune86l mont 2kprime pprime.dat \
         *.da *.dyn *.dpi *~

CC = icc

CFLAGS += -I../

# optimize for SPEED
#
# -mcpu= can be pentium, pentiumpro (covers PII through PIII) or pentium4
# -ax?   specifies make code specifically for ? but compatible with IA-32
# -x?    specifies compile solely for ? [not specifically IA-32 compatible]
#
# where ? is 
#   K - PIII
#   W - first P4 [Williamette]
#   N - P4 Northwood
#   P - P4 Prescott
#   B - Blend of P4 and PM [mobile]
#
# Default to just generic max opts
CFLAGS += -O3 -xP -ip

# default lib name (requires install with root)
# LIBNAME=-ltommath

# libname when you can't install the lib with install
LIBNAME=../libtommath.a

#provable primes
pprime: pprime.o
	$(CC) pprime.o $(LIBNAME) -o pprime

# portable [well requires clock()] tuning app
tune: tune.o
	$(CC) tune.o $(LIBNAME) -o tune
	
# same app but using RDTSC for higher precision [requires 80586+], coff based gcc installs [e.g. ming, cygwin, djgpp]
tune86: tune.c
	nasm -f coff timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86
	
# for cygwin
tune86c: tune.c
	nasm -f gnuwin32 timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86

#make tune86 for linux or any ELF format
tune86l: tune.c
	nasm -f elf -DUSE_ELF timer.asm
	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o $(LIBNAME) -o tune86l
        
# spits out mersenne primes
mersenne: mersenne.o
	$(CC) mersenne.o $(LIBNAME) -o mersenne

# fines DR safe primes for the given config
drprime: drprime.o
	$(CC) drprime.o $(LIBNAME) -o drprime
	
# fines 2k safe primes for the given config
2kprime: 2kprime.o
	$(CC) 2kprime.o $(LIBNAME) -o 2kprime

mont: mont.o
	$(CC) mont.o $(LIBNAME) -o mont

        
clean:
	rm -f *.log *.o *.obj *.exe pprime tune mersenne drprime tune86 tune86l mont 2kprime pprime.dat *.il

#MSVC Makefile
#
#Tom St Denis

CFLAGS = /I../ /Ox /DWIN32 /W3

pprime: pprime.obj
	cl pprime.obj ../tommath.lib 

mersenne: mersenne.obj
	cl mersenne.obj ../tommath.lib
	
tune: tune.obj
	cl tune.obj ../tommath.lib

mont: mont.obj
	cl mont.obj ../tommath.lib
	
drprime: drprime.obj
	cl drprime.obj ../tommath.lib

2kprime: 2kprime.obj
	cl 2kprime.obj ../tommath.lib

/* Finds Mersenne primes using the Lucas-Lehmer test 
 *
 * Tom St Denis, [email protected]
 */
#include <time.h>
#include <tommath.h>

int
is_mersenne (long s, int *pp)
{
  mp_int  n, u;
  int     res, k;
  
  *pp = 0;

  if ((res = mp_init (&n)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init (&u)) != MP_OKAY) {
    goto LBL_N;
  }

  /* n = 2^s - 1 */
  if ((res = mp_2expt(&n, s)) != MP_OKAY) {
     goto LBL_MU;
  }
  if ((res = mp_sub_d (&n, 1, &n)) != MP_OKAY) {
    goto LBL_MU;
  }

  /* set u=4 */
  mp_set (&u, 4);

  /* for k=1 to s-2 do */
  for (k = 1; k <= s - 2; k++) {
    /* u = u^2 - 2 mod n */
    if ((res = mp_sqr (&u, &u)) != MP_OKAY) {
      goto LBL_MU;
    }
    if ((res = mp_sub_d (&u, 2, &u)) != MP_OKAY) {
      goto LBL_MU;
    }

    /* make sure u is positive */
    while (u.sign == MP_NEG) {
      if ((res = mp_add (&u, &n, &u)) != MP_OKAY) {
         goto LBL_MU;
      }
    }

    /* reduce */
    if ((res = mp_reduce_2k (&u, &n, 1)) != MP_OKAY) {
      goto LBL_MU;
    }
  }

  /* if u == 0 then its prime */
  if (mp_iszero (&u) == 1) {
    mp_prime_is_prime(&n, 8, pp);
  if (*pp != 1) printf("FAILURE\n");
  }

  res = MP_OKAY;
LBL_MU:mp_clear (&u);
LBL_N:mp_clear (&n);
  return res;
}

/* square root of a long < 65536 */
long
i_sqrt (long x)
{
  long    x1, x2;

  x2 = 16;
  do {
    x1 = x2;
    x2 = x1 - ((x1 * x1) - x) / (2 * x1);
  } while (x1 != x2);

  if (x1 * x1 > x) {
    --x1;
  }

  return x1;
}

/* is the long prime by brute force */
int
isprime (long k)
{
  long    y, z;

  y = i_sqrt (k);
  for (z = 2; z <= y; z++) {
    if ((k % z) == 0)
      return 0;
  }
  return 1;
}


int
main (void)
{
  int     pp;
  long    k;
  clock_t tt;

  k = 3;

  for (;;) {
    /* start time */
    tt = clock ();

    /* test if 2^k - 1 is prime */
    if (is_mersenne (k, &pp) != MP_OKAY) {
      printf ("Whoa error\n");
      return -1;
    }

    if (pp == 1) {
      /* count time */
      tt = clock () - tt;

      /* display if prime */
      printf ("2^%-5ld - 1 is prime, test took %ld ticks\n", k, tt);
    }

    /* goto next odd exponent */
    k += 2;

    /* but make sure its prime */
    while (isprime (k) == 0) {
      k += 2;
    }
  }
  return 0;
}

/* tests the montgomery routines */
#include <tommath.h>

int main(void)
{
   mp_int modulus, R, p, pp;
   mp_digit mp;
   long x, y;

   srand(time(NULL));
   mp_init_multi(&modulus, &R, &p, &pp, NULL);

   /* loop through various sizes */
   for (x = 4; x < 256; x++) {
       printf("DIGITS == %3ld...", x); fflush(stdout);
       
       /* make up the odd modulus */
       mp_rand(&modulus, x);
       modulus.dp[0] |= 1;
       
       /* now find the R value */
       mp_montgomery_calc_normalization(&R, &modulus);
       mp_montgomery_setup(&modulus, &mp);
       
       /* now run through a bunch tests */
       for (y = 0; y < 1000; y++) {
           mp_rand(&p, x/2);        /* p = random */
           mp_mul(&p, &R, &pp);     /* pp = R * p */
           mp_montgomery_reduce(&pp, &modulus, mp);
           
           /* should be equal to p */
           if (mp_cmp(&pp, &p) != MP_EQ) {
              printf("FAILURE!\n");
              exit(-1);
           }
       }
       printf("PASSED\n");
    }
    
    return 0;
}

/* Generates provable primes
 *
 * See http://gmail.com:8080/papers/pp.pdf for more info.
 *
 * Tom St Denis, [email protected], http://tom.gmail.com
 */
#include <time.h>
#include "tommath.h"

int   n_prime;
FILE *primes;

/* fast square root */
static  mp_digit
i_sqrt (mp_word x)
{
  mp_word x1, x2;

  x2 = x;
  do {
    x1 = x2;
    x2 = x1 - ((x1 * x1) - x) / (2 * x1);
  } while (x1 != x2);

  if (x1 * x1 > x) {
    --x1;
  }

  return x1;
}


/* generates a prime digit */
static void gen_prime (void)
{
  mp_digit r, x, y, next;
  FILE *out;

  out = fopen("pprime.dat", "wb");

  /* write first set of primes */
  r = 3; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 5; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 7; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 11; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 13; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 17; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 19; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 23; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 29; fwrite(&r, 1, sizeof(mp_digit), out);
  r = 31; fwrite(&r, 1, sizeof(mp_digit), out);

  /* get square root, since if 'r' is composite its factors must be < than this */
  y = i_sqrt (r);
  next = (y + 1) * (y + 1);

  for (;;) {
  do {
    r += 2;			/* next candidate */
    r &= MP_MASK;
    if (r < 31) break;

    /* update sqrt ? */
    if (next <= r) {
      ++y;
      next = (y + 1) * (y + 1);
    }

    /* loop if divisible by 3,5,7,11,13,17,19,23,29  */
    if ((r % 3) == 0) {
      x = 0;
      continue;
    }
    if ((r % 5) == 0) {
      x = 0;
      continue;
    }
    if ((r % 7) == 0) {
      x = 0;
      continue;
    }
    if ((r % 11) == 0) {
      x = 0;
      continue;
    }
    if ((r % 13) == 0) {
      x = 0;
      continue;
    }
    if ((r % 17) == 0) {
      x = 0;
      continue;
    }
    if ((r % 19) == 0) {
      x = 0;
      continue;
    }
    if ((r % 23) == 0) {
      x = 0;
      continue;
    }
    if ((r % 29) == 0) {
      x = 0;
      continue;
    }

    /* now check if r is divisible by x + k={1,7,11,13,17,19,23,29} */
    for (x = 30; x <= y; x += 30) {
      if ((r % (x + 1)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 7)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 11)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 13)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 17)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 19)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 23)) == 0) {
	x = 0;
	break;
      }
      if ((r % (x + 29)) == 0) {
	x = 0;
	break;
      }
    }
  } while (x == 0);
  if (r > 31) { fwrite(&r, 1, sizeof(mp_digit), out); printf("%9d\r", r); fflush(stdout); }
  if (r < 31) break;
  }

  fclose(out);
}

void load_tab(void)
{
   primes = fopen("pprime.dat", "rb");
   if (primes == NULL) {
      gen_prime();
      primes = fopen("pprime.dat", "rb");
   }
   fseek(primes, 0, SEEK_END);
   n_prime = ftell(primes) / sizeof(mp_digit);
}

mp_digit prime_digit(void)
{
   int n;
   mp_digit d;

   n = abs(rand()) % n_prime;
   fseek(primes, n * sizeof(mp_digit), SEEK_SET);
   fread(&d, 1, sizeof(mp_digit), primes);
   return d;
}


/* makes a prime of at least k bits */
int
pprime (int k, int li, mp_int * p, mp_int * q)
{
  mp_int  a, b, c, n, x, y, z, v;
  int     res, ii;
  static const mp_digit bases[] = { 2, 3, 5, 7, 11, 13, 17, 19 };

  /* single digit ? */
  if (k <= (int) DIGIT_BIT) {
    mp_set (p, prime_digit ());
    return MP_OKAY;
  }

  if ((res = mp_init (&c)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init (&v)) != MP_OKAY) {
    goto LBL_C;
  }

  /* product of first 50 primes */
  if ((res =
       mp_read_radix (&v,
		      "19078266889580195013601891820992757757219839668357012055907516904309700014933909014729740190",
		      10)) != MP_OKAY) {
    goto LBL_V;
  }

  if ((res = mp_init (&a)) != MP_OKAY) {
    goto LBL_V;
  }

  /* set the prime */
  mp_set (&a, prime_digit ());

  if ((res = mp_init (&b)) != MP_OKAY) {
    goto LBL_A;
  }

  if ((res = mp_init (&n)) != MP_OKAY) {
    goto LBL_B;
  }

  if ((res = mp_init (&x)) != MP_OKAY) {
    goto LBL_N;
  }

  if ((res = mp_init (&y)) != MP_OKAY) {
    goto LBL_X;
  }

  if ((res = mp_init (&z)) != MP_OKAY) {
    goto LBL_Y;
  }

  /* now loop making the single digit */
  while (mp_count_bits (&a) < k) {
    fprintf (stderr, "prime has %4d bits left\r", k - mp_count_bits (&a));
    fflush (stderr);
  top:
    mp_set (&b, prime_digit ());

    /* now compute z = a * b * 2 */
    if ((res = mp_mul (&a, &b, &z)) != MP_OKAY) {	/* z = a * b */
      goto LBL_Z;
    }

    if ((res = mp_copy (&z, &c)) != MP_OKAY) {	/* c = a * b */
      goto LBL_Z;
    }

    if ((res = mp_mul_2 (&z, &z)) != MP_OKAY) {	/* z = 2 * a * b */
      goto LBL_Z;
    }

    /* n = z + 1 */
    if ((res = mp_add_d (&z, 1, &n)) != MP_OKAY) {	/* n = z + 1 */
      goto LBL_Z;
    }

    /* check (n, v) == 1 */
    if ((res = mp_gcd (&n, &v, &y)) != MP_OKAY) {	/* y = (n, v) */
      goto LBL_Z;
    }

    if (mp_cmp_d (&y, 1) != MP_EQ)
      goto top;

    /* now try base x=bases[ii]  */
    for (ii = 0; ii < li; ii++) {
      mp_set (&x, bases[ii]);

      /* compute x^a mod n */
      if ((res = mp_exptmod (&x, &a, &n, &y)) != MP_OKAY) {	/* y = x^a mod n */
	goto LBL_Z;
      }

      /* if y == 1 loop */
      if (mp_cmp_d (&y, 1) == MP_EQ)
	continue;

      /* now x^2a mod n */
      if ((res = mp_sqrmod (&y, &n, &y)) != MP_OKAY) {	/* y = x^2a mod n */
	goto LBL_Z;
      }

      if (mp_cmp_d (&y, 1) == MP_EQ)
	continue;

      /* compute x^b mod n */
      if ((res = mp_exptmod (&x, &b, &n, &y)) != MP_OKAY) {	/* y = x^b mod n */
	goto LBL_Z;
      }

      /* if y == 1 loop */
      if (mp_cmp_d (&y, 1) == MP_EQ)
	continue;

      /* now x^2b mod n */
      if ((res = mp_sqrmod (&y, &n, &y)) != MP_OKAY) {	/* y = x^2b mod n */
	goto LBL_Z;
      }

      if (mp_cmp_d (&y, 1) == MP_EQ)
	continue;

      /* compute x^c mod n == x^ab mod n */
      if ((res = mp_exptmod (&x, &c, &n, &y)) != MP_OKAY) {	/* y = x^ab mod n */
	goto LBL_Z;
      }

      /* if y == 1 loop */
      if (mp_cmp_d (&y, 1) == MP_EQ)
	continue;

      /* now compute (x^c mod n)^2 */
      if ((res = mp_sqrmod (&y, &n, &y)) != MP_OKAY) {	/* y = x^2ab mod n */
	goto LBL_Z;
      }

      /* y should be 1 */
      if (mp_cmp_d (&y, 1) != MP_EQ)
	continue;
      break;
    }

    /* no bases worked? */
    if (ii == li)
      goto top;

{
   char buf[4096];

   mp_toradix(&n, buf, 10);
   printf("Certificate of primality for:\n%s\n\n", buf);
   mp_toradix(&a, buf, 10);
   printf("A == \n%s\n\n", buf);
   mp_toradix(&b, buf, 10);
   printf("B == \n%s\n\nG == %d\n", buf, bases[ii]);
   printf("----------------------------------------------------------------\n");
}

    /* a = n */
    mp_copy (&n, &a);
  }

  /* get q to be the order of the large prime subgroup */
  mp_sub_d (&n, 1, q);
  mp_div_2 (q, q);
  mp_div (q, &b, q, NULL);

  mp_exch (&n, p);

  res = MP_OKAY;
LBL_Z:mp_clear (&z);
LBL_Y:mp_clear (&y);
LBL_X:mp_clear (&x);
LBL_N:mp_clear (&n);
LBL_B:mp_clear (&b);
LBL_A:mp_clear (&a);
LBL_V:mp_clear (&v);
LBL_C:mp_clear (&c);
  return res;
}


int
main (void)
{
  mp_int  p, q;
  char    buf[4096];
  int     k, li;
  clock_t t1;

  srand (time (NULL));
  load_tab();

  printf ("Enter # of bits: \n");
  fgets (buf, sizeof (buf), stdin);
  sscanf (buf, "%d", &k);

  printf ("Enter number of bases to try (1 to 8):\n");
  fgets (buf, sizeof (buf), stdin);
  sscanf (buf, "%d", &li);


  mp_init (&p);
  mp_init (&q);

  t1 = clock ();
  pprime (k, li, &p, &q);
  t1 = clock () - t1;

  printf ("\n\nTook %ld ticks, %d bits\n", t1, mp_count_bits (&p));

  mp_toradix (&p, buf, 10);
  printf ("P == %s\n", buf);
  mp_toradix (&q, buf, 10);
  printf ("Q == %s\n", buf);

  return 0;
}

Enter # of bits: 
Enter number of bases to try (1 to 8):
Certificate of primality for:
36360080703173363

A == 
89963569

B == 
202082249

G == 2
----------------------------------------------------------------
Certificate of primality for:
4851595597739856136987139

A == 
36360080703173363

B == 
66715963

G == 2
----------------------------------------------------------------
Certificate of primality for:
19550639734462621430325731591027

A == 
4851595597739856136987139

B == 
2014867

G == 2
----------------------------------------------------------------
Certificate of primality for:
10409036141344317165691858509923818734539

A == 
19550639734462621430325731591027

B == 
266207047

G == 2
----------------------------------------------------------------
Certificate of primality for:
1049829549988285012736475602118094726647504414203

A == 
10409036141344317165691858509923818734539

B == 
50428759

G == 2
----------------------------------------------------------------
Certificate of primality for:
77194737385528288387712399596835459931920358844586615003

A == 
1049829549988285012736475602118094726647504414203

B == 
36765367

G == 2
----------------------------------------------------------------
Certificate of primality for:
35663756695365208574443215955488689578374232732893628896541201763

A == 
77194737385528288387712399596835459931920358844586615003

B == 
230998627

G == 2
----------------------------------------------------------------
Certificate of primality for:
16711831463502165169495622246023119698415848120292671294127567620396469803

A == 
35663756695365208574443215955488689578374232732893628896541201763

B == 
234297127

G == 2
----------------------------------------------------------------
Certificate of primality for:
6163534781560285962890718925972249753147470953579266394395432475622345597103528739

A == 
16711831463502165169495622246023119698415848120292671294127567620396469803

B == 
184406323

G == 2
----------------------------------------------------------------
Certificate of primality for:
814258256205243497704094951432575867360065658372158511036259934640748088306764553488803787

A == 
6163534781560285962890718925972249753147470953579266394395432475622345597103528739

B == 
66054487

G == 2
----------------------------------------------------------------
Certificate of primality for:
176469695533271657902814176811660357049007467856432383037590673407330246967781451723764079581998187

A == 
814258256205243497704094951432575867360065658372158511036259934640748088306764553488803787

B == 
108362239

G == 2
----------------------------------------------------------------
Certificate of primality for:
44924492859445516541759485198544012102424796403707253610035148063863073596051272171194806669756971406400419

A == 
176469695533271657902814176811660357049007467856432383037590673407330246967781451723764079581998187

B == 
127286707

G == 2
----------------------------------------------------------------
Certificate of primality for:
20600996927219343383225424320134474929609459588323857796871086845924186191561749519858600696159932468024710985371059

A == 
44924492859445516541759485198544012102424796403707253610035148063863073596051272171194806669756971406400419

B == 
229284691

G == 2
----------------------------------------------------------------
Certificate of primality for:
6295696427695493110141186605837397185848992307978456138112526915330347715236378041486547994708748840844217371233735072572979

A == 
20600996927219343383225424320134474929609459588323857796871086845924186191561749519858600696159932468024710985371059

B == 
152800771

G == 2
----------------------------------------------------------------
Certificate of primality for:
3104984078042317488749073016454213579257792635142218294052134804187631661145261015102617582090263808696699966840735333252107678792123

A == 
6295696427695493110141186605837397185848992307978456138112526915330347715236378041486547994708748840844217371233735072572979

B == 
246595759

G == 2
----------------------------------------------------------------
Certificate of primality for:
26405175827665701256325699315126705508919255051121452292124404943796947287968603975320562847910946802396632302209435206627913466015741799499

A == 
3104984078042317488749073016454213579257792635142218294052134804187631661145261015102617582090263808696699966840735333252107678792123

B == 
4252063

G == 2
----------------------------------------------------------------
Certificate of primality for:
11122146237908413610034600609460545703591095894418599759742741406628055069007082998134905595800236452010905900391505454890446585211975124558601770163

A == 
26405175827665701256325699315126705508919255051121452292124404943796947287968603975320562847910946802396632302209435206627913466015741799499

B == 
210605419

G == 2
----------------------------------------------------------------
Certificate of primality for:
1649861642047798890580354082088712649911849362201343649289384923147797960364736011515757482030049342943790127685185806092659832129486307035500638595572396187

A == 
11122146237908413610034600609460545703591095894418599759742741406628055069007082998134905595800236452010905900391505454890446585211975124558601770163

B == 
74170111

G == 2
----------------------------------------------------------------
Certificate of primality for:
857983367126266717607389719637086684134462613006415859877666235955788392464081914127715967940968197765042399904117392707518175220864852816390004264107201177394565363

A == 
1649861642047798890580354082088712649911849362201343649289384923147797960364736011515757482030049342943790127685185806092659832129486307035500638595572396187

B == 
260016763

G == 2
----------------------------------------------------------------
Certificate of primality for:
175995909353623703257072120479340610010337144085688850745292031336724691277374210929188442230237711063783727092685448718515661641054886101716698390145283196296702450566161283

A == 
857983367126266717607389719637086684134462613006415859877666235955788392464081914127715967940968197765042399904117392707518175220864852816390004264107201177394565363

B == 
102563707

G == 2
----------------------------------------------------------------
Certificate of primality for:
48486002551155667224487059713350447239190772068092630563272168418880661006593537218144160068395218642353495339720640699721703003648144463556291315694787862009052641640656933232794283

A == 
175995909353623703257072120479340610010337144085688850745292031336724691277374210929188442230237711063783727092685448718515661641054886101716698390145283196296702450566161283

B == 
137747527

G == 2
----------------------------------------------------------------
Certificate of primality for:
13156468011529105025061495011938518171328604045212410096476697450506055664012861932372156505805788068791146986282263016790631108386790291275939575123375304599622623328517354163964228279867403

A == 
48486002551155667224487059713350447239190772068092630563272168418880661006593537218144160068395218642353495339720640699721703003648144463556291315694787862009052641640656933232794283

B == 
135672847

G == 2
----------------------------------------------------------------
Certificate of primality for:
6355194692790533601105154341731997464407930009404822926832136060319955058388106456084549316415200519472481147942263916585428906582726749131479465958107142228236909665306781538860053107680830113869123

A == 
13156468011529105025061495011938518171328604045212410096476697450506055664012861932372156505805788068791146986282263016790631108386790291275939575123375304599622623328517354163964228279867403

B == 
241523587

G == 2
----------------------------------------------------------------
Certificate of primality for:
3157116676535430302794438027544146642863331358530722860333745617571010460905857862561870488000265751138954271040017454405707755458702044884023184574412221802502351503929935224995314581932097706874819348858083

A == 
6355194692790533601105154341731997464407930009404822926832136060319955058388106456084549316415200519472481147942263916585428906582726749131479465958107142228236909665306781538860053107680830113869123

B == 
248388667

G == 2
----------------------------------------------------------------
Certificate of primality for:
390533129219992506725320633489467713907837370444962163378727819939092929448752905310115311180032249230394348337568973177802874166228132778126338883671958897238722734394783244237133367055422297736215754829839364158067

A == 
3157116676535430302794438027544146642863331358530722860333745617571010460905857862561870488000265751138954271040017454405707755458702044884023184574412221802502351503929935224995314581932097706874819348858083

B == 
61849651

G == 2
----------------------------------------------------------------
Certificate of primality for:
48583654555070224891047847050732516652910250240135992225139515777200432486685999462997073444468380434359929499498804723793106565291183220444221080449740542884172281158126259373095216435009661050109711341419005972852770440739

A == 
390533129219992506725320633489467713907837370444962163378727819939092929448752905310115311180032249230394348337568973177802874166228132778126338883671958897238722734394783244237133367055422297736215754829839364158067

B == 
62201707

G == 2
----------------------------------------------------------------
Certificate of primality for:
25733035251905120039135866524384525138869748427727001128764704499071378939227862068500633813538831598776578372709963673670934388213622433800015759585470542686333039614931682098922935087822950084908715298627996115185849260703525317419

A == 
48583654555070224891047847050732516652910250240135992225139515777200432486685999462997073444468380434359929499498804723793106565291183220444221080449740542884172281158126259373095216435009661050109711341419005972852770440739

B == 
264832231

G == 2
----------------------------------------------------------------
Certificate of primality for:
2804594464939948901906623499531073917980499195397462605359913717827014360538186518540781517129548650937632008683280555602633122170458773895504894807182664540529077836857897972175530148107545939211339044386106111633510166695386323426241809387

A == 
25733035251905120039135866524384525138869748427727001128764704499071378939227862068500633813538831598776578372709963673670934388213622433800015759585470542686333039614931682098922935087822950084908715298627996115185849260703525317419

B == 
54494047

G == 2
----------------------------------------------------------------
Certificate of primality for:
738136612083433720096707308165797114449914259256979340471077690416567237592465306112484843530074782721390528773594351482384711900456440808251196845265132086486672447136822046628407467459921823150600138073268385534588238548865012638209515923513516547

A == 
2804594464939948901906623499531073917980499195397462605359913717827014360538186518540781517129548650937632008683280555602633122170458773895504894807182664540529077836857897972175530148107545939211339044386106111633510166695386323426241809387

B == 
131594179

G == 2
----------------------------------------------------------------
Certificate of primality for:
392847529056126766528615419937165193421166694172790666626558750047057558168124866940509180171236517681470100877687445134633784815352076138790217228749332398026714192707447855731679485746120589851992221508292976900578299504461333767437280988393026452846013683

A == 
738136612083433720096707308165797114449914259256979340471077690416567237592465306112484843530074782721390528773594351482384711900456440808251196845265132086486672447136822046628407467459921823150600138073268385534588238548865012638209515923513516547

B == 
266107603

G == 2
----------------------------------------------------------------
Certificate of primality for:
168459393231883505975876919268398655632763956627405508859662408056221544310200546265681845397346956580604208064328814319465940958080244889692368602591598503944015835190587740756859842792554282496742843600573336023639256008687581291233481455395123454655488735304365627

A == 
392847529056126766528615419937165193421166694172790666626558750047057558168124866940509180171236517681470100877687445134633784815352076138790217228749332398026714192707447855731679485746120589851992221508292976900578299504461333767437280988393026452846013683

B == 
214408111

G == 2
----------------------------------------------------------------
Certificate of primality for:
14865774288636941404884923981945833072113667565310054952177860608355263252462409554658728941191929400198053290113492910272458441655458514080123870132092365833472436407455910185221474386718838138135065780840839893113912689594815485706154461164071775481134379794909690501684643

A == 
168459393231883505975876919268398655632763956627405508859662408056221544310200546265681845397346956580604208064328814319465940958080244889692368602591598503944015835190587740756859842792554282496742843600573336023639256008687581291233481455395123454655488735304365627

B == 
44122723

G == 2
----------------------------------------------------------------
Certificate of primality for:
1213301773203241614897109856134894783021668292000023984098824423682568173639394290886185366993108292039068940333907505157813934962357206131450244004178619265868614859794316361031904412926604138893775068853175215502104744339658944443630407632290152772487455298652998368296998719996019

A == 
14865774288636941404884923981945833072113667565310054952177860608355263252462409554658728941191929400198053290113492910272458441655458514080123870132092365833472436407455910185221474386718838138135065780840839893113912689594815485706154461164071775481134379794909690501684643

B == 
40808563

G == 2
----------------------------------------------------------------
Certificate of primality for:
186935245989515158127969129347464851990429060640910951266513740972248428651109062997368144722015290092846666943896556191257222521203647606911446635194198213436423080005867489516421559330500722264446765608763224572386410155413161172707802334865729654109050873820610813855041667633843601286843

A == 
1213301773203241614897109856134894783021668292000023984098824423682568173639394290886185366993108292039068940333907505157813934962357206131450244004178619265868614859794316361031904412926604138893775068853175215502104744339658944443630407632290152772487455298652998368296998719996019

B == 
77035759

G == 2
----------------------------------------------------------------
Certificate of primality for:
83142661079751490510739960019112406284111408348732592580459037404394946037094409915127399165633756159385609671956087845517678367844901424617866988187132480585966721962585586730693443536100138246516868613250009028187662080828012497191775172228832247706080044971423654632146928165751885302331924491683

A == 
186935245989515158127969129347464851990429060640910951266513740972248428651109062997368144722015290092846666943896556191257222521203647606911446635194198213436423080005867489516421559330500722264446765608763224572386410155413161172707802334865729654109050873820610813855041667633843601286843

B == 
222383587

G == 2
----------------------------------------------------------------
Certificate of primality for:
3892354773803809855317742245039794448230625839512638747643814927766738642436392673485997449586432241626440927010641564064764336402368634186618250134234189066179771240232458249806850838490410473462391401438160528157981942499581634732706904411807195259620779379274017704050790865030808501633772117217899534443

A == 
83142661079751490510739960019112406284111408348732592580459037404394946037094409915127399165633756159385609671956087845517678367844901424617866988187132480585966721962585586730693443536100138246516868613250009028187662080828012497191775172228832247706080044971423654632146928165751885302331924491683

B == 
23407687

G == 2
----------------------------------------------------------------
Certificate of primality for:
1663606652988091811284014366560171522582683318514519379924950390627250155440313691226744227787921928894551755219495501365555370027257568506349958010457682898612082048959464465369892842603765280317696116552850664773291371490339084156052244256635115997453399761029567033971998617303988376172539172702246575225837054723

A == 
3892354773803809855317742245039794448230625839512638747643814927766738642436392673485997449586432241626440927010641564064764336402368634186618250134234189066179771240232458249806850838490410473462391401438160528157981942499581634732706904411807195259620779379274017704050790865030808501633772117217899534443

B == 
213701827

G == 2
----------------------------------------------------------------


Took 33057 ticks, 1048 bits
P == 1663606652988091811284014366560171522582683318514519379924950390627250155440313691226744227787921928894551755219495501365555370027257568506349958010457682898612082048959464465369892842603765280317696116552850664773291371490339084156052244256635115997453399761029567033971998617303988376172539172702246575225837054723
Q == 3892354773803809855317742245039794448230625839512638747643814927766738642436392673485997449586432241626440927010641564064764336402368634186618250134234189066179771240232458249806850838490410473462391401438160528157981942499581634732706904411807195259620779379274017704050790865030808501633772117217899534443

Enter # of bits: 
Enter number of bases to try (1 to 8):
Certificate of primality for:
85933926807634727

A == 
253758023

B == 
169322581

G == 5
----------------------------------------------------------------
Certificate of primality for:
23930198825086241462113799

A == 
85933926807634727

B == 
139236037

G == 11
----------------------------------------------------------------
Certificate of primality for:
6401844647261612602378676572510019

A == 
23930198825086241462113799

B == 
133760791

G == 2
----------------------------------------------------------------
Certificate of primality for:
269731366027728777712034888684015329354259

A == 
6401844647261612602378676572510019

B == 
21066691

G == 2
----------------------------------------------------------------
Certificate of primality for:
37942338209025571690075025099189467992329684223707

A == 
269731366027728777712034888684015329354259

B == 
70333567

G == 2
----------------------------------------------------------------
Certificate of primality for:
15306904714258982484473490774101705363308327436988160248323

A == 
37942338209025571690075025099189467992329684223707

B == 
201712723

G == 2
----------------------------------------------------------------
Certificate of primality for:
1616744757018513392810355191503853040357155275733333124624513530099

A == 
15306904714258982484473490774101705363308327436988160248323

B == 
52810963

G == 2
----------------------------------------------------------------
Certificate of primality for:
464222094814208047161771036072622485188658077940154689939306386289983787983

A == 
1616744757018513392810355191503853040357155275733333124624513530099

B == 
143566909

G == 5
----------------------------------------------------------------
Certificate of primality for:
187429931674053784626487560729643601208757374994177258429930699354770049369025096447

A == 
464222094814208047161771036072622485188658077940154689939306386289983787983

B == 
201875281

G == 5
----------------------------------------------------------------
Certificate of primality for:
100579220846502621074093727119851331775052664444339632682598589456666938521976625305832917563

A == 
187429931674053784626487560729643601208757374994177258429930699354770049369025096447

B == 
268311523

G == 2
----------------------------------------------------------------
Certificate of primality for:
1173616081309758475197022137833792133815753368965945885089720153370737965497134878651384030219765163

A == 
100579220846502621074093727119851331775052664444339632682598589456666938521976625305832917563

B == 
5834287

G == 2
----------------------------------------------------------------
Certificate of primality for:
191456913489905913185935197655672585713573070349044195411728114905691721186574907738081340754373032735283623

A == 
1173616081309758475197022137833792133815753368965945885089720153370737965497134878651384030219765163

B == 
81567097

G == 5
----------------------------------------------------------------
Certificate of primality for:
57856530489201750164178576399448868489243874083056587683743345599898489554401618943240901541005080049321706789987519

A == 
191456913489905913185935197655672585713573070349044195411728114905691721186574907738081340754373032735283623

B == 
151095433

G == 7
----------------------------------------------------------------
Certificate of primality for:
13790529750452576698109671710773784949185621244122040804792403407272729038377767162233653248852099545134831722512085881814803

A == 
57856530489201750164178576399448868489243874083056587683743345599898489554401618943240901541005080049321706789987519

B == 
119178679

G == 2
----------------------------------------------------------------
Certificate of primality for:
7075985989000817742677547821106534174334812111605018857703825637170140040509067704269696198231266351631132464035671858077052876058979

A == 
13790529750452576698109671710773784949185621244122040804792403407272729038377767162233653248852099545134831722512085881814803

B == 
256552363

G == 2
----------------------------------------------------------------
Certificate of primality for:
1227273006232588072907488910282307435921226646895131225407452056677899411162892829564455154080310937471747140942360789623819327234258162420463

A == 
7075985989000817742677547821106534174334812111605018857703825637170140040509067704269696198231266351631132464035671858077052876058979

B == 
86720989

G == 5
----------------------------------------------------------------
Certificate of primality for:
446764896913554613686067036908702877942872355053329937790398156069936255759889884246832779737114032666318220500106499161852193765380831330106375235763

A == 
1227273006232588072907488910282307435921226646895131225407452056677899411162892829564455154080310937471747140942360789623819327234258162420463

B == 
182015287

G == 2
----------------------------------------------------------------
Certificate of primality for:
5290203010849586596974953717018896543907195901082056939587768479377028575911127944611236020459652034082251335583308070846379514569838984811187823420951275243

A == 
446764896913554613686067036908702877942872355053329937790398156069936255759889884246832779737114032666318220500106499161852193765380831330106375235763

B == 
5920567

G == 2
----------------------------------------------------------------


Took 3454 ticks, 521 bits
P == 5290203010849586596974953717018896543907195901082056939587768479377028575911127944611236020459652034082251335583308070846379514569838984811187823420951275243
Q == 446764896913554613686067036908702877942872355053329937790398156069936255759889884246832779737114032666318220500106499161852193765380831330106375235763

; x86 timer in NASM
;
; Tom St Denis, [email protected]
[bits 32]
[section .data]
time dd 0, 0

[section .text]

%ifdef USE_ELF
[global t_start]
t_start:
%else
[global _t_start]
_t_start:
%endif
   push edx
   push eax
   rdtsc
   mov [time+0],edx
   mov [time+4],eax
   pop eax
   pop edx
   ret
   
%ifdef USE_ELF
[global t_read]
t_read:
%else
[global _t_read]
_t_read:
%endif
   rdtsc
   sub eax,[time+4]
   sbb edx,[time+0]
   ret
   

/* Tune the Karatsuba parameters
 *
 * Tom St Denis, [email protected]
 */
#include <tommath.h>
#include <time.h>

/* how many times todo each size mult.  Depends on your computer.  For slow computers
 * this can be low like 5 or 10.  For fast [re: Athlon] should be 25 - 50 or so 
 */
#define TIMES (1UL<<14UL)

/* RDTSC from Scott Duplichan */
static ulong64 TIMFUNC (void)
   {
   #if defined __GNUC__
      #if defined(__i386__) || defined(__x86_64__)
         unsigned long long a;
         __asm__ __volatile__ ("rdtsc\nmovl %%eax,%0\nmovl %%edx,4+%0\n"::"m"(a):"%eax","%edx");
         return a;
      #else /* gcc-IA64 version */
         unsigned long result;
         __asm__ __volatile__("mov %0=ar.itc" : "=r"(result) :: "memory");
         while (__builtin_expect ((int) result == -1, 0))
         __asm__ __volatile__("mov %0=ar.itc" : "=r"(result) :: "memory");
         return result;
      #endif

   // Microsoft and Intel Windows compilers
   #elif defined _M_IX86
     __asm rdtsc
   #elif defined _M_AMD64
     return __rdtsc ();
   #elif defined _M_IA64
     #if defined __INTEL_COMPILER
       #include <ia64intrin.h>
     #endif
      return __getReg (3116);
   #else
     #error need rdtsc function for this build
   #endif
   }


#ifndef X86_TIMER

/* generic ISO C timer */
ulong64 LBL_T;
void t_start(void) { LBL_T = TIMFUNC(); }
ulong64 t_read(void) { return TIMFUNC() - LBL_T; }

#else
extern void t_start(void);
extern ulong64 t_read(void);
#endif

ulong64 time_mult(int size, int s)
{
  unsigned long     x;
  mp_int  a, b, c;
  ulong64 t1;

  mp_init (&a);
  mp_init (&b);
  mp_init (&c);

  mp_rand (&a, size);
  mp_rand (&b, size);

  if (s == 1) { 
      KARATSUBA_MUL_CUTOFF = size;
  } else {
      KARATSUBA_MUL_CUTOFF = 100000;
  }

  t_start();
  for (x = 0; x < TIMES; x++) {
      mp_mul(&a,&b,&c);
  }
  t1 = t_read();
  mp_clear (&a);
  mp_clear (&b);
  mp_clear (&c);
  return t1;
}

ulong64 time_sqr(int size, int s)
{
  unsigned long     x;
  mp_int  a, b;
  ulong64 t1;

  mp_init (&a);
  mp_init (&b);

  mp_rand (&a, size);

  if (s == 1) { 
      KARATSUBA_SQR_CUTOFF = size;
  } else {
      KARATSUBA_SQR_CUTOFF = 100000;
  }

  t_start();
  for (x = 0; x < TIMES; x++) {
      mp_sqr(&a,&b);
  }
  t1 = t_read();
  mp_clear (&a);
  mp_clear (&b);
  return t1;
}

int
main (void)
{
  ulong64 t1, t2;
  int x, y;

  for (x = 8; ; x += 2) { 
     t1 = time_mult(x, 0);
     t2 = time_mult(x, 1);
     printf("%d: %9llu %9llu, %9llu\n", x, t1, t2, t2 - t1);
     if (t2 < t1) break;
  }
  y = x;

  for (x = 8; ; x += 2) { 
     t1 = time_sqr(x, 0);
     t2 = time_sqr(x, 1);
     printf("%d: %9llu %9llu, %9llu\n", x, t1, t2, t2 - t1);
     if (t2 < t1) break;
  }
  printf("KARATSUBA_MUL_CUTOFF = %d\n", y);
  printf("KARATSUBA_SQR_CUTOFF = %d\n", x);

  return 0;
}

To use the pretty graphs you have to first build/run the ltmtest from the root directory of the package.  
Todo this type 

make timing ; ltmtest

in the root.  It will run for a while [about ten minutes on most PCs] and produce a series of .log files in logs/.

After doing that run "gnuplot graphs.dem" to make the PNGs.  If you managed todo that all so far just open index.html to view
them all :-)

Have fun

Tom

480        87
960       111
1440       135
1920       159
2400       200
2880       224
3360       248
3840       272
4320       296
4800       320
5280       344
5760       368
6240       392
6720       416
7200       440
7680       464

513   1435869
769   3544970
1025   7791638
2049  46902238
2561  85334899
3073 141451412
4097 308770310

607   2109225
1279  10148314
2203  34126877
3217  82716424
4253 161569606

1024   7705271
2048  34286851
4096 165207491
521   1618631

532   1928550
784   3763908
1036   7564221
1540  16566059
2072  32283784
3080  79851565
4116 157843530

set terminal png
set size 1.75
set ylabel "Cycles per Operation"
set xlabel "Operand size (bits)"

set output "addsub.png"
plot 'add.log' smooth bezier title "Addition", 'sub.log' smooth bezier title "Subtraction"

set output "mult.png"
plot 'sqr.log' smooth bezier title "Squaring (without Karatsuba)", 'sqr_kara.log' smooth bezier title "Squaring (Karatsuba)", 'mult.log' smooth bezier title "Multiplication (without Karatsuba)", 'mult_kara.log' smooth bezier title "Multiplication (Karatsuba)"

set output "expt.png"
plot 'expt.log' smooth bezier title "Exptmod (Montgomery)", 'expt_dr.log' smooth bezier title "Exptmod (Dimminished Radix)", 'expt_2k.log' smooth bezier title "Exptmod (2k Reduction)"

set output "invmod.png"
plot 'invmod.log' smooth bezier title "Modular Inverse"

<html>
<head>
<title>LibTomMath Log Plots</title>
</head>
<body>

<h1>Addition and Subtraction</h1>
<center><img src=addsub.png></center>
<hr>

<h1>Multipliers</h1>
<center><img src=mult.png></center>
<hr>

<h1>Exptmod</h1>
<center><img src=expt.png></center>
<hr>

<h1>Modular Inverse</h1>
<center><img src=invmod.png></center>
<hr>

</body>
</html>

271       555
390       855
508      1161
631      1605
749      2117
871      2687
991      3329
1108      4084
1231      4786
1351      5624
1470      6392
1586      7364
1710      8218
1830      9255
1951     10217
2067     11461
2191     12463
2308     13677
2430     14800
2551     16232
2671     17460
2791     18899
2902     20247
3028     21902
3151     23240
3267     24927
3391     26441
3511     28277
3631     29838
3749     31751
3869     33673
3989     35431
4111     37518
4231     39426
4349     41504
4471     43567
4591     45786
4711     47876
4831     50299
4951     52427
5071     54785
5189     57241
5307     59730
5431     62194
5551     64761
5670     67322
5789     70073
5907     72663
6030     75437
6151     78242
6268     81202
6389     83948
6509     86985
6631     89903
6747     93184
6869     96044
6991     99286
7109    102395
7229    105917
7351    108940
7470    112490
7589    115702
7711    119508
7831    122632
7951    126410
8071    129808
8190    133895
8311    137146
8431    141218
8549    144732
8667    149131
8790    152462
8911    156754
9030    160479
9149    165138
9271    168601
9391    173185
9511    176988
9627    181976
9751    185539
9870    190388
9991    194335
10110    199605
10228    203298

271       560
391       870
511      1159
631      1605
750      2111
871      2737
991      3361
1111      4054
1231      4778
1351      5600
1471      6404
1591      7323
1710      8255
1831      9239
1948     10257
2070     11397
2190     12531
2308     13665
2429     14870
2550     16175
2671     17539
2787     18879
2911     20350
3031     21807
3150     23415
3270     24897
3388     26567
3511     28205
3627     30076
3751     31744
3869     33657
3991     35425
4111     37522
4229     39363
4351     41503
4470     43491
4590     45827
4711     47795
4828     50166
4951     52318
5070     54911
5191     57036
5308     58237
5431     60248
5551     62678
5671     64786
5791     67294
5908     69343
6031     71607
6151     74166
6271     76590
6391     78734
6511     81175
6631     83742
6750     86403
6868     88873
6990     91150
7110     94211
7228     96922
7351     99445
7469    102216
7589    104968
7711    108113
7827    110758
7950    113714
8071    116511
8186    119643
8310    122679
8425    125581
8551    128715
8669    131778
8788    135116
8910    138138
9031    141628
9148    144754
9268    148367
9391    151551
9511    155033
9631    158652
9751    162125
9871    165248
9988    168627
10111    172427
10231    176412

265       562
389       882
509      1207
631      1572
750      1990
859      2433
991      2894
1109      3555
1230      4228
1350      5018
1471      5805
1591      6579
1709      7415
1829      8329
1949      9225
2071     10139
2188     11239
2309     12178
2431     13212
2551     14294
2671     15551
2791     16512
2911     17718
3030     18876
3150     20259
3270     21374
3391     22650
3511     23948
3631     25493
3750     26756
3870     28225
3989     29705
4110     31409
4230     32834
4351     34327
4471     35818
4591     37636
4711     39228
4830     40868
4949     42393
5070     44541
5191     46269
5310     48162
5429     49728
5548     51985
5671     53948
5791     55885
5910     57584
6031     60082
6150     62239
6270     64309
6390     66014
6511     68766
6631     71012
6750     73172
6871     74952
6991     77909
7111     80371
7231     82666
7351     84531
7469     87698
7589     90318
7711    225384
7830    232428
7950    240009
8070    246522
8190    253662
8310    260961
8431    269253
8549    275743
8671    283769
8789    290811
8911    300034
9030    306873
9149    315085
9270    323944
9390    332390
9508    337519
9631    348986
9749    356904
9871    367013
9989    373831
10108    381033
10230    393475

271       560
388       878
511      1179
629      1625
751      1988
871      2423
989      2896
1111      3561
1231      4209
1350      5015
1470      5804
1591      6556
1709      7420
1831      8263
1951      9173
2070     10153
2191     11229
2310     12167
2431     13211
2550     14309
2671     15524
2788     16525
2910     17712
3028     18822
3148     20220
3271     21343
3391     22652
3511     23944
3630     25485
3750     26778
3868     28201
3990     29653
4111     31393
4225     32841
4350     34328
4471     35786
4590     37652
4711     39245
4830     40876
4951     42433
5068     44547
5191     46321
5311     48140
5430     49727
5550     52034
5671     53954
5791     55921
5908     57597
6031     60084
6148     62226
6270     64295
6390     66045
6511     68779
6629     71003
6751     73169
6871     74992
6991     77895
7110     80376
7231     82628
7351     84468
7470     87664
7591     90284
7711     91352
7828     93995
7950     96276
8071     98691
8190    101256
8308    103631
8431    105222
8550    108343
8671    110281
8787    112764
8911    115397
9031    117690
9151    120266
9271    122715
9391    124624
9510    127937
9630    130313
9750    132914
9871    136129
9991    138517
10108    141525
10231    144225

480        94
960       116
1440       140
1920       164
2400       205
2880       229
3360       253
3840       277
4320       299
4800       321
5280       345
5760       371
6240       395
6720       419
7200       441
7680       465

const float s_logv_2[] = {
   0.000000000, 0.000000000, 1.000000000, 0.630929754, 	/*  0  1  2  3 */
   0.500000000, 0.430676558, 0.386852807, 0.356207187, 	/*  4  5  6  7 */
   0.333333333, 0.315464877, 0.301029996, 0.289064826, 	/*  8  9 10 11 */
   0.278942946, 0.270238154, 0.262649535, 0.255958025, 	/* 12 13 14 15 */
   0.250000000, 0.244650542, 0.239812467, 0.235408913, 	/* 16 17 18 19 */
   0.231378213, 0.227670249, 0.224243824, 0.221064729, 	/* 20 21 22 23 */
   0.218104292, 0.215338279, 0.212746054, 0.210309918, 	/* 24 25 26 27 */
   0.208014598, 0.205846832, 0.203795047, 0.201849087, 	/* 28 29 30 31 */
   0.200000000, 0.198239863, 0.196561632, 0.194959022, 	/* 32 33 34 35 */
   0.193426404, 0.191958720, 0.190551412, 0.189200360, 	/* 36 37 38 39 */
   0.187901825, 0.186652411, 0.185449023, 0.184288833, 	/* 40 41 42 43 */
   0.183169251, 0.182087900, 0.181042597, 0.180031327, 	/* 44 45 46 47 */
   0.179052232, 0.178103594, 0.177183820, 0.176291434, 	/* 48 49 50 51 */
   0.175425064, 0.174583430, 0.173765343, 0.172969690, 	/* 52 53 54 55 */
   0.172195434, 0.171441601, 0.170707280, 0.169991616, 	/* 56 57 58 59 */
   0.169293808, 0.168613099, 0.167948779, 0.167300179, 	/* 60 61 62 63 */
   0.166666667
};

/* Default configuration for MPI library */

#ifndef MPI_CONFIG_H_
#define MPI_CONFIG_H_

/*
  For boolean options, 
  0 = no
  1 = yes

  Other options are documented individually.

 */

#ifndef MP_IOFUNC
#define MP_IOFUNC     0  /* include mp_print() ?                */
#endif

#ifndef MP_MODARITH
#define MP_MODARITH   1  /* include modular arithmetic ?        */
#endif

#ifndef MP_NUMTH
#define MP_NUMTH      1  /* include number theoretic functions? */
#endif

#ifndef MP_LOGTAB
#define MP_LOGTAB     1  /* use table of logs instead of log()? */
#endif

#ifndef MP_MEMSET
#define MP_MEMSET     1  /* use memset() to zero buffers?       */
#endif

#ifndef MP_MEMCPY
#define MP_MEMCPY     1  /* use memcpy() to copy buffers?       */
#endif

#ifndef MP_CRYPTO
#define MP_CRYPTO     1  /* erase memory on free?               */
#endif

#ifndef MP_ARGCHK
/*
  0 = no parameter checks
  1 = runtime checks, continue execution and return an error to caller
  2 = assertions; dump core on parameter errors
 */
#define MP_ARGCHK     2  /* how to check input arguments        */
#endif

#ifndef MP_DEBUG
#define MP_DEBUG      0  /* print diagnostic output?            */
#endif

#ifndef MP_DEFPREC
#define MP_DEFPREC    64 /* default precision, in digits        */
#endif

#ifndef MP_MACRO
#define MP_MACRO      1  /* use macros for frequent calls?      */
#endif

#ifndef MP_SQUARE
#define MP_SQUARE     1  /* use separate squaring code?         */
#endif

#ifndef MP_PTAB_SIZE
/*
  When building mpprime.c, we build in a table of small prime
  values to use for primality testing.  The more you include,
  the more space they take up.  See primes.c for the possible
  values (currently 16, 32, 64, 128, 256, and 6542)
 */
#define MP_PTAB_SIZE  128  /* how many built-in primes?         */
#endif

#ifndef MP_COMPAT_MACROS
#define MP_COMPAT_MACROS 1   /* define compatibility macros?    */
#endif

#endif /* ifndef MPI_CONFIG_H_ */


/* crc==3287762869, version==2, Sat Feb 02 06:43:53 2002 */

/* Type definitions generated by 'types.pl' */
typedef char               mp_sign;
typedef unsigned short     mp_digit;  /* 2 byte type */
typedef unsigned int       mp_word;   /* 4 byte type */
typedef unsigned int       mp_size;
typedef int                mp_err;

#define MP_DIGIT_BIT       (CHAR_BIT*sizeof(mp_digit))
#define MP_DIGIT_MAX       USHRT_MAX
#define MP_WORD_BIT        (CHAR_BIT*sizeof(mp_word))
#define MP_WORD_MAX        UINT_MAX

#define MP_DIGIT_SIZE      2
#define DIGIT_FMT          "%04X"
#define RADIX              (MP_DIGIT_MAX+1)

/*
    mpi.c

    by Michael J. Fromberger <[email protected]>
    Copyright (C) 1998 Michael J. Fromberger, All Rights Reserved

    Arbitrary precision integer arithmetic library
 */

#include "mpi.h"
#include <stdlib.h>
#include <string.h>
#include <ctype.h>

#if MP_DEBUG
#include <stdio.h>

#define DIAG(T,V) {fprintf(stderr,T);mp_print(V,stderr);fputc('\n',stderr);}
#else
#define DIAG(T,V)
#endif

/* 
   If MP_LOGTAB is not defined, use the math library to compute the
   logarithms on the fly.  Otherwise, use the static table below.
   Pick which works best for your system.
 */
#if MP_LOGTAB

/* {{{ s_logv_2[] - log table for 2 in various bases */

/*
  A table of the logs of 2 for various bases (the 0 and 1 entries of
  this table are meaningless and should not be referenced).  

  This table is used to compute output lengths for the mp_toradix()
  function.  Since a number n in radix r takes up about log_r(n)
  digits, we estimate the output size by taking the least integer
  greater than log_r(n), where:

  log_r(n) = log_2(n) * log_r(2)

  This table, therefore, is a table of log_r(2) for 2 <= r <= 36,
  which are the output bases supported.  
 */

#include "logtab.h"

/* }}} */
#define LOG_V_2(R)  s_logv_2[(R)]

#else

#include <math.h>
#define LOG_V_2(R)  (log(2.0)/log(R))

#endif

/* Default precision for newly created mp_int's      */
static unsigned int s_mp_defprec = MP_DEFPREC;

/* {{{ Digit arithmetic macros */

/*
  When adding and multiplying digits, the results can be larger than
  can be contained in an mp_digit.  Thus, an mp_word is used.  These
  macros mask off the upper and lower digits of the mp_word (the
  mp_word may be more than 2 mp_digits wide, but we only concern
  ourselves with the low-order 2 mp_digits)

  If your mp_word DOES have more than 2 mp_digits, you need to
  uncomment the first line, and comment out the second.
 */

/* #define  CARRYOUT(W)  (((W)>>DIGIT_BIT)&MP_DIGIT_MAX) */
#define  CARRYOUT(W)  ((W)>>DIGIT_BIT)
#define  ACCUM(W)     ((W)&MP_DIGIT_MAX)

/* }}} */

/* {{{ Comparison constants */

#define  MP_LT       -1
#define  MP_EQ        0
#define  MP_GT        1

/* }}} */

/* {{{ Constant strings */

/* Constant strings returned by mp_strerror() */
static const char *const mp_err_string[] = {
  "unknown result code",     /* say what?            */
  "boolean true",            /* MP_OKAY, MP_YES      */
  "boolean false",           /* MP_NO                */
  "out of memory",           /* MP_MEM               */
  "argument out of range",   /* MP_RANGE             */
  "invalid input parameter", /* MP_BADARG            */
  "result is undefined"      /* MP_UNDEF             */
};

/* Value to digit maps for radix conversion   */

/* s_dmap_1 - standard digits and letters */
static const char *s_dmap_1 = 
  "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz+/";

#if 0
/* s_dmap_2 - base64 ordering for digits  */
static const char *s_dmap_2 =
  "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
#endif

/* }}} */

/* {{{ Static function declarations */

/* 
   If MP_MACRO is false, these will be defined as actual functions;
   otherwise, suitable macro definitions will be used.  This works
   around the fact that ANSI C89 doesn't support an 'inline' keyword
   (although I hear C9x will ... about bloody time).  At present, the
   macro definitions are identical to the function bodies, but they'll
   expand in place, instead of generating a function call.

   I chose these particular functions to be made into macros because
   some profiling showed they are called a lot on a typical workload,
   and yet they are primarily housekeeping.
 */
#if MP_MACRO == 0
 void     s_mp_setz(mp_digit *dp, mp_size count); /* zero digits           */
 void     s_mp_copy(mp_digit *sp, mp_digit *dp, mp_size count); /* copy    */
 void    *s_mp_alloc(size_t nb, size_t ni);       /* general allocator     */
 void     s_mp_free(void *ptr);                   /* general free function */
#else

 /* Even if these are defined as macros, we need to respect the settings
    of the MP_MEMSET and MP_MEMCPY configuration options...
  */
 #if MP_MEMSET == 0
  #define  s_mp_setz(dp, count) \
       {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=0;}
 #else
  #define  s_mp_setz(dp, count) memset(dp, 0, (count) * sizeof(mp_digit))
 #endif /* MP_MEMSET */

 #if MP_MEMCPY == 0
  #define  s_mp_copy(sp, dp, count) \
       {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=(sp)[ix];}
 #else
  #define  s_mp_copy(sp, dp, count) memcpy(dp, sp, (count) * sizeof(mp_digit))
 #endif /* MP_MEMCPY */

 #define  s_mp_alloc(nb, ni)  calloc(nb, ni)
 #define  s_mp_free(ptr) {if(ptr) free(ptr);}
#endif /* MP_MACRO */

mp_err   s_mp_grow(mp_int *mp, mp_size min);   /* increase allocated size */
mp_err   s_mp_pad(mp_int *mp, mp_size min);    /* left pad with zeroes    */

void     s_mp_clamp(mp_int *mp);               /* clip leading zeroes     */

void     s_mp_exch(mp_int *a, mp_int *b);      /* swap a and b in place   */

mp_err   s_mp_lshd(mp_int *mp, mp_size p);     /* left-shift by p digits  */
void     s_mp_rshd(mp_int *mp, mp_size p);     /* right-shift by p digits */
void     s_mp_div_2d(mp_int *mp, mp_digit d);  /* divide by 2^d in place  */
void     s_mp_mod_2d(mp_int *mp, mp_digit d);  /* modulo 2^d in place     */
mp_err   s_mp_mul_2d(mp_int *mp, mp_digit d);  /* multiply by 2^d in place*/
void     s_mp_div_2(mp_int *mp);               /* divide by 2 in place    */
mp_err   s_mp_mul_2(mp_int *mp);               /* multiply by 2 in place  */
mp_digit s_mp_norm(mp_int *a, mp_int *b);      /* normalize for division  */
mp_err   s_mp_add_d(mp_int *mp, mp_digit d);   /* unsigned digit addition */
mp_err   s_mp_sub_d(mp_int *mp, mp_digit d);   /* unsigned digit subtract */
mp_err   s_mp_mul_d(mp_int *mp, mp_digit d);   /* unsigned digit multiply */
mp_err   s_mp_div_d(mp_int *mp, mp_digit d, mp_digit *r);
		                               /* unsigned digit divide   */
mp_err   s_mp_reduce(mp_int *x, mp_int *m, mp_int *mu);
                                               /* Barrett reduction       */
mp_err   s_mp_add(mp_int *a, mp_int *b);       /* magnitude addition      */
mp_err   s_mp_sub(mp_int *a, mp_int *b);       /* magnitude subtract      */
mp_err   s_mp_mul(mp_int *a, mp_int *b);       /* magnitude multiply      */
#if 0
void     s_mp_kmul(mp_digit *a, mp_digit *b, mp_digit *out, mp_size len);
                                               /* multiply buffers in place */
#endif
#if MP_SQUARE
mp_err   s_mp_sqr(mp_int *a);                  /* magnitude square        */
#else
#define  s_mp_sqr(a) s_mp_mul(a, a)
#endif
mp_err   s_mp_div(mp_int *a, mp_int *b);       /* magnitude divide        */
mp_err   s_mp_2expt(mp_int *a, mp_digit k);    /* a = 2^k                 */
int      s_mp_cmp(mp_int *a, mp_int *b);       /* magnitude comparison    */
int      s_mp_cmp_d(mp_int *a, mp_digit d);    /* magnitude digit compare */
int      s_mp_ispow2(mp_int *v);               /* is v a power of 2?      */
int      s_mp_ispow2d(mp_digit d);             /* is d a power of 2?      */

int      s_mp_tovalue(char ch, int r);          /* convert ch to value    */
char     s_mp_todigit(int val, int r, int low); /* convert val to digit   */
int      s_mp_outlen(int bits, int r);          /* output length in bytes */

/* }}} */

/* {{{ Default precision manipulation */

unsigned int mp_get_prec(void)
{
  return s_mp_defprec;

} /* end mp_get_prec() */

void         mp_set_prec(unsigned int prec)
{
  if(prec == 0)
    s_mp_defprec = MP_DEFPREC;
  else
    s_mp_defprec = prec;

} /* end mp_set_prec() */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ mp_init(mp) */

/*
  mp_init(mp)

  Initialize a new zero-valued mp_int.  Returns MP_OKAY if successful,
  MP_MEM if memory could not be allocated for the structure.
 */

mp_err mp_init(mp_int *mp)
{
  return mp_init_size(mp, s_mp_defprec);

} /* end mp_init() */

/* }}} */

/* {{{ mp_init_array(mp[], count) */

mp_err mp_init_array(mp_int mp[], int count)
{
  mp_err  res;
  int     pos;

  ARGCHK(mp !=NULL && count > 0, MP_BADARG);

  for(pos = 0; pos < count; ++pos) {
    if((res = mp_init(&mp[pos])) != MP_OKAY)
      goto CLEANUP;
  }

  return MP_OKAY;

 CLEANUP:
  while(--pos >= 0) 
    mp_clear(&mp[pos]);

  return res;

} /* end mp_init_array() */

/* }}} */

/* {{{ mp_init_size(mp, prec) */

/*
  mp_init_size(mp, prec)

  Initialize a new zero-valued mp_int with at least the given
  precision; returns MP_OKAY if successful, or MP_MEM if memory could
  not be allocated for the structure.
 */

mp_err mp_init_size(mp_int *mp, mp_size prec)
{
  ARGCHK(mp != NULL && prec > 0, MP_BADARG);

  if((DIGITS(mp) = s_mp_alloc(prec, sizeof(mp_digit))) == NULL)
    return MP_MEM;

  SIGN(mp) = MP_ZPOS;
  USED(mp) = 1;
  ALLOC(mp) = prec;

  return MP_OKAY;

} /* end mp_init_size() */

/* }}} */

/* {{{ mp_init_copy(mp, from) */

/*
  mp_init_copy(mp, from)

  Initialize mp as an exact copy of from.  Returns MP_OKAY if
  successful, MP_MEM if memory could not be allocated for the new
  structure.
 */

mp_err mp_init_copy(mp_int *mp, mp_int *from)
{
  ARGCHK(mp != NULL && from != NULL, MP_BADARG);

  if(mp == from)
    return MP_OKAY;

  if((DIGITS(mp) = s_mp_alloc(USED(from), sizeof(mp_digit))) == NULL)
    return MP_MEM;

  s_mp_copy(DIGITS(from), DIGITS(mp), USED(from));
  USED(mp) = USED(from);
  ALLOC(mp) = USED(from);
  SIGN(mp) = SIGN(from);

  return MP_OKAY;

} /* end mp_init_copy() */

/* }}} */

/* {{{ mp_copy(from, to) */

/*
  mp_copy(from, to)

  Copies the mp_int 'from' to the mp_int 'to'.  It is presumed that
  'to' has already been initialized (if not, use mp_init_copy()
  instead). If 'from' and 'to' are identical, nothing happens.
 */

mp_err mp_copy(mp_int *from, mp_int *to)
{
  ARGCHK(from != NULL && to != NULL, MP_BADARG);

  if(from == to)
    return MP_OKAY;

  { /* copy */
    mp_digit   *tmp;

    /*
      If the allocated buffer in 'to' already has enough space to hold
      all the used digits of 'from', we'll re-use it to avoid hitting
      the memory allocater more than necessary; otherwise, we'd have
      to grow anyway, so we just allocate a hunk and make the copy as
      usual
     */
    if(ALLOC(to) >= USED(from)) {
      s_mp_setz(DIGITS(to) + USED(from), ALLOC(to) - USED(from));
      s_mp_copy(DIGITS(from), DIGITS(to), USED(from));
      
    } else {
      if((tmp = s_mp_alloc(USED(from), sizeof(mp_digit))) == NULL)
	return MP_MEM;

      s_mp_copy(DIGITS(from), tmp, USED(from));

      if(DIGITS(to) != NULL) {
#if MP_CRYPTO
	s_mp_setz(DIGITS(to), ALLOC(to));
#endif
	s_mp_free(DIGITS(to));
      }

      DIGITS(to) = tmp;
      ALLOC(to) = USED(from);
    }

    /* Copy the precision and sign from the original */
    USED(to) = USED(from);
    SIGN(to) = SIGN(from);
  } /* end copy */

  return MP_OKAY;

} /* end mp_copy() */

/* }}} */

/* {{{ mp_exch(mp1, mp2) */

/*
  mp_exch(mp1, mp2)

  Exchange mp1 and mp2 without allocating any intermediate memory
  (well, unless you count the stack space needed for this call and the
  locals it creates...).  This cannot fail.
 */

void mp_exch(mp_int *mp1, mp_int *mp2)
{
#if MP_ARGCHK == 2
  assert(mp1 != NULL && mp2 != NULL);
#else
  if(mp1 == NULL || mp2 == NULL)
    return;
#endif

  s_mp_exch(mp1, mp2);

} /* end mp_exch() */

/* }}} */

/* {{{ mp_clear(mp) */

/*
  mp_clear(mp)

  Release the storage used by an mp_int, and void its fields so that
  if someone calls mp_clear() again for the same int later, we won't
  get tollchocked.
 */

void   mp_clear(mp_int *mp)
{
  if(mp == NULL)
    return;

  if(DIGITS(mp) != NULL) {
#if MP_CRYPTO
    s_mp_setz(DIGITS(mp), ALLOC(mp));
#endif
    s_mp_free(DIGITS(mp));
    DIGITS(mp) = NULL;
  }

  USED(mp) = 0;
  ALLOC(mp) = 0;

} /* end mp_clear() */

/* }}} */

/* {{{ mp_clear_array(mp[], count) */

void   mp_clear_array(mp_int mp[], int count)
{
  ARGCHK(mp != NULL && count > 0, MP_BADARG);

  while(--count >= 0) 
    mp_clear(&mp[count]);

} /* end mp_clear_array() */

/* }}} */

/* {{{ mp_zero(mp) */

/*
  mp_zero(mp) 

  Set mp to zero.  Does not change the allocated size of the structure,
  and therefore cannot fail (except on a bad argument, which we ignore)
 */
void   mp_zero(mp_int *mp)
{
  if(mp == NULL)
    return;

  s_mp_setz(DIGITS(mp), ALLOC(mp));
  USED(mp) = 1;
  SIGN(mp) = MP_ZPOS;

} /* end mp_zero() */

/* }}} */

/* {{{ mp_set(mp, d) */

void   mp_set(mp_int *mp, mp_digit d)
{
  if(mp == NULL)
    return;

  mp_zero(mp);
  DIGIT(mp, 0) = d;

} /* end mp_set() */

/* }}} */

/* {{{ mp_set_int(mp, z) */

mp_err mp_set_int(mp_int *mp, long z)
{
  int            ix;
  unsigned long  v = abs(z);
  mp_err         res;

  ARGCHK(mp != NULL, MP_BADARG);

  mp_zero(mp);
  if(z == 0)
    return MP_OKAY;  /* shortcut for zero */

  for(ix = sizeof(long) - 1; ix >= 0; ix--) {

    if((res = s_mp_mul_2d(mp, CHAR_BIT)) != MP_OKAY)
      return res;

    res = s_mp_add_d(mp, 
		     (mp_digit)((v >> (ix * CHAR_BIT)) & UCHAR_MAX));
    if(res != MP_OKAY)
      return res;

  }

  if(z < 0)
    SIGN(mp) = MP_NEG;

  return MP_OKAY;

} /* end mp_set_int() */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ Digit arithmetic */

/* {{{ mp_add_d(a, d, b) */

/*
  mp_add_d(a, d, b)

  Compute the sum b = a + d, for a single digit d.  Respects the sign of
  its primary addend (single digits are unsigned anyway).
 */

mp_err mp_add_d(mp_int *a, mp_digit d, mp_int *b)
{
  mp_err   res = MP_OKAY;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  if(SIGN(b) == MP_ZPOS) {
    res = s_mp_add_d(b, d);
  } else if(s_mp_cmp_d(b, d) >= 0) {
    res = s_mp_sub_d(b, d);
  } else {
    SIGN(b) = MP_ZPOS;

    DIGIT(b, 0) = d - DIGIT(b, 0);
  }

  return res;

} /* end mp_add_d() */

/* }}} */

/* {{{ mp_sub_d(a, d, b) */

/*
  mp_sub_d(a, d, b)

  Compute the difference b = a - d, for a single digit d.  Respects the
  sign of its subtrahend (single digits are unsigned anyway).
 */

mp_err mp_sub_d(mp_int *a, mp_digit d, mp_int *b)
{
  mp_err   res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  if(SIGN(b) == MP_NEG) {
    if((res = s_mp_add_d(b, d)) != MP_OKAY)
      return res;

  } else if(s_mp_cmp_d(b, d) >= 0) {
    if((res = s_mp_sub_d(b, d)) != MP_OKAY)
      return res;

  } else {
    mp_neg(b, b);

    DIGIT(b, 0) = d - DIGIT(b, 0);
    SIGN(b) = MP_NEG;
  }

  if(s_mp_cmp_d(b, 0) == 0)
    SIGN(b) = MP_ZPOS;

  return MP_OKAY;

} /* end mp_sub_d() */

/* }}} */

/* {{{ mp_mul_d(a, d, b) */

/*
  mp_mul_d(a, d, b)

  Compute the product b = a * d, for a single digit d.  Respects the sign
  of its multiplicand (single digits are unsigned anyway)
 */

mp_err mp_mul_d(mp_int *a, mp_digit d, mp_int *b)
{
  mp_err  res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if(d == 0) {
    mp_zero(b);
    return MP_OKAY;
  }

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  res = s_mp_mul_d(b, d);

  return res;

} /* end mp_mul_d() */

/* }}} */

/* {{{ mp_mul_2(a, c) */

mp_err mp_mul_2(mp_int *a, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if((res = mp_copy(a, c)) != MP_OKAY)
    return res;

  return s_mp_mul_2(c);

} /* end mp_mul_2() */

/* }}} */

/* {{{ mp_div_d(a, d, q, r) */

/*
  mp_div_d(a, d, q, r)

  Compute the quotient q = a / d and remainder r = a mod d, for a
  single digit d.  Respects the sign of its divisor (single digits are
  unsigned anyway).
 */

mp_err mp_div_d(mp_int *a, mp_digit d, mp_int *q, mp_digit *r)
{
  mp_err   res;
  mp_digit rem;
  int      pow;

  ARGCHK(a != NULL, MP_BADARG);

  if(d == 0)
    return MP_RANGE;

  /* Shortcut for powers of two ... */
  if((pow = s_mp_ispow2d(d)) >= 0) {
    mp_digit  mask;

    mask = (1 << pow) - 1;
    rem = DIGIT(a, 0) & mask;

    if(q) {
      mp_copy(a, q);
      s_mp_div_2d(q, pow);
    }

    if(r)
      *r = rem;

    return MP_OKAY;
  }

  /*
    If the quotient is actually going to be returned, we'll try to
    avoid hitting the memory allocator by copying the dividend into it
    and doing the division there.  This can't be any _worse_ than
    always copying, and will sometimes be better (since it won't make
    another copy)

    If it's not going to be returned, we need to allocate a temporary
    to hold the quotient, which will just be discarded.
   */
  if(q) {
    if((res = mp_copy(a, q)) != MP_OKAY)
      return res;

    res = s_mp_div_d(q, d, &rem);
    if(s_mp_cmp_d(q, 0) == MP_EQ)
      SIGN(q) = MP_ZPOS;

  } else {
    mp_int  qp;

    if((res = mp_init_copy(&qp, a)) != MP_OKAY)
      return res;

    res = s_mp_div_d(&qp, d, &rem);
    if(s_mp_cmp_d(&qp, 0) == 0)
      SIGN(&qp) = MP_ZPOS;

    mp_clear(&qp);
  }

  if(r)
    *r = rem;

  return res;

} /* end mp_div_d() */

/* }}} */

/* {{{ mp_div_2(a, c) */

/*
  mp_div_2(a, c)

  Compute c = a / 2, disregarding the remainder.
 */

mp_err mp_div_2(mp_int *a, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if((res = mp_copy(a, c)) != MP_OKAY)
    return res;

  s_mp_div_2(c);

  return MP_OKAY;

} /* end mp_div_2() */

/* }}} */

/* {{{ mp_expt_d(a, d, b) */

mp_err mp_expt_d(mp_int *a, mp_digit d, mp_int *c)
{
  mp_int   s, x;
  mp_err   res;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if((res = mp_init(&s)) != MP_OKAY)
    return res;
  if((res = mp_init_copy(&x, a)) != MP_OKAY)
    goto X;

  DIGIT(&s, 0) = 1;

  while(d != 0) {
    if(d & 1) {
      if((res = s_mp_mul(&s, &x)) != MP_OKAY)
	goto CLEANUP;
    }

    d >>= 1;

    if((res = s_mp_sqr(&x)) != MP_OKAY)
      goto CLEANUP;
  }

  s_mp_exch(&s, c);

CLEANUP:
  mp_clear(&x);
X:
  mp_clear(&s);

  return res;

} /* end mp_expt_d() */

/* }}} */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ Full arithmetic */

/* {{{ mp_abs(a, b) */

/*
  mp_abs(a, b)

  Compute b = |a|.  'a' and 'b' may be identical.
 */

mp_err mp_abs(mp_int *a, mp_int *b)
{
  mp_err   res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  SIGN(b) = MP_ZPOS;

  return MP_OKAY;

} /* end mp_abs() */

/* }}} */

/* {{{ mp_neg(a, b) */

/*
  mp_neg(a, b)

  Compute b = -a.  'a' and 'b' may be identical.
 */

mp_err mp_neg(mp_int *a, mp_int *b)
{
  mp_err   res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  if(s_mp_cmp_d(b, 0) == MP_EQ) 
    SIGN(b) = MP_ZPOS;
  else 
    SIGN(b) = (SIGN(b) == MP_NEG) ? MP_ZPOS : MP_NEG;

  return MP_OKAY;

} /* end mp_neg() */

/* }}} */

/* {{{ mp_add(a, b, c) */

/*
  mp_add(a, b, c)

  Compute c = a + b.  All parameters may be identical.
 */

mp_err mp_add(mp_int *a, mp_int *b, mp_int *c)
{
  mp_err  res;
  int     cmp;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  if(SIGN(a) == SIGN(b)) { /* same sign:  add values, keep sign */

    /* Commutativity of addition lets us do this in either order,
       so we avoid having to use a temporary even if the result 
       is supposed to replace the output
     */
    if(c == b) {
      if((res = s_mp_add(c, a)) != MP_OKAY)
	return res;
    } else {
      if(c != a && (res = mp_copy(a, c)) != MP_OKAY)
	return res;

      if((res = s_mp_add(c, b)) != MP_OKAY) 
	return res;
    }

  } else if((cmp = s_mp_cmp(a, b)) > 0) {  /* different sign: a > b   */

    /* If the output is going to be clobbered, we will use a temporary
       variable; otherwise, we'll do it without touching the memory 
       allocator at all, if possible
     */
    if(c == b) {
      mp_int  tmp;

      if((res = mp_init_copy(&tmp, a)) != MP_OKAY)
	return res;
      if((res = s_mp_sub(&tmp, b)) != MP_OKAY) {
	mp_clear(&tmp);
	return res;
      }

      s_mp_exch(&tmp, c);
      mp_clear(&tmp);

    } else {

      if(c != a && (res = mp_copy(a, c)) != MP_OKAY)
	return res;
      if((res = s_mp_sub(c, b)) != MP_OKAY)
	return res;

    }

  } else if(cmp == 0) {             /* different sign, a == b   */

    mp_zero(c);
    return MP_OKAY;

  } else {                          /* different sign: a < b    */

    /* See above... */
    if(c == a) {
      mp_int  tmp;

      if((res = mp_init_copy(&tmp, b)) != MP_OKAY)
	return res;
      if((res = s_mp_sub(&tmp, a)) != MP_OKAY) {
	mp_clear(&tmp);
	return res;
      }

      s_mp_exch(&tmp, c);
      mp_clear(&tmp);

    } else {

      if(c != b && (res = mp_copy(b, c)) != MP_OKAY)
	return res;
      if((res = s_mp_sub(c, a)) != MP_OKAY)
	return res;

    }
  }

  if(USED(c) == 1 && DIGIT(c, 0) == 0)
    SIGN(c) = MP_ZPOS;

  return MP_OKAY;

} /* end mp_add() */

/* }}} */

/* {{{ mp_sub(a, b, c) */

/*
  mp_sub(a, b, c)

  Compute c = a - b.  All parameters may be identical.
 */

mp_err mp_sub(mp_int *a, mp_int *b, mp_int *c)
{
  mp_err  res;
  int     cmp;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  if(SIGN(a) != SIGN(b)) {
    if(c == a) {
      if((res = s_mp_add(c, b)) != MP_OKAY)
	return res;
    } else {
      if(c != b && ((res = mp_copy(b, c)) != MP_OKAY))
	return res;
      if((res = s_mp_add(c, a)) != MP_OKAY)
	return res;
      SIGN(c) = SIGN(a);
    }

  } else if((cmp = s_mp_cmp(a, b)) > 0) { /* Same sign, a > b */
    if(c == b) {
      mp_int  tmp;

      if((res = mp_init_copy(&tmp, a)) != MP_OKAY)
	return res;
      if((res = s_mp_sub(&tmp, b)) != MP_OKAY) {
	mp_clear(&tmp);
	return res;
      }
      s_mp_exch(&tmp, c);
      mp_clear(&tmp);

    } else {
      if(c != a && ((res = mp_copy(a, c)) != MP_OKAY))
	return res;

      if((res = s_mp_sub(c, b)) != MP_OKAY)
	return res;
    }

  } else if(cmp == 0) {  /* Same sign, equal magnitude */
    mp_zero(c);
    return MP_OKAY;

  } else {               /* Same sign, b > a */
    if(c == a) {
      mp_int  tmp;

      if((res = mp_init_copy(&tmp, b)) != MP_OKAY)
	return res;

      if((res = s_mp_sub(&tmp, a)) != MP_OKAY) {
	mp_clear(&tmp);
	return res;
      }
      s_mp_exch(&tmp, c);
      mp_clear(&tmp);

    } else {
      if(c != b && ((res = mp_copy(b, c)) != MP_OKAY)) 
	return res;

      if((res = s_mp_sub(c, a)) != MP_OKAY)
	return res;
    }

    SIGN(c) = !SIGN(b);
  }

  if(USED(c) == 1 && DIGIT(c, 0) == 0)
    SIGN(c) = MP_ZPOS;

  return MP_OKAY;

} /* end mp_sub() */

/* }}} */

/* {{{ mp_mul(a, b, c) */

/*
  mp_mul(a, b, c)

  Compute c = a * b.  All parameters may be identical.
 */

mp_err mp_mul(mp_int *a, mp_int *b, mp_int *c)
{
  mp_err   res;
  mp_sign  sgn;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  sgn = (SIGN(a) == SIGN(b)) ? MP_ZPOS : MP_NEG;

  if(c == b) {
    if((res = s_mp_mul(c, a)) != MP_OKAY)
      return res;

  } else {
    if((res = mp_copy(a, c)) != MP_OKAY)
      return res;

    if((res = s_mp_mul(c, b)) != MP_OKAY)
      return res;
  }
  
  if(sgn == MP_ZPOS || s_mp_cmp_d(c, 0) == MP_EQ)
    SIGN(c) = MP_ZPOS;
  else
    SIGN(c) = sgn;
  
  return MP_OKAY;

} /* end mp_mul() */

/* }}} */

/* {{{ mp_mul_2d(a, d, c) */

/*
  mp_mul_2d(a, d, c)

  Compute c = a * 2^d.  a may be the same as c.
 */

mp_err mp_mul_2d(mp_int *a, mp_digit d, mp_int *c)
{
  mp_err   res;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if((res = mp_copy(a, c)) != MP_OKAY)
    return res;

  if(d == 0)
    return MP_OKAY;

  return s_mp_mul_2d(c, d);

} /* end mp_mul() */

/* }}} */

/* {{{ mp_sqr(a, b) */

#if MP_SQUARE
mp_err mp_sqr(mp_int *a, mp_int *b)
{
  mp_err   res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if((res = mp_copy(a, b)) != MP_OKAY)
    return res;

  if((res = s_mp_sqr(b)) != MP_OKAY)
    return res;

  SIGN(b) = MP_ZPOS;

  return MP_OKAY;

} /* end mp_sqr() */
#endif

/* }}} */

/* {{{ mp_div(a, b, q, r) */

/*
  mp_div(a, b, q, r)

  Compute q = a / b and r = a mod b.  Input parameters may be re-used
  as output parameters.  If q or r is NULL, that portion of the
  computation will be discarded (although it will still be computed)

  Pay no attention to the hacker behind the curtain.
 */

mp_err mp_div(mp_int *a, mp_int *b, mp_int *q, mp_int *r)
{
  mp_err   res;
  mp_int   qtmp, rtmp;
  int      cmp;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  if(mp_cmp_z(b) == MP_EQ)
    return MP_RANGE;

  /* If a <= b, we can compute the solution without division, and
     avoid any memory allocation
   */
  if((cmp = s_mp_cmp(a, b)) < 0) {
    if(r) {
      if((res = mp_copy(a, r)) != MP_OKAY)
	return res;
    }

    if(q) 
      mp_zero(q);

    return MP_OKAY;

  } else if(cmp == 0) {

    /* Set quotient to 1, with appropriate sign */
    if(q) {
      int qneg = (SIGN(a) != SIGN(b));

      mp_set(q, 1);
      if(qneg)
	SIGN(q) = MP_NEG;
    }

    if(r)
      mp_zero(r);

    return MP_OKAY;
  }

  /* If we get here, it means we actually have to do some division */

  /* Set up some temporaries... */
  if((res = mp_init_copy(&qtmp, a)) != MP_OKAY)
    return res;
  if((res = mp_init_copy(&rtmp, b)) != MP_OKAY)
    goto CLEANUP;

  if((res = s_mp_div(&qtmp, &rtmp)) != MP_OKAY)
    goto CLEANUP;

  /* Compute the signs for the output  */
  SIGN(&rtmp) = SIGN(a); /* Sr = Sa              */
  if(SIGN(a) == SIGN(b))
    SIGN(&qtmp) = MP_ZPOS;  /* Sq = MP_ZPOS if Sa = Sb */
  else
    SIGN(&qtmp) = MP_NEG;   /* Sq = MP_NEG if Sa != Sb */

  if(s_mp_cmp_d(&qtmp, 0) == MP_EQ)
    SIGN(&qtmp) = MP_ZPOS;
  if(s_mp_cmp_d(&rtmp, 0) == MP_EQ)
    SIGN(&rtmp) = MP_ZPOS;

  /* Copy output, if it is needed      */
  if(q) 
    s_mp_exch(&qtmp, q);

  if(r) 
    s_mp_exch(&rtmp, r);

CLEANUP:
  mp_clear(&rtmp);
  mp_clear(&qtmp);

  return res;

} /* end mp_div() */

/* }}} */

/* {{{ mp_div_2d(a, d, q, r) */

mp_err mp_div_2d(mp_int *a, mp_digit d, mp_int *q, mp_int *r)
{
  mp_err  res;

  ARGCHK(a != NULL, MP_BADARG);

  if(q) {
    if((res = mp_copy(a, q)) != MP_OKAY)
      return res;

    s_mp_div_2d(q, d);
  }

  if(r) {
    if((res = mp_copy(a, r)) != MP_OKAY)
      return res;

    s_mp_mod_2d(r, d);
  }

  return MP_OKAY;

} /* end mp_div_2d() */

/* }}} */

/* {{{ mp_expt(a, b, c) */

/*
  mp_expt(a, b, c)

  Compute c = a ** b, that is, raise a to the b power.  Uses a
  standard iterative square-and-multiply technique.
 */

mp_err mp_expt(mp_int *a, mp_int *b, mp_int *c)
{
  mp_int   s, x;
  mp_err   res;
  mp_digit d;
  int      dig, bit;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  if(mp_cmp_z(b) < 0)
    return MP_RANGE;

  if((res = mp_init(&s)) != MP_OKAY)
    return res;

  mp_set(&s, 1);

  if((res = mp_init_copy(&x, a)) != MP_OKAY)
    goto X;

  /* Loop over low-order digits in ascending order */
  for(dig = 0; dig < (USED(b) - 1); dig++) {
    d = DIGIT(b, dig);

    /* Loop over bits of each non-maximal digit */
    for(bit = 0; bit < DIGIT_BIT; bit++) {
      if(d & 1) {
	if((res = s_mp_mul(&s, &x)) != MP_OKAY) 
	  goto CLEANUP;
      }

      d >>= 1;
      
      if((res = s_mp_sqr(&x)) != MP_OKAY)
	goto CLEANUP;
    }
  }

  /* Consider now the last digit... */
  d = DIGIT(b, dig);

  while(d) {
    if(d & 1) {
      if((res = s_mp_mul(&s, &x)) != MP_OKAY)
	goto CLEANUP;
    }

    d >>= 1;

    if((res = s_mp_sqr(&x)) != MP_OKAY)
      goto CLEANUP;
  }
  
  if(mp_iseven(b))
    SIGN(&s) = SIGN(a);

  res = mp_copy(&s, c);

CLEANUP:
  mp_clear(&x);
X:
  mp_clear(&s);

  return res;

} /* end mp_expt() */

/* }}} */

/* {{{ mp_2expt(a, k) */

/* Compute a = 2^k */

mp_err mp_2expt(mp_int *a, mp_digit k)
{
  ARGCHK(a != NULL, MP_BADARG);

  return s_mp_2expt(a, k);

} /* end mp_2expt() */

/* }}} */

/* {{{ mp_mod(a, m, c) */

/*
  mp_mod(a, m, c)

  Compute c = a (mod m).  Result will always be 0 <= c < m.
 */

mp_err mp_mod(mp_int *a, mp_int *m, mp_int *c)
{
  mp_err  res;
  int     mag;

  ARGCHK(a != NULL && m != NULL && c != NULL, MP_BADARG);

  if(SIGN(m) == MP_NEG)
    return MP_RANGE;

  /*
     If |a| > m, we need to divide to get the remainder and take the
     absolute value.  

     If |a| < m, we don't need to do any division, just copy and adjust
     the sign (if a is negative).

     If |a| == m, we can simply set the result to zero.

     This order is intended to minimize the average path length of the
     comparison chain on common workloads -- the most frequent cases are
     that |a| != m, so we do those first.
   */
  if((mag = s_mp_cmp(a, m)) > 0) {
    if((res = mp_div(a, m, NULL, c)) != MP_OKAY)
      return res;
    
    if(SIGN(c) == MP_NEG) {
      if((res = mp_add(c, m, c)) != MP_OKAY)
	return res;
    }

  } else if(mag < 0) {
    if((res = mp_copy(a, c)) != MP_OKAY)
      return res;

    if(mp_cmp_z(a) < 0) {
      if((res = mp_add(c, m, c)) != MP_OKAY)
	return res;

    }
    
  } else {
    mp_zero(c);

  }

  return MP_OKAY;

} /* end mp_mod() */

/* }}} */

/* {{{ mp_mod_d(a, d, c) */

/*
  mp_mod_d(a, d, c)

  Compute c = a (mod d).  Result will always be 0 <= c < d
 */
mp_err mp_mod_d(mp_int *a, mp_digit d, mp_digit *c)
{
  mp_err   res;
  mp_digit rem;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if(s_mp_cmp_d(a, d) > 0) {
    if((res = mp_div_d(a, d, NULL, &rem)) != MP_OKAY)
      return res;

  } else {
    if(SIGN(a) == MP_NEG)
      rem = d - DIGIT(a, 0);
    else
      rem = DIGIT(a, 0);
  }

  if(c)
    *c = rem;

  return MP_OKAY;

} /* end mp_mod_d() */

/* }}} */

/* {{{ mp_sqrt(a, b) */

/*
  mp_sqrt(a, b)

  Compute the integer square root of a, and store the result in b.
  Uses an integer-arithmetic version of Newton's iterative linear
  approximation technique to determine this value; the result has the
  following two properties:

     b^2 <= a
     (b+1)^2 >= a

  It is a range error to pass a negative value.
 */
mp_err mp_sqrt(mp_int *a, mp_int *b)
{
  mp_int   x, t;
  mp_err   res;

  ARGCHK(a != NULL && b != NULL, MP_BADARG);

  /* Cannot take square root of a negative value */
  if(SIGN(a) == MP_NEG)
    return MP_RANGE;

  /* Special cases for zero and one, trivial     */
  if(mp_cmp_d(a, 0) == MP_EQ || mp_cmp_d(a, 1) == MP_EQ) 
    return mp_copy(a, b);
    
  /* Initialize the temporaries we'll use below  */
  if((res = mp_init_size(&t, USED(a))) != MP_OKAY)
    return res;

  /* Compute an initial guess for the iteration as a itself */
  if((res = mp_init_copy(&x, a)) != MP_OKAY)
    goto X;

s_mp_rshd(&x, (USED(&x)/2)+1);
mp_add_d(&x, 1, &x);

  for(;;) {
    /* t = (x * x) - a */
    mp_copy(&x, &t);      /* can't fail, t is big enough for original x */
    if((res = mp_sqr(&t, &t)) != MP_OKAY ||
       (res = mp_sub(&t, a, &t)) != MP_OKAY)
      goto CLEANUP;

    /* t = t / 2x       */
    s_mp_mul_2(&x);
    if((res = mp_div(&t, &x, &t, NULL)) != MP_OKAY)
      goto CLEANUP;
    s_mp_div_2(&x);

    /* Terminate the loop, if the quotient is zero */
    if(mp_cmp_z(&t) == MP_EQ)
      break;

    /* x = x - t       */
    if((res = mp_sub(&x, &t, &x)) != MP_OKAY)
      goto CLEANUP;

  }

  /* Copy result to output parameter */
  mp_sub_d(&x, 1, &x);
  s_mp_exch(&x, b);

 CLEANUP:
  mp_clear(&x);
 X:
  mp_clear(&t); 

  return res;

} /* end mp_sqrt() */

/* }}} */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ Modular arithmetic */

#if MP_MODARITH
/* {{{ mp_addmod(a, b, m, c) */

/*
  mp_addmod(a, b, m, c)

  Compute c = (a + b) mod m
 */

mp_err mp_addmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && b != NULL && m != NULL && c != NULL, MP_BADARG);

  if((res = mp_add(a, b, c)) != MP_OKAY)
    return res;
  if((res = mp_mod(c, m, c)) != MP_OKAY)
    return res;

  return MP_OKAY;

}

/* }}} */

/* {{{ mp_submod(a, b, m, c) */

/*
  mp_submod(a, b, m, c)

  Compute c = (a - b) mod m
 */

mp_err mp_submod(mp_int *a, mp_int *b, mp_int *m, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && b != NULL && m != NULL && c != NULL, MP_BADARG);

  if((res = mp_sub(a, b, c)) != MP_OKAY)
    return res;
  if((res = mp_mod(c, m, c)) != MP_OKAY)
    return res;

  return MP_OKAY;

}

/* }}} */

/* {{{ mp_mulmod(a, b, m, c) */

/*
  mp_mulmod(a, b, m, c)

  Compute c = (a * b) mod m
 */

mp_err mp_mulmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && b != NULL && m != NULL && c != NULL, MP_BADARG);

  if((res = mp_mul(a, b, c)) != MP_OKAY)
    return res;
  if((res = mp_mod(c, m, c)) != MP_OKAY)
    return res;

  return MP_OKAY;

}

/* }}} */

/* {{{ mp_sqrmod(a, m, c) */

#if MP_SQUARE
mp_err mp_sqrmod(mp_int *a, mp_int *m, mp_int *c)
{
  mp_err  res;

  ARGCHK(a != NULL && m != NULL && c != NULL, MP_BADARG);

  if((res = mp_sqr(a, c)) != MP_OKAY)
    return res;
  if((res = mp_mod(c, m, c)) != MP_OKAY)
    return res;

  return MP_OKAY;

} /* end mp_sqrmod() */
#endif

/* }}} */

/* {{{ mp_exptmod(a, b, m, c) */

/*
  mp_exptmod(a, b, m, c)

  Compute c = (a ** b) mod m.  Uses a standard square-and-multiply
  method with modular reductions at each step. (This is basically the
  same code as mp_expt(), except for the addition of the reductions)
  
  The modular reductions are done using Barrett's algorithm (see
  s_mp_reduce() below for details)
 */

mp_err mp_exptmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c)
{
  mp_int   s, x, mu;
  mp_err   res;
  mp_digit d, *db = DIGITS(b);
  mp_size  ub = USED(b);
  int      dig, bit;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  if(mp_cmp_z(b) < 0 || mp_cmp_z(m) <= 0)
    return MP_RANGE;

  if((res = mp_init(&s)) != MP_OKAY)
    return res;
  if((res = mp_init_copy(&x, a)) != MP_OKAY)
    goto X;
  if((res = mp_mod(&x, m, &x)) != MP_OKAY ||
     (res = mp_init(&mu)) != MP_OKAY)
    goto MU;

  mp_set(&s, 1);

  /* mu = b^2k / m */
  s_mp_add_d(&mu, 1); 
  s_mp_lshd(&mu, 2 * USED(m));
  if((res = mp_div(&mu, m, &mu, NULL)) != MP_OKAY)
    goto CLEANUP;

  /* Loop over digits of b in ascending order, except highest order */
  for(dig = 0; dig < (ub - 1); dig++) {
    d = *db++;

    /* Loop over the bits of the lower-order digits */
    for(bit = 0; bit < DIGIT_BIT; bit++) {
      if(d & 1) {
	if((res = s_mp_mul(&s, &x)) != MP_OKAY)
	  goto CLEANUP;
	if((res = s_mp_reduce(&s, m, &mu)) != MP_OKAY)
	  goto CLEANUP;
      }

      d >>= 1;

      if((res = s_mp_sqr(&x)) != MP_OKAY)
	goto CLEANUP;
      if((res = s_mp_reduce(&x, m, &mu)) != MP_OKAY)
	goto CLEANUP;
    }
  }

  /* Now do the last digit... */
  d = *db;

  while(d) {
    if(d & 1) {
      if((res = s_mp_mul(&s, &x)) != MP_OKAY)
	goto CLEANUP;
      if((res = s_mp_reduce(&s, m, &mu)) != MP_OKAY)
	goto CLEANUP;
    }

    d >>= 1;

    if((res = s_mp_sqr(&x)) != MP_OKAY)
      goto CLEANUP;
    if((res = s_mp_reduce(&x, m, &mu)) != MP_OKAY)
      goto CLEANUP;
  }

  s_mp_exch(&s, c);

 CLEANUP:
  mp_clear(&mu);
 MU:
  mp_clear(&x);
 X:
  mp_clear(&s);

  return res;

} /* end mp_exptmod() */

/* }}} */

/* {{{ mp_exptmod_d(a, d, m, c) */

mp_err mp_exptmod_d(mp_int *a, mp_digit d, mp_int *m, mp_int *c)
{
  mp_int   s, x;
  mp_err   res;

  ARGCHK(a != NULL && c != NULL, MP_BADARG);

  if((res = mp_init(&s)) != MP_OKAY)
    return res;
  if((res = mp_init_copy(&x, a)) != MP_OKAY)
    goto X;

  mp_set(&s, 1);

  while(d != 0) {
    if(d & 1) {
      if((res = s_mp_mul(&s, &x)) != MP_OKAY ||
	 (res = mp_mod(&s, m, &s)) != MP_OKAY)
	goto CLEANUP;
    }

    d /= 2;

    if((res = s_mp_sqr(&x)) != MP_OKAY ||
       (res = mp_mod(&x, m, &x)) != MP_OKAY)
      goto CLEANUP;
  }

  s_mp_exch(&s, c);

CLEANUP:
  mp_clear(&x);
X:
  mp_clear(&s);

  return res;

} /* end mp_exptmod_d() */

/* }}} */
#endif /* if MP_MODARITH */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ Comparison functions */

/* {{{ mp_cmp_z(a) */

/*
  mp_cmp_z(a)

  Compare a <=> 0.  Returns <0 if a<0, 0 if a=0, >0 if a>0.
 */

int    mp_cmp_z(mp_int *a)
{
  if(SIGN(a) == MP_NEG)
    return MP_LT;
  else if(USED(a) == 1 && DIGIT(a, 0) == 0)
    return MP_EQ;
  else
    return MP_GT;

} /* end mp_cmp_z() */

/* }}} */

/* {{{ mp_cmp_d(a, d) */

/*
  mp_cmp_d(a, d)

  Compare a <=> d.  Returns <0 if a<d, 0 if a=d, >0 if a>d
 */

int    mp_cmp_d(mp_int *a, mp_digit d)
{
  ARGCHK(a != NULL, MP_EQ);

  if(SIGN(a) == MP_NEG)
    return MP_LT;

  return s_mp_cmp_d(a, d);

} /* end mp_cmp_d() */

/* }}} */

/* {{{ mp_cmp(a, b) */

int    mp_cmp(mp_int *a, mp_int *b)
{
  ARGCHK(a != NULL && b != NULL, MP_EQ);

  if(SIGN(a) == SIGN(b)) {
    int  mag;

    if((mag = s_mp_cmp(a, b)) == MP_EQ)
      return MP_EQ;

    if(SIGN(a) == MP_ZPOS)
      return mag;
    else
      return -mag;

  } else if(SIGN(a) == MP_ZPOS) {
    return MP_GT;
  } else {
    return MP_LT;
  }

} /* end mp_cmp() */

/* }}} */

/* {{{ mp_cmp_mag(a, b) */

/*
  mp_cmp_mag(a, b)

  Compares |a| <=> |b|, and returns an appropriate comparison result
 */

int    mp_cmp_mag(mp_int *a, mp_int *b)
{
  ARGCHK(a != NULL && b != NULL, MP_EQ);

  return s_mp_cmp(a, b);

} /* end mp_cmp_mag() */

/* }}} */

/* {{{ mp_cmp_int(a, z) */

/*
  This just converts z to an mp_int, and uses the existing comparison
  routines.  This is sort of inefficient, but it's not clear to me how
  frequently this wil get used anyway.  For small positive constants,
  you can always use mp_cmp_d(), and for zero, there is mp_cmp_z().
 */
int    mp_cmp_int(mp_int *a, long z)
{
  mp_int  tmp;
  int     out;

  ARGCHK(a != NULL, MP_EQ);
  
  mp_init(&tmp); mp_set_int(&tmp, z);
  out = mp_cmp(a, &tmp);
  mp_clear(&tmp);

  return out;

} /* end mp_cmp_int() */

/* }}} */

/* {{{ mp_isodd(a) */

/*
  mp_isodd(a)

  Returns a true (non-zero) value if a is odd, false (zero) otherwise.
 */
int    mp_isodd(mp_int *a)
{
  ARGCHK(a != NULL, 0);

  return (DIGIT(a, 0) & 1);

} /* end mp_isodd() */

/* }}} */

/* {{{ mp_iseven(a) */

int    mp_iseven(mp_int *a)
{
  return !mp_isodd(a);

} /* end mp_iseven() */

/* }}} */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ Number theoretic functions */

#if MP_NUMTH
/* {{{ mp_gcd(a, b, c) */

/*
  Like the old mp_gcd() function, except computes the GCD using the
  binary algorithm due to Josef Stein in 1961 (via Knuth).
 */
mp_err mp_gcd(mp_int *a, mp_int *b, mp_int *c)
{
  mp_err   res;
  mp_int   u, v, t;
  mp_size  k = 0;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  if(mp_cmp_z(a) == MP_EQ && mp_cmp_z(b) == MP_EQ)
      return MP_RANGE;
  if(mp_cmp_z(a) == MP_EQ) {
    return mp_copy(b, c);
  } else if(mp_cmp_z(b) == MP_EQ) {
    return mp_copy(a, c);
  }

  if((res = mp_init(&t)) != MP_OKAY)
    return res;
  if((res = mp_init_copy(&u, a)) != MP_OKAY)
    goto U;
  if((res = mp_init_copy(&v, b)) != MP_OKAY)
    goto V;

  SIGN(&u) = MP_ZPOS;
  SIGN(&v) = MP_ZPOS;

  /* Divide out common factors of 2 until at least 1 of a, b is even */
  while(mp_iseven(&u) && mp_iseven(&v)) {
    s_mp_div_2(&u);
    s_mp_div_2(&v);
    ++k;
  }

  /* Initialize t */
  if(mp_isodd(&u)) {
    if((res = mp_copy(&v, &t)) != MP_OKAY)
      goto CLEANUP;
    
    /* t = -v */
    if(SIGN(&v) == MP_ZPOS)
      SIGN(&t) = MP_NEG;
    else
      SIGN(&t) = MP_ZPOS;
    
  } else {
    if((res = mp_copy(&u, &t)) != MP_OKAY)
      goto CLEANUP;

  }

  for(;;) {
    while(mp_iseven(&t)) {
      s_mp_div_2(&t);
    }

    if(mp_cmp_z(&t) == MP_GT) {
      if((res = mp_copy(&t, &u)) != MP_OKAY)
	goto CLEANUP;

    } else {
      if((res = mp_copy(&t, &v)) != MP_OKAY)
	goto CLEANUP;

      /* v = -t */
      if(SIGN(&t) == MP_ZPOS)
	SIGN(&v) = MP_NEG;
      else
	SIGN(&v) = MP_ZPOS;
    }

    if((res = mp_sub(&u, &v, &t)) != MP_OKAY)
      goto CLEANUP;

    if(s_mp_cmp_d(&t, 0) == MP_EQ)
      break;
  }

  s_mp_2expt(&v, k);       /* v = 2^k   */
  res = mp_mul(&u, &v, c); /* c = u * v */

 CLEANUP:
  mp_clear(&v);
 V:
  mp_clear(&u);
 U:
  mp_clear(&t);

  return res;

} /* end mp_bgcd() */

/* }}} */

/* {{{ mp_lcm(a, b, c) */

/* We compute the least common multiple using the rule:

   ab = [a, b](a, b)

   ... by computing the product, and dividing out the gcd.
 */

mp_err mp_lcm(mp_int *a, mp_int *b, mp_int *c)
{
  mp_int  gcd, prod;
  mp_err  res;

  ARGCHK(a != NULL && b != NULL && c != NULL, MP_BADARG);

  /* Set up temporaries */
  if((res = mp_init(&gcd)) != MP_OKAY)
    return res;
  if((res = mp_init(&prod)) != MP_OKAY)
    goto GCD;

  if((res = mp_mul(a, b, &prod)) != MP_OKAY)
    goto CLEANUP;
  if((res = mp_gcd(a, b, &gcd)) != MP_OKAY)
    goto CLEANUP;

  res = mp_div(&prod, &gcd, c, NULL);

 CLEANUP:
  mp_clear(&prod);
 GCD:
  mp_clear(&gcd);

  return res;

} /* end mp_lcm() */

/* }}} */

/* {{{ mp_xgcd(a, b, g, x, y) */

/*
  mp_xgcd(a, b, g, x, y)

  Compute g = (a, b) and values x and y satisfying Bezout's identity
  (that is, ax + by = g).  This uses the extended binary GCD algorithm
  based on the Stein algorithm used for mp_gcd()
 */

mp_err mp_xgcd(mp_int *a, mp_int *b, mp_int *g, mp_int *x, mp_int *y)
{
  mp_int   gx, xc, yc, u, v, A, B, C, D;
  mp_int  *clean[9];
  mp_err   res;
  int      last = -1;

  if(mp_cmp_z(b) == 0)
    return MP_RANGE;

  /* Initialize all these variables we need */
  if((res = mp_init(&u)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &u;
  if((res = mp_init(&v)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &v;
  if((res = mp_init(&gx)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &gx;
  if((res = mp_init(&A)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &A;
  if((res = mp_init(&B)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &B;
  if((res = mp_init(&C)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &C;
  if((res = mp_init(&D)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &D;
  if((res = mp_init_copy(&xc, a)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &xc;
  mp_abs(&xc, &xc);
  if((res = mp_init_copy(&yc, b)) != MP_OKAY) goto CLEANUP;
  clean[++last] = &yc;
  mp_abs(&yc, &yc);

  mp_set(&gx, 1);

  /* Divide by two until at least one of them is even */
  while(mp_iseven(&xc) && mp_iseven(&yc)) {
    s_mp_div_2(&xc);
    s_mp_div_2(&yc);
    if((res = s_mp_mul_2(&gx)) != MP_OKAY)
      goto CLEANUP;
  }

  mp_copy(&xc, &u);
  mp_copy(&yc, &v);
  mp_set(&A, 1); mp_set(&D, 1);

  /* Loop through binary GCD algorithm */
  for(;;) {
    while(mp_iseven(&u)) {
      s_mp_div_2(&u);

      if(mp_iseven(&A) && mp_iseven(&B)) {
	s_mp_div_2(&A); s_mp_div_2(&B);
      } else {
	if((res = mp_add(&A, &yc, &A)) != MP_OKAY) goto CLEANUP;
	s_mp_div_2(&A);
	if((res = mp_sub(&B, &xc, &B)) != MP_OKAY) goto CLEANUP;
	s_mp_div_2(&B);
      }
    }

    while(mp_iseven(&v)) {
      s_mp_div_2(&v);

      if(mp_iseven(&C) && mp_iseven(&D)) {
	s_mp_div_2(&C); s_mp_div_2(&D);
      } else {
	if((res = mp_add(&C, &yc, &C)) != MP_OKAY) goto CLEANUP;
	s_mp_div_2(&C);
	if((res = mp_sub(&D, &xc, &D)) != MP_OKAY) goto CLEANUP;
	s_mp_div_2(&D);
      }
    }

    if(mp_cmp(&u, &v) >= 0) {
      if((res = mp_sub(&u, &v, &u)) != MP_OKAY) goto CLEANUP;
      if((res = mp_sub(&A, &C, &A)) != MP_OKAY) goto CLEANUP;
      if((res = mp_sub(&B, &D, &B)) != MP_OKAY) goto CLEANUP;

    } else {
      if((res = mp_sub(&v, &u, &v)) != MP_OKAY) goto CLEANUP;
      if((res = mp_sub(&C, &A, &C)) != MP_OKAY) goto CLEANUP;
      if((res = mp_sub(&D, &B, &D)) != MP_OKAY) goto CLEANUP;

    }

    /* If we're done, copy results to output */
    if(mp_cmp_z(&u) == 0) {
      if(x)
	if((res = mp_copy(&C, x)) != MP_OKAY) goto CLEANUP;

      if(y)
	if((res = mp_copy(&D, y)) != MP_OKAY) goto CLEANUP;
      
      if(g)
	if((res = mp_mul(&gx, &v, g)) != MP_OKAY) goto CLEANUP;

      break;
    }
  }

 CLEANUP:
  while(last >= 0)
    mp_clear(clean[last--]);

  return res;

} /* end mp_xgcd() */

/* }}} */

/* {{{ mp_invmod(a, m, c) */

/*
  mp_invmod(a, m, c)

  Compute c = a^-1 (mod m), if there is an inverse for a (mod m).
  This is equivalent to the question of whether (a, m) = 1.  If not,
  MP_UNDEF is returned, and there is no inverse.
 */

mp_err mp_invmod(mp_int *a, mp_int *m, mp_int *c)
{
  mp_int  g, x;
  mp_err  res;

  ARGCHK(a && m && c, MP_BADARG);

  if(mp_cmp_z(a) == 0 || mp_cmp_z(m) == 0)
    return MP_RANGE;

  if((res = mp_init(&g)) != MP_OKAY)
    return res;
  if((res = mp_init(&x)) != MP_OKAY)
    goto X;

  if((res = mp_xgcd(a, m, &g, &x, NULL)) != MP_OKAY)
    goto CLEANUP;

  if(mp_cmp_d(&g, 1) != MP_EQ) {
    res = MP_UNDEF;
    goto CLEANUP;
  }

  res = mp_mod(&x, m, c);
  SIGN(c) = SIGN(a);

CLEANUP:
  mp_clear(&x);
X:
  mp_clear(&g);

  return res;

} /* end mp_invmod() */

/* }}} */
#endif /* if MP_NUMTH */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ mp_print(mp, ofp) */

#if MP_IOFUNC
/*
  mp_print(mp, ofp)

  Print a textual representation of the given mp_int on the output
  stream 'ofp'.  Output is generated using the internal radix.
 */

void   mp_print(mp_int *mp, FILE *ofp)
{
  int   ix;

  if(mp == NULL || ofp == NULL)
    return;

  fputc((SIGN(mp) == MP_NEG) ? '-' : '+', ofp);

  for(ix = USED(mp) - 1; ix >= 0; ix--) {
    fprintf(ofp, DIGIT_FMT, DIGIT(mp, ix));
  }

} /* end mp_print() */

#endif /* if MP_IOFUNC */

/* }}} */

/*------------------------------------------------------------------------*/
/* {{{ More I/O Functions */

/* {{{ mp_read_signed_bin(mp, str, len) */

/* 
   mp_read_signed_bin(mp, str, len)

   Read in a raw value (base 256) into the given mp_int
 */

mp_err  mp_read_signed_bin(mp_int *mp, unsigned char *str, int len)
{
  mp_err         res;

  ARGCHK(mp != NULL && str != NULL && len > 0, MP_BADARG);

  if((res = mp_read_unsigned_bin(mp, str + 1, len - 1)) == MP_OKAY) {
    /* Get sign from first byte */
    if(str[0])
      SIGN(mp) = MP_NEG;
    else
      SIGN(mp) = MP_ZPOS;
  }

  return res;

} /* end mp_read_signed_bin() */

/* }}} */

/* {{{ mp_signed_bin_size(mp) */

int    mp_signed_bin_size(mp_int *mp)
{
  ARGCHK(mp != NULL, 0);

  return mp_unsigned_bin_size(mp) + 1;

} /* end mp_signed_bin_size() */

/* }}} */

/* {{{ mp_to_signed_bin(mp, str) */

mp_err mp_to_signed_bin(mp_int *mp, unsigned char *str)
{
  ARGCHK(mp != NULL && str != NULL, MP_BADARG);

  /* Caller responsible for allocating enough memory (use mp_raw_size(mp)) */
  str[0] = (char)SIGN(mp);

  return mp_to_unsigned_bin(mp, str + 1);

} /* end mp_to_signed_bin() */

/* }}} */

/* {{{ mp_read_unsigned_bin(mp, str, len) */

/*
  mp_read_unsigned_bin(mp, str, len)

  Read in an unsigned value (base 256) into the given mp_int
 */

mp_err  mp_read_unsigned_bin(mp_int *mp, unsigned char *str, int len)
{
  int     ix;
  mp_err  res;

  ARGCHK(mp != NULL && str != NULL && len > 0, MP_BADARG);

  mp_zero(mp);

  for(ix = 0; ix < len; ix++) {
    if((res = s_mp_mul_2d(mp, CHAR_BIT)) != MP_OKAY)
      return res;

    if((res = mp_add_d(mp, str[ix], mp)) != MP_OKAY)
      return res;
  }
  
  return MP_OKAY;
  
} /* end mp_read_unsigned_bin() */

/* }}} */

/* {{{ mp_unsigned_bin_size(mp) */

int     mp_unsigned_bin_size(mp_int *mp) 
{
  mp_digit   topdig;
  int        count;

  ARGCHK(mp != NULL, 0);

  /* Special case for the value zero */
  if(USED(mp) == 1 && DIGIT(mp, 0) == 0)
    return 1;

  count = (USED(mp) - 1) * sizeof(mp_digit);
  topdig = DIGIT(mp, USED(mp) - 1);

  while(topdig != 0) {
    ++count;
    topdig >>= CHAR_BIT;
  }

  return count;

} /* end mp_unsigned_bin_size() */

/* }}} */

/* {{{ mp_to_unsigned_bin(mp, str) */

mp_err mp_to_unsigned_bin(mp_int *mp, unsigned char *str)
{
  mp_digit      *dp, *end, d;
  unsigned char *spos;

  ARGCHK(mp != NULL && str != NULL, MP_BADARG);

  dp = DIGITS(mp);
  end = dp + USED(mp) - 1;
  spos = str;

  /* Special case for zero, quick test */
  if(dp == end && *dp == 0) {
    *str = '\0';
    return MP_OKAY;
  }

  /* Generate digits in reverse order */
  while(dp < end) {
    int      ix;

    d = *dp;
    for(ix = 0; ix < sizeof(mp_digit); ++ix) {
      *spos = d & UCHAR_MAX;
      d >>= CHAR_BIT;
      ++spos;
    }

    ++dp;
  }

  /* Now handle last digit specially, high order zeroes are not written */
  d = *end;
  while(d != 0) {
    *spos = d & UCHAR_MAX;
    d >>= CHAR_BIT;
    ++spos;
  }

  /* Reverse everything to get digits in the correct order */
  while(--spos > str) {
    unsigned char t = *str;
    *str = *spos;
    *spos = t;

    ++str;
  }

  return MP_OKAY;

} /* end mp_to_unsigned_bin() */

/* }}} */

/* {{{ mp_count_bits(mp) */

int    mp_count_bits(mp_int *mp)
{
  int      len;
  mp_digit d;

  ARGCHK(mp != NULL, MP_BADARG);

  len = DIGIT_BIT * (USED(mp) - 1);
  d = DIGIT(mp, USED(mp) - 1);

  while(d != 0) {
    ++len;
    d >>= 1;
  }

  return len;
  
} /* end mp_count_bits() */

/* }}} */

/* {{{ mp_read_radix(mp, str, radix) */

/*
  mp_read_radix(mp, str, radix)

  Read an integer from the given string, and set mp to the resulting
  value.  The input is presumed to be in base 10.  Leading non-digit
  characters are ignored, and the function reads until a non-digit
  character or the end of the string.
 */

mp_err  mp_read_radix(mp_int *mp, unsigned char *str, int radix)
{
  int     ix = 0, val = 0;
  mp_err  res;
  mp_sign sig = MP_ZPOS;

  ARGCHK(mp != NULL && str != NULL && radix >= 2 && radix <= MAX_RADIX, 
	 MP_BADARG);

  mp_zero(mp);

  /* Skip leading non-digit characters until a digit or '-' or '+' */
  while(str[ix] && 
	(s_mp_tovalue(str[ix], radix) < 0) && 
	str[ix] != '-' &&
	str[ix] != '+') {
    ++ix;
  }

  if(str[ix] == '-') {
    sig = MP_NEG;
    ++ix;
  } else if(str[ix] == '+') {
    sig = MP_ZPOS; /* this is the default anyway... */
    ++ix;
  }

  while((val = s_mp_tovalue(str[ix], radix)) >= 0) {
    if((res = s_mp_mul_d(mp, radix)) != MP_OKAY)
      return res;
    if((res = s_mp_add_d(mp, val)) != MP_OKAY)
      return res;
    ++ix;
  }

  if(s_mp_cmp_d(mp, 0) == MP_EQ)
    SIGN(mp) = MP_ZPOS;
  else
    SIGN(mp) = sig;

  return MP_OKAY;

} /* end mp_read_radix() */

/* }}} */

/* {{{ mp_radix_size(mp, radix) */

int    mp_radix_size(mp_int *mp, int radix)
{
  int  len;
  ARGCHK(mp != NULL, 0);

  len = s_mp_outlen(mp_count_bits(mp), radix) + 1; /* for NUL terminator */

  if(mp_cmp_z(mp) < 0)
    ++len; /* for sign */

  return len;

} /* end mp_radix_size() */

/* }}} */

/* {{{ mp_value_radix_size(num, qty, radix) */

/* num = number of digits
   qty = number of bits per digit
   radix = target base
   
   Return the number of digits in the specified radix that would be
   needed to express 'num' digits of 'qty' bits each.
 */
int    mp_value_radix_size(int num, int qty, int radix)
{
  ARGCHK(num >= 0 && qty > 0 && radix >= 2 && radix <= MAX_RADIX, 0);

  return s_mp_outlen(num * qty, radix);

} /* end mp_value_radix_size() */

/* }}} */

/* {{{ mp_toradix(mp, str, radix) */

mp_err mp_toradix(mp_int *mp, unsigned char *str, int radix)
{
  int  ix, pos = 0;

  ARGCHK(mp != NULL && str != NULL, MP_BADARG);
  ARGCHK(radix > 1 && radix <= MAX_RADIX, MP_RANGE);

  if(mp_cmp_z(mp) == MP_EQ) {
    str[0] = '0';
    str[1] = '\0';
  } else {
    mp_err   res;
    mp_int   tmp;
    mp_sign  sgn;
    mp_digit rem, rdx = (mp_digit)radix;
    char     ch;

    if((res = mp_init_copy(&tmp, mp)) != MP_OKAY)
      return res;

    /* Save sign for later, and take absolute value */
    sgn = SIGN(&tmp); SIGN(&tmp) = MP_ZPOS;

    /* Generate output digits in reverse order      */
    while(mp_cmp_z(&tmp) != 0) {
      if((res = s_mp_div_d(&tmp, rdx, &rem)) != MP_OKAY) {
	mp_clear(&tmp);
	return res;
      }

      /* Generate digits, use capital letters */
      ch = s_mp_todigit(rem, radix, 0);

      str[pos++] = ch;
    }

    /* Add - sign if original value was negative */
    if(sgn == MP_NEG)
      str[pos++] = '-';

    /* Add trailing NUL to end the string        */
    str[pos--] = '\0';

    /* Reverse the digits and sign indicator     */
    ix = 0;
    while(ix < pos) {
      char tmp = str[ix];

      str[ix] = str[pos];
      str[pos] = tmp;
      ++ix;
      --pos;
    }
    
    mp_clear(&tmp);
  }

  return MP_OKAY;

} /* end mp_toradix() */

/* }}} */

/* {{{ mp_char2value(ch, r) */

int    mp_char2value(char ch, int r)
{
  return s_mp_tovalue(ch, r);

} /* end mp_tovalue() */

/* }}} */

/* }}} */

/* {{{ mp_strerror(ec) */

/*
  mp_strerror(ec)

  Return a string describing the meaning of error code 'ec'.  The
  string returned is allocated in static memory, so the caller should
  not attempt to modify or free the memory associated with this
  string.
 */
const char  *mp_strerror(mp_err ec)
{
  int   aec = (ec < 0) ? -ec : ec;

  /* Code values are negative, so the senses of these comparisons
     are accurate */
  if(ec < MP_LAST_CODE || ec > MP_OKAY) {
    return mp_err_string[0];  /* unknown error code */
  } else {
    return mp_err_string[aec + 1];
  }

} /* end mp_strerror() */

/* }}} */

/*========================================================================*/
/*------------------------------------------------------------------------*/
/* Static function definitions (internal use only)                        */

/* {{{ Memory management */

/* {{{ s_mp_grow(mp, min) */

/* Make sure there are at least 'min' digits allocated to mp              */
mp_err   s_mp_grow(mp_int *mp, mp_size min)
{
  if(min > ALLOC(mp)) {
    mp_digit   *tmp;

    /* Set min to next nearest default precision block size */
    min = ((min + (s_mp_defprec - 1)) / s_mp_defprec) * s_mp_defprec;

    if((tmp = s_mp_alloc(min, sizeof(mp_digit))) == NULL)
      return MP_MEM;

    s_mp_copy(DIGITS(mp), tmp, USED(mp));

#if MP_CRYPTO
    s_mp_setz(DIGITS(mp), ALLOC(mp));
#endif
    s_mp_free(DIGITS(mp));
    DIGITS(mp) = tmp;
    ALLOC(mp) = min;
  }

  return MP_OKAY;

} /* end s_mp_grow() */

/* }}} */

/* {{{ s_mp_pad(mp, min) */

/* Make sure the used size of mp is at least 'min', growing if needed     */
mp_err   s_mp_pad(mp_int *mp, mp_size min)
{
  if(min > USED(mp)) {
    mp_err  res;

    /* Make sure there is room to increase precision  */
    if(min > ALLOC(mp) && (res = s_mp_grow(mp, min)) != MP_OKAY)
      return res;

    /* Increase precision; should already be 0-filled */
    USED(mp) = min;
  }

  return MP_OKAY;

} /* end s_mp_pad() */

/* }}} */

/* {{{ s_mp_setz(dp, count) */

#if MP_MACRO == 0
/* Set 'count' digits pointed to by dp to be zeroes                       */
void s_mp_setz(mp_digit *dp, mp_size count)
{
#if MP_MEMSET == 0
  int  ix;

  for(ix = 0; ix < count; ix++)
    dp[ix] = 0;
#else
  memset(dp, 0, count * sizeof(mp_digit));
#endif

} /* end s_mp_setz() */
#endif

/* }}} */

/* {{{ s_mp_copy(sp, dp, count) */

#if MP_MACRO == 0
/* Copy 'count' digits from sp to dp                                      */
void s_mp_copy(mp_digit *sp, mp_digit *dp, mp_size count)
{
#if MP_MEMCPY == 0
  int  ix;

  for(ix = 0; ix < count; ix++)
    dp[ix] = sp[ix];
#else
  memcpy(dp, sp, count * sizeof(mp_digit));
#endif

} /* end s_mp_copy() */
#endif

/* }}} */

/* {{{ s_mp_alloc(nb, ni) */

#if MP_MACRO == 0
/* Allocate ni records of nb bytes each, and return a pointer to that     */
void    *s_mp_alloc(size_t nb, size_t ni)
{
  return calloc(nb, ni);

} /* end s_mp_alloc() */
#endif

/* }}} */

/* {{{ s_mp_free(ptr) */

#if MP_MACRO == 0
/* Free the memory pointed to by ptr                                      */
void     s_mp_free(void *ptr)
{
  if(ptr)
    free(ptr);

} /* end s_mp_free() */
#endif

/* }}} */

/* {{{ s_mp_clamp(mp) */

/* Remove leading zeroes from the given value                             */
void     s_mp_clamp(mp_int *mp)
{
  mp_size   du = USED(mp);
  mp_digit *zp = DIGITS(mp) + du - 1;

  while(du > 1 && !*zp--)
    --du;

  USED(mp) = du;

} /* end s_mp_clamp() */


/* }}} */

/* {{{ s_mp_exch(a, b) */

/* Exchange the data for a and b; (b, a) = (a, b)                         */
void     s_mp_exch(mp_int *a, mp_int *b)
{
  mp_int   tmp;

  tmp = *a;
  *a = *b;
  *b = tmp;

} /* end s_mp_exch() */

/* }}} */

/* }}} */

/* {{{ Arithmetic helpers */

/* {{{ s_mp_lshd(mp, p) */

/* 
   Shift mp leftward by p digits, growing if needed, and zero-filling
   the in-shifted digits at the right end.  This is a convenient
   alternative to multiplication by powers of the radix
 */   

mp_err   s_mp_lshd(mp_int *mp, mp_size p)
{
  mp_err   res;
  mp_size  pos;
  mp_digit *dp;
  int     ix;

  if(p == 0)
    return MP_OKAY;

  if((res = s_mp_pad(mp, USED(mp) + p)) != MP_OKAY)
    return res;

  pos = USED(mp) - 1;
  dp = DIGITS(mp);

  /* Shift all the significant figures over as needed */
  for(ix = pos - p; ix >= 0; ix--) 
    dp[ix + p] = dp[ix];

  /* Fill the bottom digits with zeroes */
  for(ix = 0; ix < p; ix++)
    dp[ix] = 0;

  return MP_OKAY;

} /* end s_mp_lshd() */

/* }}} */

/* {{{ s_mp_rshd(mp, p) */

/* 
   Shift mp rightward by p digits.  Maintains the invariant that
   digits above the precision are all zero.  Digits shifted off the
   end are lost.  Cannot fail.
 */

void     s_mp_rshd(mp_int *mp, mp_size p)
{
  mp_size  ix;
  mp_digit *dp;

  if(p == 0)
    return;

  /* Shortcut when all digits are to be shifted off */
  if(p >= USED(mp)) {
    s_mp_setz(DIGITS(mp), ALLOC(mp));
    USED(mp) = 1;
    SIGN(mp) = MP_ZPOS;
    return;
  }

  /* Shift all the significant figures over as needed */
  dp = DIGITS(mp);
  for(ix = p; ix < USED(mp); ix++)
    dp[ix - p] = dp[ix];

  /* Fill the top digits with zeroes */
  ix -= p;
  while(ix < USED(mp))
    dp[ix++] = 0;

  /* Strip off any leading zeroes    */
  s_mp_clamp(mp);

} /* end s_mp_rshd() */

/* }}} */

/* {{{ s_mp_div_2(mp) */

/* Divide by two -- take advantage of radix properties to do it fast      */
void     s_mp_div_2(mp_int *mp)
{
  s_mp_div_2d(mp, 1);

} /* end s_mp_div_2() */

/* }}} */

/* {{{ s_mp_mul_2(mp) */

mp_err s_mp_mul_2(mp_int *mp)
{
  int      ix;
  mp_digit kin = 0, kout, *dp = DIGITS(mp);
  mp_err   res;

  /* Shift digits leftward by 1 bit */
  for(ix = 0; ix < USED(mp); ix++) {
    kout = (dp[ix] >> (DIGIT_BIT - 1)) & 1;
    dp[ix] = (dp[ix] << 1) | kin;

    kin = kout;
  }

  /* Deal with rollover from last digit */
  if(kin) {
    if(ix >= ALLOC(mp)) {
      if((res = s_mp_grow(mp, ALLOC(mp) + 1)) != MP_OKAY)
	return res;
      dp = DIGITS(mp);
    }

    dp[ix] = kin;
    USED(mp) += 1;
  }

  return MP_OKAY;

} /* end s_mp_mul_2() */

/* }}} */

/* {{{ s_mp_mod_2d(mp, d) */

/*
  Remainder the integer by 2^d, where d is a number of bits.  This
  amounts to a bitwise AND of the value, and does not require the full
  division code
 */
void     s_mp_mod_2d(mp_int *mp, mp_digit d)
{
  unsigned int  ndig = (d / DIGIT_BIT), nbit = (d % DIGIT_BIT);
  unsigned int  ix;
  mp_digit      dmask, *dp = DIGITS(mp);

  if(ndig >= USED(mp))
    return;

  /* Flush all the bits above 2^d in its digit */
  dmask = (1 << nbit) - 1;
  dp[ndig] &= dmask;

  /* Flush all digits above the one with 2^d in it */
  for(ix = ndig + 1; ix < USED(mp); ix++)
    dp[ix] = 0;

  s_mp_clamp(mp);

} /* end s_mp_mod_2d() */

/* }}} */

/* {{{ s_mp_mul_2d(mp, d) */

/*
  Multiply by the integer 2^d, where d is a number of bits.  This
  amounts to a bitwise shift of the value, and does not require the
  full multiplication code.
 */
mp_err    s_mp_mul_2d(mp_int *mp, mp_digit d)
{
  mp_err   res;
  mp_digit save, next, mask, *dp;
  mp_size  used;
  int      ix;

  if((res = s_mp_lshd(mp, d / DIGIT_BIT)) != MP_OKAY)
    return res;

  dp = DIGITS(mp); used = USED(mp);
  d %= DIGIT_BIT;

  mask = (1 << d) - 1;

  /* If the shift requires another digit, make sure we've got one to
     work with */
  if((dp[used - 1] >> (DIGIT_BIT - d)) & mask) {
    if((res = s_mp_grow(mp, used + 1)) != MP_OKAY)
      return res;
    dp = DIGITS(mp);
  }

  /* Do the shifting... */
  save = 0;
  for(ix = 0; ix < used; ix++) {
    next = (dp[ix] >> (DIGIT_BIT - d)) & mask;
    dp[ix] = (dp[ix] << d) | save;
    save = next;
  }

  /* If, at this point, we have a nonzero carryout into the next
     digit, we'll increase the size by one digit, and store it...
   */
  if(save) {
    dp[used] = save;
    USED(mp) += 1;
  }

  s_mp_clamp(mp);
  return MP_OKAY;

} /* end s_mp_mul_2d() */

/* }}} */

/* {{{ s_mp_div_2d(mp, d) */

/*
  Divide the integer by 2^d, where d is a number of bits.  This
  amounts to a bitwise shift of the value, and does not require the
  full division code (used in Barrett reduction, see below)
 */
void     s_mp_div_2d(mp_int *mp, mp_digit d)
{
  int       ix;
  mp_digit  save, next, mask, *dp = DIGITS(mp);

  s_mp_rshd(mp, d / DIGIT_BIT);
  d %= DIGIT_BIT;

  mask = (1 << d) - 1;

  save = 0;
  for(ix = USED(mp) - 1; ix >= 0; ix--) {
    next = dp[ix] & mask;
    dp[ix] = (dp[ix] >> d) | (save << (DIGIT_BIT - d));
    save = next;
  }

  s_mp_clamp(mp);

} /* end s_mp_div_2d() */

/* }}} */

/* {{{ s_mp_norm(a, b) */

/*
  s_mp_norm(a, b)

  Normalize a and b for division, where b is the divisor.  In order
  that we might make good guesses for quotient digits, we want the
  leading digit of b to be at least half the radix, which we
  accomplish by multiplying a and b by a constant.  This constant is
  returned (so that it can be divided back out of the remainder at the
  end of the division process).

  We multiply by the smallest power of 2 that gives us a leading digit
  at least half the radix.  By choosing a power of 2, we simplify the 
  multiplication and division steps to simple shifts.
 */
mp_digit s_mp_norm(mp_int *a, mp_int *b)
{
  mp_digit  t, d = 0;

  t = DIGIT(b, USED(b) - 1);
  while(t < (RADIX / 2)) {
    t <<= 1;
    ++d;
  }
    
  if(d != 0) {
    s_mp_mul_2d(a, d);
    s_mp_mul_2d(b, d);
  }

  return d;

} /* end s_mp_norm() */

/* }}} */

/* }}} */

/* {{{ Primitive digit arithmetic */

/* {{{ s_mp_add_d(mp, d) */

/* Add d to |mp| in place                                                 */
mp_err   s_mp_add_d(mp_int *mp, mp_digit d)    /* unsigned digit addition */
{
  mp_word   w, k = 0;
  mp_size   ix = 1, used = USED(mp);
  mp_digit *dp = DIGITS(mp);

  w = dp[0] + d;
  dp[0] = ACCUM(w);
  k = CARRYOUT(w);

  while(ix < used && k) {
    w = dp[ix] + k;
    dp[ix] = ACCUM(w);
    k = CARRYOUT(w);
    ++ix;
  }

  if(k != 0) {
    mp_err  res;

    if((res = s_mp_pad(mp, USED(mp) + 1)) != MP_OKAY)
      return res;

    DIGIT(mp, ix) = k;
  }

  return MP_OKAY;

} /* end s_mp_add_d() */

/* }}} */

/* {{{ s_mp_sub_d(mp, d) */

/* Subtract d from |mp| in place, assumes |mp| > d                        */
mp_err   s_mp_sub_d(mp_int *mp, mp_digit d)    /* unsigned digit subtract */
{
  mp_word   w, b = 0;
  mp_size   ix = 1, used = USED(mp);
  mp_digit *dp = DIGITS(mp);

  /* Compute initial subtraction    */
  w = (RADIX + dp[0]) - d;
  b = CARRYOUT(w) ? 0 : 1;
  dp[0] = ACCUM(w);

  /* Propagate borrows leftward     */
  while(b && ix < used) {
    w = (RADIX + dp[ix]) - b;
    b = CARRYOUT(w) ? 0 : 1;
    dp[ix] = ACCUM(w);
    ++ix;
  }

  /* Remove leading zeroes          */
  s_mp_clamp(mp);

  /* If we have a borrow out, it's a violation of the input invariant */
  if(b)
    return MP_RANGE;
  else
    return MP_OKAY;

} /* end s_mp_sub_d() */

/* }}} */

/* {{{ s_mp_mul_d(a, d) */

/* Compute a = a * d, single digit multiplication                         */
mp_err   s_mp_mul_d(mp_int *a, mp_digit d)
{
  mp_word w, k = 0;
  mp_size ix, max;
  mp_err  res;
  mp_digit *dp = DIGITS(a);

  /*
    Single-digit multiplication will increase the precision of the
    output by at most one digit.  However, we can detect when this
    will happen -- if the high-order digit of a, times d, gives a
    two-digit result, then the precision of the result will increase;
    otherwise it won't.  We use this fact to avoid calling s_mp_pad()
    unless absolutely necessary.
   */
  max = USED(a);
  w = dp[max - 1] * d;
  if(CARRYOUT(w) != 0) {
    if((res = s_mp_pad(a, max + 1)) != MP_OKAY)
      return res;
    dp = DIGITS(a);
  }

  for(ix = 0; ix < max; ix++) {
    w = (dp[ix] * d) + k;
    dp[ix] = ACCUM(w);
    k = CARRYOUT(w);
  }

  /* If there is a precision increase, take care of it here; the above
     test guarantees we have enough storage to do this safely.
   */
  if(k) {
    dp[max] = k; 
    USED(a) = max + 1;
  }

  s_mp_clamp(a);

  return MP_OKAY;
  
} /* end s_mp_mul_d() */

/* }}} */

/* {{{ s_mp_div_d(mp, d, r) */

/*
  s_mp_div_d(mp, d, r)

  Compute the quotient mp = mp / d and remainder r = mp mod d, for a
  single digit d.  If r is null, the remainder will be discarded.
 */

mp_err   s_mp_div_d(mp_int *mp, mp_digit d, mp_digit *r)
{
  mp_word   w = 0, t;
  mp_int    quot;
  mp_err    res;
  mp_digit *dp = DIGITS(mp), *qp;
  int       ix;

  if(d == 0)
    return MP_RANGE;

  /* Make room for the quotient */
  if((res = mp_init_size(&quot, USED(mp))) != MP_OKAY)
    return res;

  USED(&quot) = USED(mp); /* so clamping will work below */
  qp = DIGITS(&quot);

  /* Divide without subtraction */
  for(ix = USED(mp) - 1; ix >= 0; ix--) {
    w = (w << DIGIT_BIT) | dp[ix];

    if(w >= d) {
      t = w / d;
      w = w % d;
    } else {
      t = 0;
    }

    qp[ix] = t;
  }

  /* Deliver the remainder, if desired */
  if(r)
    *r = w;

  s_mp_clamp(&quot);
  mp_exch(&quot, mp);
  mp_clear(&quot);

  return MP_OKAY;

} /* end s_mp_div_d() */

/* }}} */

/* }}} */

/* {{{ Primitive full arithmetic */

/* {{{ s_mp_add(a, b) */

/* Compute a = |a| + |b|                                                  */
mp_err   s_mp_add(mp_int *a, mp_int *b)        /* magnitude addition      */
{
  mp_word   w = 0;
  mp_digit *pa, *pb;
  mp_size   ix, used = USED(b);
  mp_err    res;

  /* Make sure a has enough precision for the output value */
  if((used > USED(a)) && (res = s_mp_pad(a, used)) != MP_OKAY)
    return res;

  /*
    Add up all digits up to the precision of b.  If b had initially
    the same precision as a, or greater, we took care of it by the
    padding step above, so there is no problem.  If b had initially
    less precision, we'll have to make sure the carry out is duly
    propagated upward among the higher-order digits of the sum.
   */
  pa = DIGITS(a);
  pb = DIGITS(b);
  for(ix = 0; ix < used; ++ix) {
    w += *pa + *pb++;
    *pa++ = ACCUM(w);
    w = CARRYOUT(w);
  }

  /* If we run out of 'b' digits before we're actually done, make
     sure the carries get propagated upward...  
   */
  used = USED(a);
  while(w && ix < used) {
    w += *pa;
    *pa++ = ACCUM(w);
    w = CARRYOUT(w);
    ++ix;
  }

  /* If there's an overall carry out, increase precision and include
     it.  We could have done this initially, but why touch the memory
     allocator unless we're sure we have to?
   */
  if(w) {
    if((res = s_mp_pad(a, used + 1)) != MP_OKAY)
      return res;

    DIGIT(a, ix) = w;  /* pa may not be valid after s_mp_pad() call */
  }

  return MP_OKAY;

} /* end s_mp_add() */

/* }}} */

/* {{{ s_mp_sub(a, b) */

/* Compute a = |a| - |b|, assumes |a| >= |b|                              */
mp_err   s_mp_sub(mp_int *a, mp_int *b)        /* magnitude subtract      */
{
  mp_word   w = 0;
  mp_digit *pa, *pb;
  mp_size   ix, used = USED(b);

  /*
    Subtract and propagate borrow.  Up to the precision of b, this
    accounts for the digits of b; after that, we just make sure the
    carries get to the right place.  This saves having to pad b out to
    the precision of a just to make the loops work right...
   */
  pa = DIGITS(a);
  pb = DIGITS(b);

  for(ix = 0; ix < used; ++ix) {
    w = (RADIX + *pa) - w - *pb++;
    *pa++ = ACCUM(w);
    w = CARRYOUT(w) ? 0 : 1;
  }

  used = USED(a);
  while(ix < used) {
    w = RADIX + *pa - w;
    *pa++ = ACCUM(w);
    w = CARRYOUT(w) ? 0 : 1;
    ++ix;
  }

  /* Clobber any leading zeroes we created    */
  s_mp_clamp(a);

  /* 
     If there was a borrow out, then |b| > |a| in violation
     of our input invariant.  We've already done the work,
     but we'll at least complain about it...
   */
  if(w)
    return MP_RANGE;
  else
    return MP_OKAY;

} /* end s_mp_sub() */

/* }}} */

mp_err   s_mp_reduce(mp_int *x, mp_int *m, mp_int *mu)
{
  mp_int   q;
  mp_err   res;
  mp_size  um = USED(m);

  if((res = mp_init_copy(&q, x)) != MP_OKAY)
    return res;

  s_mp_rshd(&q, um - 1);       /* q1 = x / b^(k-1)  */
  s_mp_mul(&q, mu);            /* q2 = q1 * mu      */
  s_mp_rshd(&q, um + 1);       /* q3 = q2 / b^(k+1) */

  /* x = x mod b^(k+1), quick (no division) */
  s_mp_mod_2d(x, (mp_digit)(DIGIT_BIT * (um + 1)));

  /* q = q * m mod b^(k+1), quick (no division), uses the short multiplier */
#ifndef SHRT_MUL
  s_mp_mul(&q, m);
  s_mp_mod_2d(&q, (mp_digit)(DIGIT_BIT * (um + 1)));
#else
  s_mp_mul_dig(&q, m, um + 1);
#endif  

  /* x = x - q */
  if((res = mp_sub(x, &q, x)) != MP_OKAY)
    goto CLEANUP;

  /* If x < 0, add b^(k+1) to it */
  if(mp_cmp_z(x) < 0) {
    mp_set(&q, 1);
    if((res = s_mp_lshd(&q, um + 1)) != MP_OKAY)
      goto CLEANUP;
    if((res = mp_add(x, &q, x)) != MP_OKAY)
      goto CLEANUP;
  }

  /* Back off if it's too big */
  while(mp_cmp(x, m) >= 0) {
    if((res = s_mp_sub(x, m)) != MP_OKAY)
      break;
  }

 CLEANUP:
  mp_clear(&q);

  return res;

} /* end s_mp_reduce() */



/* {{{ s_mp_mul(a, b) */

/* Compute a = |a| * |b|                                                  */
mp_err   s_mp_mul(mp_int *a, mp_int *b)
{
  mp_word   w, k = 0;
  mp_int    tmp;
  mp_err    res;
  mp_size   ix, jx, ua = USED(a), ub = USED(b);
  mp_digit *pa, *pb, *pt, *pbt;

  if((res = mp_init_size(&tmp, ua + ub)) != MP_OKAY)
    return res;

  /* This has the effect of left-padding with zeroes... */
  USED(&tmp) = ua + ub;

  /* We're going to need the base value each iteration */
  pbt = DIGITS(&tmp);

  /* Outer loop:  Digits of b */

  pb = DIGITS(b);
  for(ix = 0; ix < ub; ++ix, ++pb) {
    if(*pb == 0) 
      continue;

    /* Inner product:  Digits of a */
    pa = DIGITS(a);
    for(jx = 0; jx < ua; ++jx, ++pa) {
      pt = pbt + ix + jx;
      w = *pb * *pa + k + *pt;
      *pt = ACCUM(w);
      k = CARRYOUT(w);
    }

    pbt[ix + jx] = k;
    k = 0;
  }

  s_mp_clamp(&tmp);
  s_mp_exch(&tmp, a);

  mp_clear(&tmp);

  return MP_OKAY;

} /* end s_mp_mul() */

/* }}} */

/* {{{ s_mp_kmul(a, b, out, len) */

#if 0
void   s_mp_kmul(mp_digit *a, mp_digit *b, mp_digit *out, mp_size len)
{
  mp_word   w, k = 0;
  mp_size   ix, jx;
  mp_digit *pa, *pt;

  for(ix = 0; ix < len; ++ix, ++b) {
    if(*b == 0)
      continue;
    
    pa = a;
    for(jx = 0; jx < len; ++jx, ++pa) {
      pt = out + ix + jx;
      w = *b * *pa + k + *pt;
      *pt = ACCUM(w);
      k = CARRYOUT(w);
    }

    out[ix + jx] = k;
    k = 0;
  }

} /* end s_mp_kmul() */
#endif

/* }}} */

/* {{{ s_mp_sqr(a) */

/*
  Computes the square of a, in place.  This can be done more
  efficiently than a general multiplication, because many of the
  computation steps are redundant when squaring.  The inner product
  step is a bit more complicated, but we save a fair number of
  iterations of the multiplication loop.
 */
#if MP_SQUARE
mp_err   s_mp_sqr(mp_int *a)
{
  mp_word  w, k = 0;
  mp_int   tmp;
  mp_err   res;
  mp_size  ix, jx, kx, used = USED(a);
  mp_digit *pa1, *pa2, *pt, *pbt;

  if((res = mp_init_size(&tmp, 2 * used)) != MP_OKAY)
    return res;

  /* Left-pad with zeroes */
  USED(&tmp) = 2 * used;

  /* We need the base value each time through the loop */
  pbt = DIGITS(&tmp);

  pa1 = DIGITS(a);
  for(ix = 0; ix < used; ++ix, ++pa1) {
    if(*pa1 == 0)
      continue;

    w = DIGIT(&tmp, ix + ix) + (*pa1 * *pa1);

    pbt[ix + ix] = ACCUM(w);
    k = CARRYOUT(w);

    /*
      The inner product is computed as:

         (C, S) = t[i,j] + 2 a[i] a[j] + C

      This can overflow what can be represented in an mp_word, and
      since C arithmetic does not provide any way to check for
      overflow, we have to check explicitly for overflow conditions
      before they happen.
     */
    for(jx = ix + 1, pa2 = DIGITS(a) + jx; jx < used; ++jx, ++pa2) {
      mp_word  u = 0, v;
      
      /* Store this in a temporary to avoid indirections later */
      pt = pbt + ix + jx;

      /* Compute the multiplicative step */
      w = *pa1 * *pa2;

      /* If w is more than half MP_WORD_MAX, the doubling will
	 overflow, and we need to record a carry out into the next
	 word */
      u = (w >> (MP_WORD_BIT - 1)) & 1;

      /* Double what we've got, overflow will be ignored as defined
	 for C arithmetic (we've already noted if it is to occur)
       */
      w *= 2;

      /* Compute the additive step */
      v = *pt + k;

      /* If we do not already have an overflow carry, check to see
	 if the addition will cause one, and set the carry out if so 
       */
      u |= ((MP_WORD_MAX - v) < w);

      /* Add in the rest, again ignoring overflow */
      w += v;

      /* Set the i,j digit of the output */
      *pt = ACCUM(w);

      /* Save carry information for the next iteration of the loop.
	 This is why k must be an mp_word, instead of an mp_digit */
      k = CARRYOUT(w) | (u << DIGIT_BIT);

    } /* for(jx ...) */

    /* Set the last digit in the cycle and reset the carry */
    k = DIGIT(&tmp, ix + jx) + k;
    pbt[ix + jx] = ACCUM(k);
    k = CARRYOUT(k);

    /* If we are carrying out, propagate the carry to the next digit
       in the output.  This may cascade, so we have to be somewhat
       circumspect -- but we will have enough precision in the output
       that we won't overflow 
     */
    kx = 1;
    while(k) {
      k = pbt[ix + jx + kx] + 1;
      pbt[ix + jx + kx] = ACCUM(k);
      k = CARRYOUT(k);
      ++kx;
    }
  } /* for(ix ...) */

  s_mp_clamp(&tmp);
  s_mp_exch(&tmp, a);

  mp_clear(&tmp);

  return MP_OKAY;

} /* end s_mp_sqr() */
#endif

/* }}} */

/* {{{ s_mp_div(a, b) */

/*
  s_mp_div(a, b)

  Compute a = a / b and b = a mod b.  Assumes b > a.
 */

mp_err   s_mp_div(mp_int *a, mp_int *b)
{
  mp_int   quot, rem, t;
  mp_word  q;
  mp_err   res;
  mp_digit d;
  int      ix;

  if(mp_cmp_z(b) == 0)
    return MP_RANGE;

  /* Shortcut if b is power of two */
  if((ix = s_mp_ispow2(b)) >= 0) {
    mp_copy(a, b);  /* need this for remainder */
    s_mp_div_2d(a, (mp_digit)ix);
    s_mp_mod_2d(b, (mp_digit)ix);

    return MP_OKAY;
  }

  /* Allocate space to store the quotient */
  if((res = mp_init_size(&quot, USED(a))) != MP_OKAY)
    return res;

  /* A working temporary for division     */
  if((res = mp_init_size(&t, USED(a))) != MP_OKAY)
    goto T;

  /* Allocate space for the remainder     */
  if((res = mp_init_size(&rem, USED(a))) != MP_OKAY)
    goto REM;

  /* Normalize to optimize guessing       */
  d = s_mp_norm(a, b);

  /* Perform the division itself...woo!   */
  ix = USED(a) - 1;

  while(ix >= 0) {
    /* Find a partial substring of a which is at least b */
    while(s_mp_cmp(&rem, b) < 0 && ix >= 0) {
      if((res = s_mp_lshd(&rem, 1)) != MP_OKAY) 
	goto CLEANUP;

      if((res = s_mp_lshd(&quot, 1)) != MP_OKAY)
	goto CLEANUP;

      DIGIT(&rem, 0) = DIGIT(a, ix);
      s_mp_clamp(&rem);
      --ix;
    }

    /* If we didn't find one, we're finished dividing    */
    if(s_mp_cmp(&rem, b) < 0) 
      break;    

    /* Compute a guess for the next quotient digit       */
    q = DIGIT(&rem, USED(&rem) - 1);
    if(q <= DIGIT(b, USED(b) - 1) && USED(&rem) > 1)
      q = (q << DIGIT_BIT) | DIGIT(&rem, USED(&rem) - 2);

    q /= DIGIT(b, USED(b) - 1);

    /* The guess can be as much as RADIX + 1 */
    if(q >= RADIX)
      q = RADIX - 1;

    /* See what that multiplies out to                   */
    mp_copy(b, &t);
    if((res = s_mp_mul_d(&t, q)) != MP_OKAY)
      goto CLEANUP;

    /* 
       If it's too big, back it off.  We should not have to do this
       more than once, or, in rare cases, twice.  Knuth describes a
       method by which this could be reduced to a maximum of once, but
       I didn't implement that here.
     */
    while(s_mp_cmp(&t, &rem) > 0) {
      --q;
      s_mp_sub(&t, b);
    }

    /* At this point, q should be the right next digit   */
    if((res = s_mp_sub(&rem, &t)) != MP_OKAY)
      goto CLEANUP;

    /*
      Include the digit in the quotient.  We allocated enough memory
      for any quotient we could ever possibly get, so we should not
      have to check for failures here
     */
    DIGIT(&quot, 0) = q;
  }

  /* Denormalize remainder                */
  if(d != 0) 
    s_mp_div_2d(&rem, d);

  s_mp_clamp(&quot);
  s_mp_clamp(&rem);

  /* Copy quotient back to output         */
  s_mp_exch(&quot, a);
  
  /* Copy remainder back to output        */
  s_mp_exch(&rem, b);

CLEANUP:
  mp_clear(&rem);
REM:
  mp_clear(&t);
T:
  mp_clear(&quot);

  return res;

} /* end s_mp_div() */

/* }}} */

/* {{{ s_mp_2expt(a, k) */

mp_err   s_mp_2expt(mp_int *a, mp_digit k)
{
  mp_err    res;
  mp_size   dig, bit;

  dig = k / DIGIT_BIT;
  bit = k % DIGIT_BIT;

  mp_zero(a);
  if((res = s_mp_pad(a, dig + 1)) != MP_OKAY)
    return res;
  
  DIGIT(a, dig) |= (1 << bit);

  return MP_OKAY;

} /* end s_mp_2expt() */

/* }}} */


/* }}} */

/* }}} */

/* {{{ Primitive comparisons */

/* {{{ s_mp_cmp(a, b) */

/* Compare |a| <=> |b|, return 0 if equal, <0 if a<b, >0 if a>b           */
int      s_mp_cmp(mp_int *a, mp_int *b)
{
  mp_size   ua = USED(a), ub = USED(b);

  if(ua > ub)
    return MP_GT;
  else if(ua < ub)
    return MP_LT;
  else {
    int      ix = ua - 1;
    mp_digit *ap = DIGITS(a) + ix, *bp = DIGITS(b) + ix;

    while(ix >= 0) {
      if(*ap > *bp)
	return MP_GT;
      else if(*ap < *bp)
	return MP_LT;

      --ap; --bp; --ix;
    }

    return MP_EQ;
  }

} /* end s_mp_cmp() */

/* }}} */

/* {{{ s_mp_cmp_d(a, d) */

/* Compare |a| <=> d, return 0 if equal, <0 if a<d, >0 if a>d             */
int      s_mp_cmp_d(mp_int *a, mp_digit d)
{
  mp_size  ua = USED(a);
  mp_digit *ap = DIGITS(a);

  if(ua > 1)
    return MP_GT;

  if(*ap < d) 
    return MP_LT;
  else if(*ap > d)
    return MP_GT;
  else
    return MP_EQ;

} /* end s_mp_cmp_d() */

/* }}} */

/* {{{ s_mp_ispow2(v) */

/*
  Returns -1 if the value is not a power of two; otherwise, it returns
  k such that v = 2^k, i.e. lg(v).
 */
int      s_mp_ispow2(mp_int *v)
{
  mp_digit d, *dp;
  mp_size  uv = USED(v);
  int      extra = 0, ix;

  d = DIGIT(v, uv - 1); /* most significant digit of v */

  while(d && ((d & 1) == 0)) {
    d >>= 1;
    ++extra;
  }

  if(d == 1) {
    ix = uv - 2;
    dp = DIGITS(v) + ix;

    while(ix >= 0) {
      if(*dp)
	return -1; /* not a power of two */

      --dp; --ix;
    }

    return ((uv - 1) * DIGIT_BIT) + extra;
  } 

  return -1;

} /* end s_mp_ispow2() */

/* }}} */

/* {{{ s_mp_ispow2d(d) */

int      s_mp_ispow2d(mp_digit d)
{
  int   pow = 0;

  while((d & 1) == 0) {
    ++pow; d >>= 1;
  }

  if(d == 1)
    return pow;

  return -1;

} /* end s_mp_ispow2d() */

/* }}} */

/* }}} */

/* {{{ Primitive I/O helpers */

/* {{{ s_mp_tovalue(ch, r) */

/*
  Convert the given character to its digit value, in the given radix.
  If the given character is not understood in the given radix, -1 is
  returned.  Otherwise the digit's numeric value is returned.

  The results will be odd if you use a radix < 2 or > 62, you are
  expected to know what you're up to.
 */
int      s_mp_tovalue(char ch, int r)
{
  int    val, xch;
  
  if(r > 36)
    xch = ch;
  else
    xch = toupper(ch);

  if(isdigit(xch))
    val = xch - '0';
  else if(isupper(xch))
    val = xch - 'A' + 10;
  else if(islower(xch))
    val = xch - 'a' + 36;
  else if(xch == '+')
    val = 62;
  else if(xch == '/')
    val = 63;
  else 
    return -1;

  if(val < 0 || val >= r)
    return -1;

  return val;

} /* end s_mp_tovalue() */

/* }}} */

/* {{{ s_mp_todigit(val, r, low) */

/*
  Convert val to a radix-r digit, if possible.  If val is out of range
  for r, returns zero.  Otherwise, returns an ASCII character denoting
  the value in the given radix.

  The results may be odd if you use a radix < 2 or > 64, you are
  expected to know what you're doing.
 */
  
char     s_mp_todigit(int val, int r, int low)
{
  char   ch;

  if(val < 0 || val >= r)
    return 0;

  ch = s_dmap_1[val];

  if(r <= 36 && low)
    ch = tolower(ch);

  return ch;

} /* end s_mp_todigit() */

/* }}} */

/* {{{ s_mp_outlen(bits, radix) */

/* 
   Return an estimate for how long a string is needed to hold a radix
   r representation of a number with 'bits' significant bits.

   Does not include space for a sign or a NUL terminator.
 */
int      s_mp_outlen(int bits, int r)
{
  return (int)((double)bits * LOG_V_2(r));

} /* end s_mp_outlen() */

/* }}} */

/* }}} */

/*------------------------------------------------------------------------*/
/* HERE THERE BE DRAGONS                                                  */
/* crc==4242132123, version==2, Sat Feb 02 06:43:52 2002 */

/*
    mpi.h

    by Michael J. Fromberger <[email protected]>
    Copyright (C) 1998 Michael J. Fromberger, All Rights Reserved

    Arbitrary precision integer arithmetic library
 */

#ifndef _H_MPI_
#define _H_MPI_

#include "mpi-config.h"

#define  MP_LT       -1
#define  MP_EQ        0
#define  MP_GT        1

#if MP_DEBUG
#undef MP_IOFUNC
#define MP_IOFUNC 1
#endif

#if MP_IOFUNC
#include <stdio.h>
#include <ctype.h>
#endif

#include <limits.h>

#define  MP_NEG  1
#define  MP_ZPOS 0

/* Included for compatibility... */
#define  NEG     MP_NEG
#define  ZPOS    MP_ZPOS

#define  MP_OKAY          0 /* no error, all is well */
#define  MP_YES           0 /* yes (boolean result)  */
#define  MP_NO           -1 /* no (boolean result)   */
#define  MP_MEM          -2 /* out of memory         */
#define  MP_RANGE        -3 /* argument out of range */
#define  MP_BADARG       -4 /* invalid parameter     */
#define  MP_UNDEF        -5 /* answer is undefined   */
#define  MP_LAST_CODE    MP_UNDEF

#include "mpi-types.h"

/* Included for compatibility... */
#define DIGIT_BIT         MP_DIGIT_BIT
#define DIGIT_MAX         MP_DIGIT_MAX

/* Macros for accessing the mp_int internals           */
#define  SIGN(MP)     ((MP)->sign)
#define  USED(MP)     ((MP)->used)
#define  ALLOC(MP)    ((MP)->alloc)
#define  DIGITS(MP)   ((MP)->dp)
#define  DIGIT(MP,N)  (MP)->dp[(N)]

#if MP_ARGCHK == 1
#define  ARGCHK(X,Y)  {if(!(X)){return (Y);}}
#elif MP_ARGCHK == 2
#include <assert.h>
#define  ARGCHK(X,Y)  assert(X)
#else
#define  ARGCHK(X,Y)  /*  */
#endif

/* This defines the maximum I/O base (minimum is 2)   */
#define MAX_RADIX         64

typedef struct {
  mp_sign       sign;    /* sign of this quantity      */
  mp_size       alloc;   /* how many digits allocated  */
  mp_size       used;    /* how many digits used       */
  mp_digit     *dp;      /* the digits themselves      */
} mp_int;

/*------------------------------------------------------------------------*/
/* Default precision                                                      */

unsigned int mp_get_prec(void);
void         mp_set_prec(unsigned int prec);

/*------------------------------------------------------------------------*/
/* Memory management                                                      */

mp_err mp_init(mp_int *mp);
mp_err mp_init_array(mp_int mp[], int count);
mp_err mp_init_size(mp_int *mp, mp_size prec);
mp_err mp_init_copy(mp_int *mp, mp_int *from);
mp_err mp_copy(mp_int *from, mp_int *to);
void   mp_exch(mp_int *mp1, mp_int *mp2);
void   mp_clear(mp_int *mp);
void   mp_clear_array(mp_int mp[], int count);
void   mp_zero(mp_int *mp);
void   mp_set(mp_int *mp, mp_digit d);
mp_err mp_set_int(mp_int *mp, long z);
mp_err mp_shrink(mp_int *a);


/*------------------------------------------------------------------------*/
/* Single digit arithmetic                                                */

mp_err mp_add_d(mp_int *a, mp_digit d, mp_int *b);
mp_err mp_sub_d(mp_int *a, mp_digit d, mp_int *b);
mp_err mp_mul_d(mp_int *a, mp_digit d, mp_int *b);
mp_err mp_mul_2(mp_int *a, mp_int *c);
mp_err mp_div_d(mp_int *a, mp_digit d, mp_int *q, mp_digit *r);
mp_err mp_div_2(mp_int *a, mp_int *c);
mp_err mp_expt_d(mp_int *a, mp_digit d, mp_int *c);

/*------------------------------------------------------------------------*/
/* Sign manipulations                                                     */

mp_err mp_abs(mp_int *a, mp_int *b);
mp_err mp_neg(mp_int *a, mp_int *b);

/*------------------------------------------------------------------------*/
/* Full arithmetic                                                        */

mp_err mp_add(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_sub(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_mul(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_mul_2d(mp_int *a, mp_digit d, mp_int *c);
#if MP_SQUARE
mp_err mp_sqr(mp_int *a, mp_int *b);
#else
#define mp_sqr(a, b) mp_mul(a, a, b)
#endif
mp_err mp_div(mp_int *a, mp_int *b, mp_int *q, mp_int *r);
mp_err mp_div_2d(mp_int *a, mp_digit d, mp_int *q, mp_int *r);
mp_err mp_expt(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_2expt(mp_int *a, mp_digit k);
mp_err mp_sqrt(mp_int *a, mp_int *b);

/*------------------------------------------------------------------------*/
/* Modular arithmetic                                                     */

#if MP_MODARITH
mp_err mp_mod(mp_int *a, mp_int *m, mp_int *c);
mp_err mp_mod_d(mp_int *a, mp_digit d, mp_digit *c);
mp_err mp_addmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c);
mp_err mp_submod(mp_int *a, mp_int *b, mp_int *m, mp_int *c);
mp_err mp_mulmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c);
#if MP_SQUARE
mp_err mp_sqrmod(mp_int *a, mp_int *m, mp_int *c);
#else
#define mp_sqrmod(a, m, c) mp_mulmod(a, a, m, c)
#endif
mp_err mp_exptmod(mp_int *a, mp_int *b, mp_int *m, mp_int *c);
mp_err mp_exptmod_d(mp_int *a, mp_digit d, mp_int *m, mp_int *c);
#endif /* MP_MODARITH */

/*------------------------------------------------------------------------*/
/* Comparisons                                                            */

int    mp_cmp_z(mp_int *a);
int    mp_cmp_d(mp_int *a, mp_digit d);
int    mp_cmp(mp_int *a, mp_int *b);
int    mp_cmp_mag(mp_int *a, mp_int *b);
int    mp_cmp_int(mp_int *a, long z);
int    mp_isodd(mp_int *a);
int    mp_iseven(mp_int *a);

/*------------------------------------------------------------------------*/
/* Number theoretic                                                       */

#if MP_NUMTH
mp_err mp_gcd(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_lcm(mp_int *a, mp_int *b, mp_int *c);
mp_err mp_xgcd(mp_int *a, mp_int *b, mp_int *g, mp_int *x, mp_int *y);
mp_err mp_invmod(mp_int *a, mp_int *m, mp_int *c);
#endif /* end MP_NUMTH */

/*------------------------------------------------------------------------*/
/* Input and output                                                       */

#if MP_IOFUNC
void   mp_print(mp_int *mp, FILE *ofp);
#endif /* end MP_IOFUNC */

/*------------------------------------------------------------------------*/
/* Base conversion                                                        */

#define BITS     1
#define BYTES    CHAR_BIT

mp_err mp_read_signed_bin(mp_int *mp, unsigned char *str, int len);
int    mp_signed_bin_size(mp_int *mp);
mp_err mp_to_signed_bin(mp_int *mp, unsigned char *str);

mp_err mp_read_unsigned_bin(mp_int *mp, unsigned char *str, int len);
int    mp_unsigned_bin_size(mp_int *mp);
mp_err mp_to_unsigned_bin(mp_int *mp, unsigned char *str);

int    mp_count_bits(mp_int *mp);

#if MP_COMPAT_MACROS
#define mp_read_raw(mp, str, len) mp_read_signed_bin((mp), (str), (len))
#define mp_raw_size(mp)           mp_signed_bin_size(mp)
#define mp_toraw(mp, str)         mp_to_signed_bin((mp), (str))
#define mp_read_mag(mp, str, len) mp_read_unsigned_bin((mp), (str), (len))
#define mp_mag_size(mp)           mp_unsigned_bin_size(mp)
#define mp_tomag(mp, str)         mp_to_unsigned_bin((mp), (str))
#endif

mp_err mp_read_radix(mp_int *mp, unsigned char *str, int radix);
int    mp_radix_size(mp_int *mp, int radix);
int    mp_value_radix_size(int num, int qty, int radix);
mp_err mp_toradix(mp_int *mp, unsigned char *str, int radix);

int    mp_char2value(char ch, int r);

#define mp_tobinary(M, S)  mp_toradix((M), (S), 2)
#define mp_tooctal(M, S)   mp_toradix((M), (S), 8)
#define mp_todecimal(M, S) mp_toradix((M), (S), 10)
#define mp_tohex(M, S)     mp_toradix((M), (S), 16)

/*------------------------------------------------------------------------*/
/* Error strings                                                          */

const  char  *mp_strerror(mp_err ec);

#endif /* end _H_MPI_ */

/* makes a bignum test harness with NUM tests per operation
 *
 * the output is made in the following format [one parameter per line]

operation
operand1
operand2
[... operandN]
result1
result2
[... resultN]

So for example "a * b mod n" would be

mulmod
a
b
n
a*b mod n

e.g. if a=3, b=4 n=11 then

mulmod
3
4
11
1

 */

#ifdef MP_8BIT
#define THE_MASK 127
#else
#define THE_MASK 32767
#endif

#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#include "mpi.c"

FILE *rng;

void rand_num(mp_int *a)
{
   int n, size;
   unsigned char buf[2048];

   size = 1 + ((fgetc(rng)<<8) + fgetc(rng)) % 101;
   buf[0] = (fgetc(rng)&1)?1:0;
   fread(buf+1, 1, size, rng);
   while (buf[1] == 0) buf[1] = fgetc(rng);
   mp_read_raw(a, buf, 1+size);
}

void rand_num2(mp_int *a)
{
   int n, size;
   unsigned char buf[2048];

   size = 10 + ((fgetc(rng)<<8) + fgetc(rng)) % 101;
   buf[0] = (fgetc(rng)&1)?1:0;
   fread(buf+1, 1, size, rng);
   while (buf[1] == 0) buf[1] = fgetc(rng);
   mp_read_raw(a, buf, 1+size);
}

#define mp_to64(a, b) mp_toradix(a, b, 64)

int main(void)
{
   int n, tmp;
   mp_int a, b, c, d, e;
   clock_t t1;
   char buf[4096];

   mp_init(&a);
   mp_init(&b);
   mp_init(&c);
   mp_init(&d);
   mp_init(&e);


   /* initial (2^n - 1)^2 testing, makes sure the comba multiplier works [it has the new carry code] */
/*
   mp_set(&a, 1);
   for (n = 1; n < 8192; n++) {
       mp_mul(&a, &a, &c);
       printf("mul\n");
       mp_to64(&a, buf);
       printf("%s\n%s\n", buf, buf);
       mp_to64(&c, buf);
       printf("%s\n", buf);

       mp_add_d(&a, 1, &a);
       mp_mul_2(&a, &a);
       mp_sub_d(&a, 1, &a);
   }
*/

   rng = fopen("/dev/urandom", "rb");
   if (rng == NULL) {
      rng = fopen("/dev/random", "rb");
      if (rng == NULL) {
         fprintf(stderr, "\nWarning:  stdin used as random source\n\n");
         rng = stdin;
      }
   }

   t1 = clock();
   for (;;) {
#if 0
      if (clock() - t1 > CLOCKS_PER_SEC) {
         sleep(2);
         t1 = clock();
      }
#endif
       n = fgetc(rng) % 15;

   if (n == 0) {
       /* add tests */
       rand_num(&a);
       rand_num(&b);
       mp_add(&a, &b, &c);
       printf("add\n");
       mp_to64(&a, buf);
       printf("%s\n", buf);
       mp_to64(&b, buf);
       printf("%s\n", buf);
       mp_to64(&c, buf);
       printf("%s\n", buf);
   } else if (n == 1) {
      /* sub tests */
       rand_num(&a);
       rand_num(&b);
       mp_sub(&a, &b, &c);
       printf("sub\n");
       mp_to64(&a, buf);
       printf("%s\n", buf);
       mp_to64(&b, buf);
       printf("%s\n", buf);
       mp_to64(&c, buf);
       printf("%s\n", buf);
   } else if (n == 2) {
       /* mul tests */
       rand_num(&a);
       rand_num(&b);
       mp_mul(&a, &b, &c);
       printf("mul\n");
       mp_to64(&a, buf);
       printf("%s\n", buf);
       mp_to64(&b, buf);
       printf("%s\n", buf);
       mp_to64(&c, buf);
       printf("%s\n", buf);
   } else if (n == 3) {
      /* div tests */
       rand_num(&a);
       rand_num(&b);
       mp_div(&a, &b, &c, &d);
       printf("div\n");
       mp_to64(&a, buf);
       printf("%s\n", buf);
       mp_to64(&b, buf);
       printf("%s\n", buf);
       mp_to64(&c, buf);
       printf("%s\n", buf);
       mp_to64(&d, buf);
       printf("%s\n", buf);
   } else if (n == 4) {
      /* sqr tests */
       rand_num(&a);
       mp_sqr(&a, &b);
       printf("sqr\n");
       mp_to64(&a, buf);
       printf("%s\n", buf);
       mp_to64(&b, buf);
       printf("%s\n", buf);
   } else if (n == 5) {
      /* mul_2d test */
      rand_num(&a);
      mp_copy(&a, &b);
      n = fgetc(rng) & 63;
      mp_mul_2d(&b, n, &b);
      mp_to64(&a, buf);
      printf("mul2d\n");
      printf("%s\n", buf);
      printf("%d\n", n);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   } else if (n == 6) {
      /* div_2d test */
      rand_num(&a);
      mp_copy(&a, &b);
      n = fgetc(rng) & 63;
      mp_div_2d(&b, n, &b, NULL);
      mp_to64(&a, buf);
      printf("div2d\n");
      printf("%s\n", buf);
      printf("%d\n", n);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   } else if (n == 7) {
      /* gcd test */
      rand_num(&a);
      rand_num(&b);
      a.sign = MP_ZPOS;
      b.sign = MP_ZPOS;
      mp_gcd(&a, &b, &c);
      printf("gcd\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
      mp_to64(&c, buf);
      printf("%s\n", buf);
   } else if (n == 8) {
      /* lcm test */
      rand_num(&a);
      rand_num(&b);
      a.sign = MP_ZPOS;
      b.sign = MP_ZPOS;
      mp_lcm(&a, &b, &c);
      printf("lcm\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
      mp_to64(&c, buf);
      printf("%s\n", buf);
   } else if (n == 9) {
      /* exptmod test */
      rand_num2(&a);
      rand_num2(&b);
      rand_num2(&c);
//      if (c.dp[0]&1) mp_add_d(&c, 1, &c);
      a.sign = b.sign = c.sign = 0;
      mp_exptmod(&a, &b, &c, &d);
      printf("expt\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
      mp_to64(&c, buf);
      printf("%s\n", buf);
      mp_to64(&d, buf);
      printf("%s\n", buf);
   } else if (n == 10) {
      /* invmod test */
      rand_num2(&a);
      rand_num2(&b);
      b.sign = MP_ZPOS;
      a.sign = MP_ZPOS;
      mp_gcd(&a, &b, &c);
      if (mp_cmp_d(&c, 1) != 0) continue;
      if (mp_cmp_d(&b, 1) == 0) continue;
      mp_invmod(&a, &b, &c);
      printf("invmod\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
      mp_to64(&c, buf);
      printf("%s\n", buf);
   } else if (n == 11) {
      rand_num(&a);
      mp_mul_2(&a, &a);
      mp_div_2(&a, &b);
      printf("div2\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   } else if (n == 12) {
      rand_num2(&a);
      mp_mul_2(&a, &b);
      printf("mul2\n");
      mp_to64(&a, buf);
      printf("%s\n", buf);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   } else if (n == 13) {
      rand_num2(&a);
      tmp = abs(rand()) & THE_MASK;
      mp_add_d(&a, tmp, &b);
      printf("add_d\n");
      mp_to64(&a, buf);
      printf("%s\n%d\n", buf, tmp);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   } else if (n == 14) {
      rand_num2(&a);
      tmp = abs(rand()) & THE_MASK;
      mp_sub_d(&a, tmp, &b);
      printf("sub_d\n");
      mp_to64(&a, buf);
      printf("%s\n%d\n", buf, tmp);
      mp_to64(&b, buf);
      printf("%s\n", buf);
   }
   }
   fclose(rng);
   return 0;
}

# makes the images... yeah

default:  pses

design_process.ps: design_process.tif
	tiff2ps -s -e design_process.tif > design_process.ps

sliding_window.ps: sliding_window.tif
	tiff2ps -s -e sliding_window.tif > sliding_window.ps
	
expt_state.ps: expt_state.tif
	tiff2ps -s -e expt_state.tif > expt_state.ps

primality.ps: primality.tif
	tiff2ps -s -e primality.tif > primality.ps

design_process.pdf: design_process.ps
	epstopdf design_process.ps

sliding_window.pdf: sliding_window.ps
	epstopdf sliding_window.ps
	
expt_state.pdf: expt_state.ps
	epstopdf expt_state.ps

primality.pdf: primality.ps
	epstopdf primality.ps


pses: sliding_window.ps expt_state.ps primality.ps design_process.ps
pdfes: sliding_window.pdf expt_state.pdf primality.pdf design_process.pdf

clean:
	rm -rf *.ps *.pdf .xvpics
   

/* Start: bn_error.c */
#include <tommath.h>
#ifdef BN_ERROR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

static const struct {
     int code;
     char *msg;
} msgs[] = {
     { MP_OKAY, "Successful" },
     { MP_MEM,  "Out of heap" },
     { MP_VAL,  "Value out of range" }
};

/* return a char * string for a given code */
char *mp_error_to_string(int code)
{
   int x;

   /* scan the lookup table for the given message */
   for (x = 0; x < (int)(sizeof(msgs) / sizeof(msgs[0])); x++) {
       if (msgs[x].code == code) {
          return msgs[x].msg;
       }
   }

   /* generic reply for invalid code */
   return "Invalid error code";
}

#endif

/* End: bn_error.c */

/* Start: bn_fast_mp_invmod.c */
#include <tommath.h>
#ifdef BN_FAST_MP_INVMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes the modular inverse via binary extended euclidean algorithm, 
 * that is c = 1/a mod b 
 *
 * Based on slow invmod except this is optimized for the case where b is 
 * odd as per HAC Note 14.64 on pp. 610
 */
int fast_mp_invmod (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int  x, y, u, v, B, D;
  int     res, neg;

  /* 2. [modified] b must be odd   */
  if (mp_iseven (b) == 1) {
    return MP_VAL;
  }

  /* init all our temps */
  if ((res = mp_init_multi(&x, &y, &u, &v, &B, &D, NULL)) != MP_OKAY) {
     return res;
  }

  /* x == modulus, y == value to invert */
  if ((res = mp_copy (b, &x)) != MP_OKAY) {
    goto LBL_ERR;
  }

  /* we need y = |a| */
  if ((res = mp_mod (a, b, &y)) != MP_OKAY) {
    goto LBL_ERR;
  }

  /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */
  if ((res = mp_copy (&x, &u)) != MP_OKAY) {
    goto LBL_ERR;
  }
  if ((res = mp_copy (&y, &v)) != MP_OKAY) {
    goto LBL_ERR;
  }
  mp_set (&D, 1);

top:
  /* 4.  while u is even do */
  while (mp_iseven (&u) == 1) {
    /* 4.1 u = u/2 */
    if ((res = mp_div_2 (&u, &u)) != MP_OKAY) {
      goto LBL_ERR;
    }
    /* 4.2 if B is odd then */
    if (mp_isodd (&B) == 1) {
      if ((res = mp_sub (&B, &x, &B)) != MP_OKAY) {
        goto LBL_ERR;
      }
    }
    /* B = B/2 */
    if ((res = mp_div_2 (&B, &B)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* 5.  while v is even do */
  while (mp_iseven (&v) == 1) {
    /* 5.1 v = v/2 */
    if ((res = mp_div_2 (&v, &v)) != MP_OKAY) {
      goto LBL_ERR;
    }
    /* 5.2 if D is odd then */
    if (mp_isodd (&D) == 1) {
      /* D = (D-x)/2 */
      if ((res = mp_sub (&D, &x, &D)) != MP_OKAY) {
        goto LBL_ERR;
      }
    }
    /* D = D/2 */
    if ((res = mp_div_2 (&D, &D)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* 6.  if u >= v then */
  if (mp_cmp (&u, &v) != MP_LT) {
    /* u = u - v, B = B - D */
    if ((res = mp_sub (&u, &v, &u)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&B, &D, &B)) != MP_OKAY) {
      goto LBL_ERR;
    }
  } else {
    /* v - v - u, D = D - B */
    if ((res = mp_sub (&v, &u, &v)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&D, &B, &D)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* if not zero goto step 4 */
  if (mp_iszero (&u) == 0) {
    goto top;
  }

  /* now a = C, b = D, gcd == g*v */

  /* if v != 1 then there is no inverse */
  if (mp_cmp_d (&v, 1) != MP_EQ) {
    res = MP_VAL;
    goto LBL_ERR;
  }

  /* b is now the inverse */
  neg = a->sign;
  while (D.sign == MP_NEG) {
    if ((res = mp_add (&D, b, &D)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }
  mp_exch (&D, c);
  c->sign = neg;
  res = MP_OKAY;

LBL_ERR:mp_clear_multi (&x, &y, &u, &v, &B, &D, NULL);
  return res;
}
#endif

/* End: bn_fast_mp_invmod.c */

/* Start: bn_fast_mp_montgomery_reduce.c */
#include <tommath.h>
#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes xR**-1 == x (mod N) via Montgomery Reduction
 *
 * This is an optimized implementation of montgomery_reduce
 * which uses the comba method to quickly calculate the columns of the
 * reduction.
 *
 * Based on Algorithm 14.32 on pp.601 of HAC.
*/
int fast_mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho)
{
  int     ix, res, olduse;
  mp_word W[MP_WARRAY];

  /* get old used count */
  olduse = x->used;

  /* grow a as required */
  if (x->alloc < n->used + 1) {
    if ((res = mp_grow (x, n->used + 1)) != MP_OKAY) {
      return res;
    }
  }

  /* first we have to get the digits of the input into
   * an array of double precision words W[...]
   */
  {
    register mp_word *_W;
    register mp_digit *tmpx;

    /* alias for the W[] array */
    _W   = W;

    /* alias for the digits of  x*/
    tmpx = x->dp;

    /* copy the digits of a into W[0..a->used-1] */
    for (ix = 0; ix < x->used; ix++) {
      *_W++ = *tmpx++;
    }

    /* zero the high words of W[a->used..m->used*2] */
    for (; ix < n->used * 2 + 1; ix++) {
      *_W++ = 0;
    }
  }

  /* now we proceed to zero successive digits
   * from the least significant upwards
   */
  for (ix = 0; ix < n->used; ix++) {
    /* mu = ai * m' mod b
     *
     * We avoid a double precision multiplication (which isn't required)
     * by casting the value down to a mp_digit.  Note this requires
     * that W[ix-1] have  the carry cleared (see after the inner loop)
     */
    register mp_digit mu;
    mu = (mp_digit) (((W[ix] & MP_MASK) * rho) & MP_MASK);

    /* a = a + mu * m * b**i
     *
     * This is computed in place and on the fly.  The multiplication
     * by b**i is handled by offseting which columns the results
     * are added to.
     *
     * Note the comba method normally doesn't handle carries in the
     * inner loop In this case we fix the carry from the previous
     * column since the Montgomery reduction requires digits of the
     * result (so far) [see above] to work.  This is
     * handled by fixing up one carry after the inner loop.  The
     * carry fixups are done in order so after these loops the
     * first m->used words of W[] have the carries fixed
     */
    {
      register int iy;
      register mp_digit *tmpn;
      register mp_word *_W;

      /* alias for the digits of the modulus */
      tmpn = n->dp;

      /* Alias for the columns set by an offset of ix */
      _W = W + ix;

      /* inner loop */
      for (iy = 0; iy < n->used; iy++) {
          *_W++ += ((mp_word)mu) * ((mp_word)*tmpn++);
      }
    }

    /* now fix carry for next digit, W[ix+1] */
    W[ix + 1] += W[ix] >> ((mp_word) DIGIT_BIT);
  }

  /* now we have to propagate the carries and
   * shift the words downward [all those least
   * significant digits we zeroed].
   */
  {
    register mp_digit *tmpx;
    register mp_word *_W, *_W1;

    /* nox fix rest of carries */

    /* alias for current word */
    _W1 = W + ix;

    /* alias for next word, where the carry goes */
    _W = W + ++ix;

    for (; ix <= n->used * 2 + 1; ix++) {
      *_W++ += *_W1++ >> ((mp_word) DIGIT_BIT);
    }

    /* copy out, A = A/b**n
     *
     * The result is A/b**n but instead of converting from an
     * array of mp_word to mp_digit than calling mp_rshd
     * we just copy them in the right order
     */

    /* alias for destination word */
    tmpx = x->dp;

    /* alias for shifted double precision result */
    _W = W + n->used;

    for (ix = 0; ix < n->used + 1; ix++) {
      *tmpx++ = (mp_digit)(*_W++ & ((mp_word) MP_MASK));
    }

    /* zero oldused digits, if the input a was larger than
     * m->used+1 we'll have to clear the digits
     */
    for (; ix < olduse; ix++) {
      *tmpx++ = 0;
    }
  }

  /* set the max used and clamp */
  x->used = n->used + 1;
  mp_clamp (x);

  /* if A >= m then A = A - m */
  if (mp_cmp_mag (x, n) != MP_LT) {
    return s_mp_sub (x, n, x);
  }
  return MP_OKAY;
}
#endif

/* End: bn_fast_mp_montgomery_reduce.c */

/* Start: bn_fast_s_mp_mul_digs.c */
#include <tommath.h>
#ifdef BN_FAST_S_MP_MUL_DIGS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Fast (comba) multiplier
 *
 * This is the fast column-array [comba] multiplier.  It is 
 * designed to compute the columns of the product first 
 * then handle the carries afterwards.  This has the effect 
 * of making the nested loops that compute the columns very
 * simple and schedulable on super-scalar processors.
 *
 * This has been modified to produce a variable number of 
 * digits of output so if say only a half-product is required 
 * you don't have to compute the upper half (a feature 
 * required for fast Barrett reduction).
 *
 * Based on Algorithm 14.12 on pp.595 of HAC.
 *
 */
int fast_s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs)
{
  int     olduse, res, pa, ix, iz;
  mp_digit W[MP_WARRAY];
  register mp_word  _W;

  /* grow the destination as required */
  if (c->alloc < digs) {
    if ((res = mp_grow (c, digs)) != MP_OKAY) {
      return res;
    }
  }

  /* number of output digits to produce */
  pa = MIN(digs, a->used + b->used);

  /* clear the carry */
  _W = 0;
  for (ix = 0; ix < pa; ix++) { 
      int      tx, ty;
      int      iy;
      mp_digit *tmpx, *tmpy;

      /* get offsets into the two bignums */
      ty = MIN(b->used-1, ix);
      tx = ix - ty;

      /* setup temp aliases */
      tmpx = a->dp + tx;
      tmpy = b->dp + ty;

      /* this is the number of times the loop will iterrate, essentially 
         while (tx++ < a->used && ty-- >= 0) { ... }
       */
      iy = MIN(a->used-tx, ty+1);

      /* execute loop */
      for (iz = 0; iz < iy; ++iz) {
         _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--);

      }

      /* store term */
      W[ix] = ((mp_digit)_W) & MP_MASK;

      /* make next carry */
      _W = _W >> ((mp_word)DIGIT_BIT);
 }

  /* setup dest */
  olduse  = c->used;
  c->used = pa;

  {
    register mp_digit *tmpc;
    tmpc = c->dp;
    for (ix = 0; ix < pa+1; ix++) {
      /* now extract the previous digit [below the carry] */
      *tmpc++ = W[ix];
    }

    /* clear unused digits [that existed in the old copy of c] */
    for (; ix < olduse; ix++) {
      *tmpc++ = 0;
    }
  }
  mp_clamp (c);
  return MP_OKAY;
}
#endif

/* End: bn_fast_s_mp_mul_digs.c */

/* Start: bn_fast_s_mp_mul_high_digs.c */
#include <tommath.h>
#ifdef BN_FAST_S_MP_MUL_HIGH_DIGS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* this is a modified version of fast_s_mul_digs that only produces
 * output digits *above* digs.  See the comments for fast_s_mul_digs
 * to see how it works.
 *
 * This is used in the Barrett reduction since for one of the multiplications
 * only the higher digits were needed.  This essentially halves the work.
 *
 * Based on Algorithm 14.12 on pp.595 of HAC.
 */
int fast_s_mp_mul_high_digs (mp_int * a, mp_int * b, mp_int * c, int digs)
{
  int     olduse, res, pa, ix, iz;
  mp_digit W[MP_WARRAY];
  mp_word  _W;

  /* grow the destination as required */
  pa = a->used + b->used;
  if (c->alloc < pa) {
    if ((res = mp_grow (c, pa)) != MP_OKAY) {
      return res;
    }
  }

  /* number of output digits to produce */
  pa = a->used + b->used;
  _W = 0;
  for (ix = digs; ix < pa; ix++) { 
      int      tx, ty, iy;
      mp_digit *tmpx, *tmpy;

      /* get offsets into the two bignums */
      ty = MIN(b->used-1, ix);
      tx = ix - ty;

      /* setup temp aliases */
      tmpx = a->dp + tx;
      tmpy = b->dp + ty;

      /* this is the number of times the loop will iterrate, essentially its 
         while (tx++ < a->used && ty-- >= 0) { ... }
       */
      iy = MIN(a->used-tx, ty+1);

      /* execute loop */
      for (iz = 0; iz < iy; iz++) {
         _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--);
      }

      /* store term */
      W[ix] = ((mp_digit)_W) & MP_MASK;

      /* make next carry */
      _W = _W >> ((mp_word)DIGIT_BIT);
  }
  
  /* setup dest */
  olduse  = c->used;
  c->used = pa;

  {
    register mp_digit *tmpc;

    tmpc = c->dp + digs;
    for (ix = digs; ix < pa; ix++) {
      /* now extract the previous digit [below the carry] */
      *tmpc++ = W[ix];
    }

    /* clear unused digits [that existed in the old copy of c] */
    for (; ix < olduse; ix++) {
      *tmpc++ = 0;
    }
  }
  mp_clamp (c);
  return MP_OKAY;
}
#endif

/* End: bn_fast_s_mp_mul_high_digs.c */

/* Start: bn_fast_s_mp_sqr.c */
#include <tommath.h>
#ifdef BN_FAST_S_MP_SQR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* the jist of squaring...
 * you do like mult except the offset of the tmpx [one that 
 * starts closer to zero] can't equal the offset of tmpy.  
 * So basically you set up iy like before then you min it with
 * (ty-tx) so that it never happens.  You double all those 
 * you add in the inner loop

After that loop you do the squares and add them in.
*/

int fast_s_mp_sqr (mp_int * a, mp_int * b)
{
  int       olduse, res, pa, ix, iz;
  mp_digit   W[MP_WARRAY], *tmpx;
  mp_word   W1;

  /* grow the destination as required */
  pa = a->used + a->used;
  if (b->alloc < pa) {
    if ((res = mp_grow (b, pa)) != MP_OKAY) {
      return res;
    }
  }

  /* number of output digits to produce */
  W1 = 0;
  for (ix = 0; ix < pa; ix++) { 
      int      tx, ty, iy;
      mp_word  _W;
      mp_digit *tmpy;

      /* clear counter */
      _W = 0;

      /* get offsets into the two bignums */
      ty = MIN(a->used-1, ix);
      tx = ix - ty;

      /* setup temp aliases */
      tmpx = a->dp + tx;
      tmpy = a->dp + ty;

      /* this is the number of times the loop will iterrate, essentially
         while (tx++ < a->used && ty-- >= 0) { ... }
       */
      iy = MIN(a->used-tx, ty+1);

      /* now for squaring tx can never equal ty 
       * we halve the distance since they approach at a rate of 2x
       * and we have to round because odd cases need to be executed
       */
      iy = MIN(iy, (ty-tx+1)>>1);

      /* execute loop */
      for (iz = 0; iz < iy; iz++) {
         _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--);
      }

      /* double the inner product and add carry */
      _W = _W + _W + W1;

      /* even columns have the square term in them */
      if ((ix&1) == 0) {
         _W += ((mp_word)a->dp[ix>>1])*((mp_word)a->dp[ix>>1]);
      }

      /* store it */
      W[ix] = (mp_digit)(_W & MP_MASK);

      /* make next carry */
      W1 = _W >> ((mp_word)DIGIT_BIT);
  }

  /* setup dest */
  olduse  = b->used;
  b->used = a->used+a->used;

  {
    mp_digit *tmpb;
    tmpb = b->dp;
    for (ix = 0; ix < pa; ix++) {
      *tmpb++ = W[ix] & MP_MASK;
    }

    /* clear unused digits [that existed in the old copy of c] */
    for (; ix < olduse; ix++) {
      *tmpb++ = 0;
    }
  }
  mp_clamp (b);
  return MP_OKAY;
}
#endif

/* End: bn_fast_s_mp_sqr.c */

/* Start: bn_mp_2expt.c */
#include <tommath.h>
#ifdef BN_MP_2EXPT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes a = 2**b 
 *
 * Simple algorithm which zeroes the int, grows it then just sets one bit
 * as required.
 */
int
mp_2expt (mp_int * a, int b)
{
  int     res;

  /* zero a as per default */
  mp_zero (a);

  /* grow a to accomodate the single bit */
  if ((res = mp_grow (a, b / DIGIT_BIT + 1)) != MP_OKAY) {
    return res;
  }

  /* set the used count of where the bit will go */
  a->used = b / DIGIT_BIT + 1;

  /* put the single bit in its place */
  a->dp[b / DIGIT_BIT] = ((mp_digit)1) << (b % DIGIT_BIT);

  return MP_OKAY;
}
#endif

/* End: bn_mp_2expt.c */

/* Start: bn_mp_abs.c */
#include <tommath.h>
#ifdef BN_MP_ABS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* b = |a| 
 *
 * Simple function copies the input and fixes the sign to positive
 */
int
mp_abs (mp_int * a, mp_int * b)
{
  int     res;

  /* copy a to b */
  if (a != b) {
     if ((res = mp_copy (a, b)) != MP_OKAY) {
       return res;
     }
  }

  /* force the sign of b to positive */
  b->sign = MP_ZPOS;

  return MP_OKAY;
}
#endif

/* End: bn_mp_abs.c */

/* Start: bn_mp_add.c */
#include <tommath.h>
#ifdef BN_MP_ADD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* high level addition (handles signs) */
int mp_add (mp_int * a, mp_int * b, mp_int * c)
{
  int     sa, sb, res;

  /* get sign of both inputs */
  sa = a->sign;
  sb = b->sign;

  /* handle two cases, not four */
  if (sa == sb) {
    /* both positive or both negative */
    /* add their magnitudes, copy the sign */
    c->sign = sa;
    res = s_mp_add (a, b, c);
  } else {
    /* one positive, the other negative */
    /* subtract the one with the greater magnitude from */
    /* the one of the lesser magnitude.  The result gets */
    /* the sign of the one with the greater magnitude. */
    if (mp_cmp_mag (a, b) == MP_LT) {
      c->sign = sb;
      res = s_mp_sub (b, a, c);
    } else {
      c->sign = sa;
      res = s_mp_sub (a, b, c);
    }
  }
  return res;
}

#endif

/* End: bn_mp_add.c */

/* Start: bn_mp_add_d.c */
#include <tommath.h>
#ifdef BN_MP_ADD_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* single digit addition */
int
mp_add_d (mp_int * a, mp_digit b, mp_int * c)
{
  int     res, ix, oldused;
  mp_digit *tmpa, *tmpc, mu;

  /* grow c as required */
  if (c->alloc < a->used + 1) {
     if ((res = mp_grow(c, a->used + 1)) != MP_OKAY) {
        return res;
     }
  }

  /* if a is negative and |a| >= b, call c = |a| - b */
  if (a->sign == MP_NEG && (a->used > 1 || a->dp[0] >= b)) {
     /* temporarily fix sign of a */
     a->sign = MP_ZPOS;

     /* c = |a| - b */
     res = mp_sub_d(a, b, c);

     /* fix sign  */
     a->sign = c->sign = MP_NEG;

     /* clamp */
     mp_clamp(c);

     return res;
  }

  /* old number of used digits in c */
  oldused = c->used;

  /* sign always positive */
  c->sign = MP_ZPOS;

  /* source alias */
  tmpa    = a->dp;

  /* destination alias */
  tmpc    = c->dp;

  /* if a is positive */
  if (a->sign == MP_ZPOS) {
     /* add digit, after this we're propagating
      * the carry.
      */
     *tmpc   = *tmpa++ + b;
     mu      = *tmpc >> DIGIT_BIT;
     *tmpc++ &= MP_MASK;

     /* now handle rest of the digits */
     for (ix = 1; ix < a->used; ix++) {
        *tmpc   = *tmpa++ + mu;
        mu      = *tmpc >> DIGIT_BIT;
        *tmpc++ &= MP_MASK;
     }
     /* set final carry */
     ix++;
     *tmpc++  = mu;

     /* setup size */
     c->used = a->used + 1;
  } else {
     /* a was negative and |a| < b */
     c->used  = 1;

     /* the result is a single digit */
     if (a->used == 1) {
        *tmpc++  =  b - a->dp[0];
     } else {
        *tmpc++  =  b;
     }

     /* setup count so the clearing of oldused
      * can fall through correctly
      */
     ix       = 1;
  }

  /* now zero to oldused */
  while (ix++ < oldused) {
     *tmpc++ = 0;
  }
  mp_clamp(c);

  return MP_OKAY;
}

#endif

/* End: bn_mp_add_d.c */

/* Start: bn_mp_addmod.c */
#include <tommath.h>
#ifdef BN_MP_ADDMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* d = a + b (mod c) */
int
mp_addmod (mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
  int     res;
  mp_int  t;

  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_add (a, b, &t)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }
  res = mp_mod (&t, c, d);
  mp_clear (&t);
  return res;
}
#endif

/* End: bn_mp_addmod.c */

/* Start: bn_mp_and.c */
#include <tommath.h>
#ifdef BN_MP_AND_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* AND two ints together */
int
mp_and (mp_int * a, mp_int * b, mp_int * c)
{
  int     res, ix, px;
  mp_int  t, *x;

  if (a->used > b->used) {
    if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
      return res;
    }
    px = b->used;
    x = b;
  } else {
    if ((res = mp_init_copy (&t, b)) != MP_OKAY) {
      return res;
    }
    px = a->used;
    x = a;
  }

  for (ix = 0; ix < px; ix++) {
    t.dp[ix] &= x->dp[ix];
  }

  /* zero digits above the last from the smallest mp_int */
  for (; ix < t.used; ix++) {
    t.dp[ix] = 0;
  }

  mp_clamp (&t);
  mp_exch (c, &t);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_mp_and.c */

/* Start: bn_mp_clamp.c */
#include <tommath.h>
#ifdef BN_MP_CLAMP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* trim unused digits 
 *
 * This is used to ensure that leading zero digits are
 * trimed and the leading "used" digit will be non-zero
 * Typically very fast.  Also fixes the sign if there
 * are no more leading digits
 */
void
mp_clamp (mp_int * a)
{
  /* decrease used while the most significant digit is
   * zero.
   */
  while (a->used > 0 && a->dp[a->used - 1] == 0) {
    --(a->used);
  }

  /* reset the sign flag if used == 0 */
  if (a->used == 0) {
    a->sign = MP_ZPOS;
  }
}
#endif

/* End: bn_mp_clamp.c */

/* Start: bn_mp_clear.c */
#include <tommath.h>
#ifdef BN_MP_CLEAR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* clear one (frees)  */
void
mp_clear (mp_int * a)
{
  int i;

  /* only do anything if a hasn't been freed previously */
  if (a->dp != NULL) {
    /* first zero the digits */
    for (i = 0; i < a->used; i++) {
        a->dp[i] = 0;
    }

    /* free ram */
    XFREE(a->dp);

    /* reset members to make debugging easier */
    a->dp    = NULL;
    a->alloc = a->used = 0;
    a->sign  = MP_ZPOS;
  }
}
#endif

/* End: bn_mp_clear.c */

/* Start: bn_mp_clear_multi.c */
#include <tommath.h>
#ifdef BN_MP_CLEAR_MULTI_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */
#include <stdarg.h>

void mp_clear_multi(mp_int *mp, ...) 
{
    mp_int* next_mp = mp;
    va_list args;
    va_start(args, mp);
    while (next_mp != NULL) {
        mp_clear(next_mp);
        next_mp = va_arg(args, mp_int*);
    }
    va_end(args);
}
#endif

/* End: bn_mp_clear_multi.c */

/* Start: bn_mp_cmp.c */
#include <tommath.h>
#ifdef BN_MP_CMP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* compare two ints (signed)*/
int
mp_cmp (mp_int * a, mp_int * b)
{
  /* compare based on sign */
  if (a->sign != b->sign) {
     if (a->sign == MP_NEG) {
        return MP_LT;
     } else {
        return MP_GT;
     }
  }
  
  /* compare digits */
  if (a->sign == MP_NEG) {
     /* if negative compare opposite direction */
     return mp_cmp_mag(b, a);
  } else {
     return mp_cmp_mag(a, b);
  }
}
#endif

/* End: bn_mp_cmp.c */

/* Start: bn_mp_cmp_d.c */
#include <tommath.h>
#ifdef BN_MP_CMP_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* compare a digit */
int mp_cmp_d(mp_int * a, mp_digit b)
{
  /* compare based on sign */
  if (a->sign == MP_NEG) {
    return MP_LT;
  }

  /* compare based on magnitude */
  if (a->used > 1) {
    return MP_GT;
  }

  /* compare the only digit of a to b */
  if (a->dp[0] > b) {
    return MP_GT;
  } else if (a->dp[0] < b) {
    return MP_LT;
  } else {
    return MP_EQ;
  }
}
#endif

/* End: bn_mp_cmp_d.c */

/* Start: bn_mp_cmp_mag.c */
#include <tommath.h>
#ifdef BN_MP_CMP_MAG_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* compare maginitude of two ints (unsigned) */
int mp_cmp_mag (mp_int * a, mp_int * b)
{
  int     n;
  mp_digit *tmpa, *tmpb;

  /* compare based on # of non-zero digits */
  if (a->used > b->used) {
    return MP_GT;
  }
  
  if (a->used < b->used) {
    return MP_LT;
  }

  /* alias for a */
  tmpa = a->dp + (a->used - 1);

  /* alias for b */
  tmpb = b->dp + (a->used - 1);

  /* compare based on digits  */
  for (n = 0; n < a->used; ++n, --tmpa, --tmpb) {
    if (*tmpa > *tmpb) {
      return MP_GT;
    }

    if (*tmpa < *tmpb) {
      return MP_LT;
    }
  }
  return MP_EQ;
}
#endif

/* End: bn_mp_cmp_mag.c */

/* Start: bn_mp_cnt_lsb.c */
#include <tommath.h>
#ifdef BN_MP_CNT_LSB_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

static const int lnz[16] = { 
   4, 0, 1, 0, 2, 0, 1, 0, 3, 0, 1, 0, 2, 0, 1, 0
};

/* Counts the number of lsbs which are zero before the first zero bit */
int mp_cnt_lsb(mp_int *a)
{
   int x;
   mp_digit q, qq;

   /* easy out */
   if (mp_iszero(a) == 1) {
      return 0;
   }

   /* scan lower digits until non-zero */
   for (x = 0; x < a->used && a->dp[x] == 0; x++);
   q = a->dp[x];
   x *= DIGIT_BIT;

   /* now scan this digit until a 1 is found */
   if ((q & 1) == 0) {
      do {
         qq  = q & 15;
         x  += lnz[qq];
         q >>= 4;
      } while (qq == 0);
   }
   return x;
}

#endif

/* End: bn_mp_cnt_lsb.c */

/* Start: bn_mp_copy.c */
#include <tommath.h>
#ifdef BN_MP_COPY_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* copy, b = a */
int
mp_copy (mp_int * a, mp_int * b)
{
  int     res, n;

  /* if dst == src do nothing */
  if (a == b) {
    return MP_OKAY;
  }

  /* grow dest */
  if (b->alloc < a->used) {
     if ((res = mp_grow (b, a->used)) != MP_OKAY) {
        return res;
     }
  }

  /* zero b and copy the parameters over */
  {
    register mp_digit *tmpa, *tmpb;

    /* pointer aliases */

    /* source */
    tmpa = a->dp;

    /* destination */
    tmpb = b->dp;

    /* copy all the digits */
    for (n = 0; n < a->used; n++) {
      *tmpb++ = *tmpa++;
    }

    /* clear high digits */
    for (; n < b->used; n++) {
      *tmpb++ = 0;
    }
  }

  /* copy used count and sign */
  b->used = a->used;
  b->sign = a->sign;
  return MP_OKAY;
}
#endif

/* End: bn_mp_copy.c */

/* Start: bn_mp_count_bits.c */
#include <tommath.h>
#ifdef BN_MP_COUNT_BITS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* returns the number of bits in an int */
int
mp_count_bits (mp_int * a)
{
  int     r;
  mp_digit q;

  /* shortcut */
  if (a->used == 0) {
    return 0;
  }

  /* get number of digits and add that */
  r = (a->used - 1) * DIGIT_BIT;
  
  /* take the last digit and count the bits in it */
  q = a->dp[a->used - 1];
  while (q > ((mp_digit) 0)) {
    ++r;
    q >>= ((mp_digit) 1);
  }
  return r;
}
#endif

/* End: bn_mp_count_bits.c */

/* Start: bn_mp_div.c */
#include <tommath.h>
#ifdef BN_MP_DIV_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

#ifdef BN_MP_DIV_SMALL

/* slower bit-bang division... also smaller */
int mp_div(mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
   mp_int ta, tb, tq, q;
   int    res, n, n2;

  /* is divisor zero ? */
  if (mp_iszero (b) == 1) {
    return MP_VAL;
  }

  /* if a < b then q=0, r = a */
  if (mp_cmp_mag (a, b) == MP_LT) {
    if (d != NULL) {
      res = mp_copy (a, d);
    } else {
      res = MP_OKAY;
    }
    if (c != NULL) {
      mp_zero (c);
    }
    return res;
  }
	
  /* init our temps */
  if ((res = mp_init_multi(&ta, &tb, &tq, &q, NULL) != MP_OKAY)) {
     return res;
  }


  mp_set(&tq, 1);
  n = mp_count_bits(a) - mp_count_bits(b);
  if (((res = mp_abs(a, &ta)) != MP_OKAY) ||
      ((res = mp_abs(b, &tb)) != MP_OKAY) || 
      ((res = mp_mul_2d(&tb, n, &tb)) != MP_OKAY) ||
      ((res = mp_mul_2d(&tq, n, &tq)) != MP_OKAY)) {
      goto LBL_ERR;
  }

  while (n-- >= 0) {
     if (mp_cmp(&tb, &ta) != MP_GT) {
        if (((res = mp_sub(&ta, &tb, &ta)) != MP_OKAY) ||
            ((res = mp_add(&q, &tq, &q)) != MP_OKAY)) {
           goto LBL_ERR;
        }
     }
     if (((res = mp_div_2d(&tb, 1, &tb, NULL)) != MP_OKAY) ||
         ((res = mp_div_2d(&tq, 1, &tq, NULL)) != MP_OKAY)) {
           goto LBL_ERR;
     }
  }

  /* now q == quotient and ta == remainder */
  n  = a->sign;
  n2 = (a->sign == b->sign ? MP_ZPOS : MP_NEG);
  if (c != NULL) {
     mp_exch(c, &q);
     c->sign  = (mp_iszero(c) == MP_YES) ? MP_ZPOS : n2;
  }
  if (d != NULL) {
     mp_exch(d, &ta);
     d->sign = (mp_iszero(d) == MP_YES) ? MP_ZPOS : n;
  }
LBL_ERR:
   mp_clear_multi(&ta, &tb, &tq, &q, NULL);
   return res;
}

#else

/* integer signed division. 
 * c*b + d == a [e.g. a/b, c=quotient, d=remainder]
 * HAC pp.598 Algorithm 14.20
 *
 * Note that the description in HAC is horribly 
 * incomplete.  For example, it doesn't consider 
 * the case where digits are removed from 'x' in 
 * the inner loop.  It also doesn't consider the 
 * case that y has fewer than three digits, etc..
 *
 * The overall algorithm is as described as 
 * 14.20 from HAC but fixed to treat these cases.
*/
int mp_div (mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
  mp_int  q, x, y, t1, t2;
  int     res, n, t, i, norm, neg;

  /* is divisor zero ? */
  if (mp_iszero (b) == 1) {
    return MP_VAL;
  }

  /* if a < b then q=0, r = a */
  if (mp_cmp_mag (a, b) == MP_LT) {
    if (d != NULL) {
      res = mp_copy (a, d);
    } else {
      res = MP_OKAY;
    }
    if (c != NULL) {
      mp_zero (c);
    }
    return res;
  }

  if ((res = mp_init_size (&q, a->used + 2)) != MP_OKAY) {
    return res;
  }
  q.used = a->used + 2;

  if ((res = mp_init (&t1)) != MP_OKAY) {
    goto LBL_Q;
  }

  if ((res = mp_init (&t2)) != MP_OKAY) {
    goto LBL_T1;
  }

  if ((res = mp_init_copy (&x, a)) != MP_OKAY) {
    goto LBL_T2;
  }

  if ((res = mp_init_copy (&y, b)) != MP_OKAY) {
    goto LBL_X;
  }

  /* fix the sign */
  neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG;
  x.sign = y.sign = MP_ZPOS;

  /* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */
  norm = mp_count_bits(&y) % DIGIT_BIT;
  if (norm < (int)(DIGIT_BIT-1)) {
     norm = (DIGIT_BIT-1) - norm;
     if ((res = mp_mul_2d (&x, norm, &x)) != MP_OKAY) {
       goto LBL_Y;
     }
     if ((res = mp_mul_2d (&y, norm, &y)) != MP_OKAY) {
       goto LBL_Y;
     }
  } else {
     norm = 0;
  }

  /* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */
  n = x.used - 1;
  t = y.used - 1;

  /* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */
  if ((res = mp_lshd (&y, n - t)) != MP_OKAY) { /* y = y*b**{n-t} */
    goto LBL_Y;
  }

  while (mp_cmp (&x, &y) != MP_LT) {
    ++(q.dp[n - t]);
    if ((res = mp_sub (&x, &y, &x)) != MP_OKAY) {
      goto LBL_Y;
    }
  }

  /* reset y by shifting it back down */
  mp_rshd (&y, n - t);

  /* step 3. for i from n down to (t + 1) */
  for (i = n; i >= (t + 1); i--) {
    if (i > x.used) {
      continue;
    }

    /* step 3.1 if xi == yt then set q{i-t-1} to b-1, 
     * otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */
    if (x.dp[i] == y.dp[t]) {
      q.dp[i - t - 1] = ((((mp_digit)1) << DIGIT_BIT) - 1);
    } else {
      mp_word tmp;
      tmp = ((mp_word) x.dp[i]) << ((mp_word) DIGIT_BIT);
      tmp |= ((mp_word) x.dp[i - 1]);
      tmp /= ((mp_word) y.dp[t]);
      if (tmp > (mp_word) MP_MASK)
        tmp = MP_MASK;
      q.dp[i - t - 1] = (mp_digit) (tmp & (mp_word) (MP_MASK));
    }

    /* while (q{i-t-1} * (yt * b + y{t-1})) > 
             xi * b**2 + xi-1 * b + xi-2 
     
       do q{i-t-1} -= 1; 
    */
    q.dp[i - t - 1] = (q.dp[i - t - 1] + 1) & MP_MASK;
    do {
      q.dp[i - t - 1] = (q.dp[i - t - 1] - 1) & MP_MASK;

      /* find left hand */
      mp_zero (&t1);
      t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1];
      t1.dp[1] = y.dp[t];
      t1.used = 2;
      if ((res = mp_mul_d (&t1, q.dp[i - t - 1], &t1)) != MP_OKAY) {
        goto LBL_Y;
      }

      /* find right hand */
      t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2];
      t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1];
      t2.dp[2] = x.dp[i];
      t2.used = 3;
    } while (mp_cmp_mag(&t1, &t2) == MP_GT);

    /* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */
    if ((res = mp_mul_d (&y, q.dp[i - t - 1], &t1)) != MP_OKAY) {
      goto LBL_Y;
    }

    if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) {
      goto LBL_Y;
    }

    if ((res = mp_sub (&x, &t1, &x)) != MP_OKAY) {
      goto LBL_Y;
    }

    /* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */
    if (x.sign == MP_NEG) {
      if ((res = mp_copy (&y, &t1)) != MP_OKAY) {
        goto LBL_Y;
      }
      if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) {
        goto LBL_Y;
      }
      if ((res = mp_add (&x, &t1, &x)) != MP_OKAY) {
        goto LBL_Y;
      }

      q.dp[i - t - 1] = (q.dp[i - t - 1] - 1UL) & MP_MASK;
    }
  }

  /* now q is the quotient and x is the remainder 
   * [which we have to normalize] 
   */
  
  /* get sign before writing to c */
  x.sign = x.used == 0 ? MP_ZPOS : a->sign;

  if (c != NULL) {
    mp_clamp (&q);
    mp_exch (&q, c);
    c->sign = neg;
  }

  if (d != NULL) {
    mp_div_2d (&x, norm, &x, NULL);
    mp_exch (&x, d);
  }

  res = MP_OKAY;

LBL_Y:mp_clear (&y);
LBL_X:mp_clear (&x);
LBL_T2:mp_clear (&t2);
LBL_T1:mp_clear (&t1);
LBL_Q:mp_clear (&q);
  return res;
}

#endif

#endif

/* End: bn_mp_div.c */

/* Start: bn_mp_div_2.c */
#include <tommath.h>
#ifdef BN_MP_DIV_2_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* b = a/2 */
int mp_div_2(mp_int * a, mp_int * b)
{
  int     x, res, oldused;

  /* copy */
  if (b->alloc < a->used) {
    if ((res = mp_grow (b, a->used)) != MP_OKAY) {
      return res;
    }
  }

  oldused = b->used;
  b->used = a->used;
  {
    register mp_digit r, rr, *tmpa, *tmpb;

    /* source alias */
    tmpa = a->dp + b->used - 1;

    /* dest alias */
    tmpb = b->dp + b->used - 1;

    /* carry */
    r = 0;
    for (x = b->used - 1; x >= 0; x--) {
      /* get the carry for the next iteration */
      rr = *tmpa & 1;

      /* shift the current digit, add in carry and store */
      *tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1));

      /* forward carry to next iteration */
      r = rr;
    }

    /* zero excess digits */
    tmpb = b->dp + b->used;
    for (x = b->used; x < oldused; x++) {
      *tmpb++ = 0;
    }
  }
  b->sign = a->sign;
  mp_clamp (b);
  return MP_OKAY;
}
#endif

/* End: bn_mp_div_2.c */

/* Start: bn_mp_div_2d.c */
#include <tommath.h>
#ifdef BN_MP_DIV_2D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* shift right by a certain bit count (store quotient in c, optional remainder in d) */
int mp_div_2d (mp_int * a, int b, mp_int * c, mp_int * d)
{
  mp_digit D, r, rr;
  int     x, res;
  mp_int  t;


  /* if the shift count is <= 0 then we do no work */
  if (b <= 0) {
    res = mp_copy (a, c);
    if (d != NULL) {
      mp_zero (d);
    }
    return res;
  }

  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  /* get the remainder */
  if (d != NULL) {
    if ((res = mp_mod_2d (a, b, &t)) != MP_OKAY) {
      mp_clear (&t);
      return res;
    }
  }

  /* copy */
  if ((res = mp_copy (a, c)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }

  /* shift by as many digits in the bit count */
  if (b >= (int)DIGIT_BIT) {
    mp_rshd (c, b / DIGIT_BIT);
  }

  /* shift any bit count < DIGIT_BIT */
  D = (mp_digit) (b % DIGIT_BIT);
  if (D != 0) {
    register mp_digit *tmpc, mask, shift;

    /* mask */
    mask = (((mp_digit)1) << D) - 1;

    /* shift for lsb */
    shift = DIGIT_BIT - D;

    /* alias */
    tmpc = c->dp + (c->used - 1);

    /* carry */
    r = 0;
    for (x = c->used - 1; x >= 0; x--) {
      /* get the lower  bits of this word in a temp */
      rr = *tmpc & mask;

      /* shift the current word and mix in the carry bits from the previous word */
      *tmpc = (*tmpc >> D) | (r << shift);
      --tmpc;

      /* set the carry to the carry bits of the current word found above */
      r = rr;
    }
  }
  mp_clamp (c);
  if (d != NULL) {
    mp_exch (&t, d);
  }
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_mp_div_2d.c */

/* Start: bn_mp_div_3.c */
#include <tommath.h>
#ifdef BN_MP_DIV_3_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* divide by three (based on routine from MPI and the GMP manual) */
int
mp_div_3 (mp_int * a, mp_int *c, mp_digit * d)
{
  mp_int   q;
  mp_word  w, t;
  mp_digit b;
  int      res, ix;
  
  /* b = 2**DIGIT_BIT / 3 */
  b = (((mp_word)1) << ((mp_word)DIGIT_BIT)) / ((mp_word)3);

  if ((res = mp_init_size(&q, a->used)) != MP_OKAY) {
     return res;
  }
  
  q.used = a->used;
  q.sign = a->sign;
  w = 0;
  for (ix = a->used - 1; ix >= 0; ix--) {
     w = (w << ((mp_word)DIGIT_BIT)) | ((mp_word)a->dp[ix]);

     if (w >= 3) {
        /* multiply w by [1/3] */
        t = (w * ((mp_word)b)) >> ((mp_word)DIGIT_BIT);

        /* now subtract 3 * [w/3] from w, to get the remainder */
        w -= t+t+t;

        /* fixup the remainder as required since
         * the optimization is not exact.
         */
        while (w >= 3) {
           t += 1;
           w -= 3;
        }
      } else {
        t = 0;
      }
      q.dp[ix] = (mp_digit)t;
  }

  /* [optional] store the remainder */
  if (d != NULL) {
     *d = (mp_digit)w;
  }

  /* [optional] store the quotient */
  if (c != NULL) {
     mp_clamp(&q);
     mp_exch(&q, c);
  }
  mp_clear(&q);
  
  return res;
}

#endif

/* End: bn_mp_div_3.c */

/* Start: bn_mp_div_d.c */
#include <tommath.h>
#ifdef BN_MP_DIV_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

static int s_is_power_of_two(mp_digit b, int *p)
{
   int x;

   /* fast return if no power of two */
   if ((b==0) || (b & (b-1))) {
      return 0;
   }

   for (x = 0; x < DIGIT_BIT; x++) {
      if (b == (((mp_digit)1)<<x)) {
         *p = x;
         return 1;
      }
   }
   return 0;
}

/* single digit division (based on routine from MPI) */
int mp_div_d (mp_int * a, mp_digit b, mp_int * c, mp_digit * d)
{
  mp_int  q;
  mp_word w;
  mp_digit t;
  int     res, ix;

  /* cannot divide by zero */
  if (b == 0) {
     return MP_VAL;
  }

  /* quick outs */
  if (b == 1 || mp_iszero(a) == 1) {
     if (d != NULL) {
        *d = 0;
     }
     if (c != NULL) {
        return mp_copy(a, c);
     }
     return MP_OKAY;
  }

  /* power of two ? */
  if (s_is_power_of_two(b, &ix) == 1) {
     if (d != NULL) {
        *d = a->dp[0] & ((((mp_digit)1)<<ix) - 1);
     }
     if (c != NULL) {
        return mp_div_2d(a, ix, c, NULL);
     }
     return MP_OKAY;
  }

#ifdef BN_MP_DIV_3_C
  /* three? */
  if (b == 3) {
     return mp_div_3(a, c, d);
  }
#endif

  /* no easy answer [c'est la vie].  Just division */
  if ((res = mp_init_size(&q, a->used)) != MP_OKAY) {
     return res;
  }
  
  q.used = a->used;
  q.sign = a->sign;
  w = 0;
  for (ix = a->used - 1; ix >= 0; ix--) {
     w = (w << ((mp_word)DIGIT_BIT)) | ((mp_word)a->dp[ix]);
     
     if (w >= b) {
        t = (mp_digit)(w / b);
        w -= ((mp_word)t) * ((mp_word)b);
      } else {
        t = 0;
      }
      q.dp[ix] = (mp_digit)t;
  }
  
  if (d != NULL) {
     *d = (mp_digit)w;
  }
  
  if (c != NULL) {
     mp_clamp(&q);
     mp_exch(&q, c);
  }
  mp_clear(&q);
  
  return res;
}

#endif

/* End: bn_mp_div_d.c */

/* Start: bn_mp_dr_is_modulus.c */
#include <tommath.h>
#ifdef BN_MP_DR_IS_MODULUS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines if a number is a valid DR modulus */
int mp_dr_is_modulus(mp_int *a)
{
   int ix;

   /* must be at least two digits */
   if (a->used < 2) {
      return 0;
   }

   /* must be of the form b**k - a [a <= b] so all
    * but the first digit must be equal to -1 (mod b).
    */
   for (ix = 1; ix < a->used; ix++) {
       if (a->dp[ix] != MP_MASK) {
          return 0;
       }
   }
   return 1;
}

#endif

/* End: bn_mp_dr_is_modulus.c */

/* Start: bn_mp_dr_reduce.c */
#include <tommath.h>
#ifdef BN_MP_DR_REDUCE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reduce "x" in place modulo "n" using the Diminished Radix algorithm.
 *
 * Based on algorithm from the paper
 *
 * "Generating Efficient Primes for Discrete Log Cryptosystems"
 *                 Chae Hoon Lim, Pil Joong Lee,
 *          POSTECH Information Research Laboratories
 *
 * The modulus must be of a special format [see manual]
 *
 * Has been modified to use algorithm 7.10 from the LTM book instead
 *
 * Input x must be in the range 0 <= x <= (n-1)**2
 */
int
mp_dr_reduce (mp_int * x, mp_int * n, mp_digit k)
{
  int      err, i, m;
  mp_word  r;
  mp_digit mu, *tmpx1, *tmpx2;

  /* m = digits in modulus */
  m = n->used;

  /* ensure that "x" has at least 2m digits */
  if (x->alloc < m + m) {
    if ((err = mp_grow (x, m + m)) != MP_OKAY) {
      return err;
    }
  }

/* top of loop, this is where the code resumes if
 * another reduction pass is required.
 */
top:
  /* aliases for digits */
  /* alias for lower half of x */
  tmpx1 = x->dp;

  /* alias for upper half of x, or x/B**m */
  tmpx2 = x->dp + m;

  /* set carry to zero */
  mu = 0;

  /* compute (x mod B**m) + k * [x/B**m] inline and inplace */
  for (i = 0; i < m; i++) {
      r         = ((mp_word)*tmpx2++) * ((mp_word)k) + *tmpx1 + mu;
      *tmpx1++  = (mp_digit)(r & MP_MASK);
      mu        = (mp_digit)(r >> ((mp_word)DIGIT_BIT));
  }

  /* set final carry */
  *tmpx1++ = mu;

  /* zero words above m */
  for (i = m + 1; i < x->used; i++) {
      *tmpx1++ = 0;
  }

  /* clamp, sub and return */
  mp_clamp (x);

  /* if x >= n then subtract and reduce again
   * Each successive "recursion" makes the input smaller and smaller.
   */
  if (mp_cmp_mag (x, n) != MP_LT) {
    s_mp_sub(x, n, x);
    goto top;
  }
  return MP_OKAY;
}
#endif

/* End: bn_mp_dr_reduce.c */

/* Start: bn_mp_dr_setup.c */
#include <tommath.h>
#ifdef BN_MP_DR_SETUP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines the setup value */
void mp_dr_setup(mp_int *a, mp_digit *d)
{
   /* the casts are required if DIGIT_BIT is one less than
    * the number of bits in a mp_digit [e.g. DIGIT_BIT==31]
    */
   *d = (mp_digit)((((mp_word)1) << ((mp_word)DIGIT_BIT)) - 
        ((mp_word)a->dp[0]));
}

#endif

/* End: bn_mp_dr_setup.c */

/* Start: bn_mp_exch.c */
#include <tommath.h>
#ifdef BN_MP_EXCH_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* swap the elements of two integers, for cases where you can't simply swap the 
 * mp_int pointers around
 */
void
mp_exch (mp_int * a, mp_int * b)
{
  mp_int  t;

  t  = *a;
  *a = *b;
  *b = t;
}
#endif

/* End: bn_mp_exch.c */

/* Start: bn_mp_expt_d.c */
#include <tommath.h>
#ifdef BN_MP_EXPT_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* calculate c = a**b  using a square-multiply algorithm */
int mp_expt_d (mp_int * a, mp_digit b, mp_int * c)
{
  int     res, x;
  mp_int  g;

  if ((res = mp_init_copy (&g, a)) != MP_OKAY) {
    return res;
  }

  /* set initial result */
  mp_set (c, 1);

  for (x = 0; x < (int) DIGIT_BIT; x++) {
    /* square */
    if ((res = mp_sqr (c, c)) != MP_OKAY) {
      mp_clear (&g);
      return res;
    }

    /* if the bit is set multiply */
    if ((b & (mp_digit) (((mp_digit)1) << (DIGIT_BIT - 1))) != 0) {
      if ((res = mp_mul (c, &g, c)) != MP_OKAY) {
         mp_clear (&g);
         return res;
      }
    }

    /* shift to next bit */
    b <<= 1;
  }

  mp_clear (&g);
  return MP_OKAY;
}
#endif

/* End: bn_mp_expt_d.c */

/* Start: bn_mp_exptmod.c */
#include <tommath.h>
#ifdef BN_MP_EXPTMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */


/* this is a shell function that calls either the normal or Montgomery
 * exptmod functions.  Originally the call to the montgomery code was
 * embedded in the normal function but that wasted alot of stack space
 * for nothing (since 99% of the time the Montgomery code would be called)
 */
int mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y)
{
  int dr;

  /* modulus P must be positive */
  if (P->sign == MP_NEG) {
     return MP_VAL;
  }

  /* if exponent X is negative we have to recurse */
  if (X->sign == MP_NEG) {
#ifdef BN_MP_INVMOD_C
     mp_int tmpG, tmpX;
     int err;

     /* first compute 1/G mod P */
     if ((err = mp_init(&tmpG)) != MP_OKAY) {
        return err;
     }
     if ((err = mp_invmod(G, P, &tmpG)) != MP_OKAY) {
        mp_clear(&tmpG);
        return err;
     }

     /* now get |X| */
     if ((err = mp_init(&tmpX)) != MP_OKAY) {
        mp_clear(&tmpG);
        return err;
     }
     if ((err = mp_abs(X, &tmpX)) != MP_OKAY) {
        mp_clear_multi(&tmpG, &tmpX, NULL);
        return err;
     }

     /* and now compute (1/G)**|X| instead of G**X [X < 0] */
     err = mp_exptmod(&tmpG, &tmpX, P, Y);
     mp_clear_multi(&tmpG, &tmpX, NULL);
     return err;
#else 
     /* no invmod */
     return MP_VAL;
#endif
  }

/* modified diminished radix reduction */
#if defined(BN_MP_REDUCE_IS_2K_L_C) && defined(BN_MP_REDUCE_2K_L_C) && defined(BN_S_MP_EXPTMOD_C)
  if (mp_reduce_is_2k_l(P) == MP_YES) {
     return s_mp_exptmod(G, X, P, Y, 1);
  }
#endif

#ifdef BN_MP_DR_IS_MODULUS_C
  /* is it a DR modulus? */
  dr = mp_dr_is_modulus(P);
#else
  /* default to no */
  dr = 0;
#endif

#ifdef BN_MP_REDUCE_IS_2K_C
  /* if not, is it a unrestricted DR modulus? */
  if (dr == 0) {
     dr = mp_reduce_is_2k(P) << 1;
  }
#endif
    
  /* if the modulus is odd or dr != 0 use the montgomery method */
#ifdef BN_MP_EXPTMOD_FAST_C
  if (mp_isodd (P) == 1 || dr !=  0) {
    return mp_exptmod_fast (G, X, P, Y, dr);
  } else {
#endif
#ifdef BN_S_MP_EXPTMOD_C
    /* otherwise use the generic Barrett reduction technique */
    return s_mp_exptmod (G, X, P, Y, 0);
#else
    /* no exptmod for evens */
    return MP_VAL;
#endif
#ifdef BN_MP_EXPTMOD_FAST_C
  }
#endif
}

#endif

/* End: bn_mp_exptmod.c */

/* Start: bn_mp_exptmod_fast.c */
#include <tommath.h>
#ifdef BN_MP_EXPTMOD_FAST_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes Y == G**X mod P, HAC pp.616, Algorithm 14.85
 *
 * Uses a left-to-right k-ary sliding window to compute the modular exponentiation.
 * The value of k changes based on the size of the exponent.
 *
 * Uses Montgomery or Diminished Radix reduction [whichever appropriate]
 */

#ifdef MP_LOW_MEM
   #define TAB_SIZE 32
#else
   #define TAB_SIZE 256
#endif

int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode)
{
  mp_int  M[TAB_SIZE], res;
  mp_digit buf, mp;
  int     err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize;

  /* use a pointer to the reduction algorithm.  This allows us to use
   * one of many reduction algorithms without modding the guts of
   * the code with if statements everywhere.
   */
  int     (*redux)(mp_int*,mp_int*,mp_digit);

  /* find window size */
  x = mp_count_bits (X);
  if (x <= 7) {
    winsize = 2;
  } else if (x <= 36) {
    winsize = 3;
  } else if (x <= 140) {
    winsize = 4;
  } else if (x <= 450) {
    winsize = 5;
  } else if (x <= 1303) {
    winsize = 6;
  } else if (x <= 3529) {
    winsize = 7;
  } else {
    winsize = 8;
  }

#ifdef MP_LOW_MEM
  if (winsize > 5) {
     winsize = 5;
  }
#endif

  /* init M array */
  /* init first cell */
  if ((err = mp_init(&M[1])) != MP_OKAY) {
     return err;
  }

  /* now init the second half of the array */
  for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
    if ((err = mp_init(&M[x])) != MP_OKAY) {
      for (y = 1<<(winsize-1); y < x; y++) {
        mp_clear (&M[y]);
      }
      mp_clear(&M[1]);
      return err;
    }
  }

  /* determine and setup reduction code */
  if (redmode == 0) {
#ifdef BN_MP_MONTGOMERY_SETUP_C     
     /* now setup montgomery  */
     if ((err = mp_montgomery_setup (P, &mp)) != MP_OKAY) {
        goto LBL_M;
     }
#else
     err = MP_VAL;
     goto LBL_M;
#endif

     /* automatically pick the comba one if available (saves quite a few calls/ifs) */
#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C
     if (((P->used * 2 + 1) < MP_WARRAY) &&
          P->used < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
        redux = fast_mp_montgomery_reduce;
     } else 
#endif
     {
#ifdef BN_MP_MONTGOMERY_REDUCE_C
        /* use slower baseline Montgomery method */
        redux = mp_montgomery_reduce;
#else
        err = MP_VAL;
        goto LBL_M;
#endif
     }
  } else if (redmode == 1) {
#if defined(BN_MP_DR_SETUP_C) && defined(BN_MP_DR_REDUCE_C)
     /* setup DR reduction for moduli of the form B**k - b */
     mp_dr_setup(P, &mp);
     redux = mp_dr_reduce;
#else
     err = MP_VAL;
     goto LBL_M;
#endif
  } else {
#if defined(BN_MP_REDUCE_2K_SETUP_C) && defined(BN_MP_REDUCE_2K_C)
     /* setup DR reduction for moduli of the form 2**k - b */
     if ((err = mp_reduce_2k_setup(P, &mp)) != MP_OKAY) {
        goto LBL_M;
     }
     redux = mp_reduce_2k;
#else
     err = MP_VAL;
     goto LBL_M;
#endif
  }

  /* setup result */
  if ((err = mp_init (&res)) != MP_OKAY) {
    goto LBL_M;
  }

  /* create M table
   *

   *
   * The first half of the table is not computed though accept for M[0] and M[1]
   */

  if (redmode == 0) {
#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C
     /* now we need R mod m */
     if ((err = mp_montgomery_calc_normalization (&res, P)) != MP_OKAY) {
       goto LBL_RES;
     }
#else 
     err = MP_VAL;
     goto LBL_RES;
#endif

     /* now set M[1] to G * R mod m */
     if ((err = mp_mulmod (G, &res, P, &M[1])) != MP_OKAY) {
       goto LBL_RES;
     }
  } else {
     mp_set(&res, 1);
     if ((err = mp_mod(G, P, &M[1])) != MP_OKAY) {
        goto LBL_RES;
     }
  }

  /* compute the value at M[1<<(winsize-1)] by squaring M[1] (winsize-1) times */
  if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) {
    goto LBL_RES;
  }

  for (x = 0; x < (winsize - 1); x++) {
    if ((err = mp_sqr (&M[1 << (winsize - 1)], &M[1 << (winsize - 1)])) != MP_OKAY) {
      goto LBL_RES;
    }
    if ((err = redux (&M[1 << (winsize - 1)], P, mp)) != MP_OKAY) {
      goto LBL_RES;
    }
  }

  /* create upper table */
  for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) {
    if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) {
      goto LBL_RES;
    }
    if ((err = redux (&M[x], P, mp)) != MP_OKAY) {
      goto LBL_RES;
    }
  }

  /* set initial mode and bit cnt */
  mode   = 0;
  bitcnt = 1;
  buf    = 0;
  digidx = X->used - 1;
  bitcpy = 0;
  bitbuf = 0;

  for (;;) {
    /* grab next digit as required */
    if (--bitcnt == 0) {
      /* if digidx == -1 we are out of digits so break */
      if (digidx == -1) {
        break;
      }
      /* read next digit and reset bitcnt */
      buf    = X->dp[digidx--];
      bitcnt = (int)DIGIT_BIT;
    }

    /* grab the next msb from the exponent */
    y     = (mp_digit)(buf >> (DIGIT_BIT - 1)) & 1;
    buf <<= (mp_digit)1;

    /* if the bit is zero and mode == 0 then we ignore it
     * These represent the leading zero bits before the first 1 bit
     * in the exponent.  Technically this opt is not required but it
     * does lower the # of trivial squaring/reductions used
     */
    if (mode == 0 && y == 0) {
      continue;
    }

    /* if the bit is zero and mode == 1 then we square */
    if (mode == 1 && y == 0) {
      if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, mp)) != MP_OKAY) {
        goto LBL_RES;
      }
      continue;
    }

    /* else we add it to the window */
    bitbuf |= (y << (winsize - ++bitcpy));
    mode    = 2;

    if (bitcpy == winsize) {
      /* ok window is filled so square as required and multiply  */
      /* square first */
      for (x = 0; x < winsize; x++) {
        if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
          goto LBL_RES;
        }
        if ((err = redux (&res, P, mp)) != MP_OKAY) {
          goto LBL_RES;
        }
      }

      /* then multiply */
      if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, mp)) != MP_OKAY) {
        goto LBL_RES;
      }

      /* empty window and reset */
      bitcpy = 0;
      bitbuf = 0;
      mode   = 1;
    }
  }

  /* if bits remain then square/multiply */
  if (mode == 2 && bitcpy > 0) {
    /* square then multiply if the bit is set */
    for (x = 0; x < bitcpy; x++) {
      if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, mp)) != MP_OKAY) {
        goto LBL_RES;
      }

      /* get next bit of the window */
      bitbuf <<= 1;
      if ((bitbuf & (1 << winsize)) != 0) {
        /* then multiply */
        if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) {
          goto LBL_RES;
        }
        if ((err = redux (&res, P, mp)) != MP_OKAY) {
          goto LBL_RES;
        }
      }
    }
  }

  if (redmode == 0) {
     /* fixup result if Montgomery reduction is used
      * recall that any value in a Montgomery system is
      * actually multiplied by R mod n.  So we have
      * to reduce one more time to cancel out the factor
      * of R.
      */
     if ((err = redux(&res, P, mp)) != MP_OKAY) {
       goto LBL_RES;
     }
  }

  /* swap res with Y */
  mp_exch (&res, Y);
  err = MP_OKAY;
LBL_RES:mp_clear (&res);
LBL_M:
  mp_clear(&M[1]);
  for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
    mp_clear (&M[x]);
  }
  return err;
}
#endif

/* End: bn_mp_exptmod_fast.c */

/* Start: bn_mp_exteuclid.c */
#include <tommath.h>
#ifdef BN_MP_EXTEUCLID_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Extended euclidean algorithm of (a, b) produces 
   a*u1 + b*u2 = u3
 */
int mp_exteuclid(mp_int *a, mp_int *b, mp_int *U1, mp_int *U2, mp_int *U3)
{
   mp_int u1,u2,u3,v1,v2,v3,t1,t2,t3,q,tmp;
   int err;

   if ((err = mp_init_multi(&u1, &u2, &u3, &v1, &v2, &v3, &t1, &t2, &t3, &q, &tmp, NULL)) != MP_OKAY) {
      return err;
   }

   /* initialize, (u1,u2,u3) = (1,0,a) */
   mp_set(&u1, 1);
   if ((err = mp_copy(a, &u3)) != MP_OKAY)                                        { goto _ERR; }

   /* initialize, (v1,v2,v3) = (0,1,b) */
   mp_set(&v2, 1);
   if ((err = mp_copy(b, &v3)) != MP_OKAY)                                        { goto _ERR; }

   /* loop while v3 != 0 */
   while (mp_iszero(&v3) == MP_NO) {
       /* q = u3/v3 */
       if ((err = mp_div(&u3, &v3, &q, NULL)) != MP_OKAY)                         { goto _ERR; }

       /* (t1,t2,t3) = (u1,u2,u3) - (v1,v2,v3)q */
       if ((err = mp_mul(&v1, &q, &tmp)) != MP_OKAY)                              { goto _ERR; }
       if ((err = mp_sub(&u1, &tmp, &t1)) != MP_OKAY)                             { goto _ERR; }
       if ((err = mp_mul(&v2, &q, &tmp)) != MP_OKAY)                              { goto _ERR; }
       if ((err = mp_sub(&u2, &tmp, &t2)) != MP_OKAY)                             { goto _ERR; }
       if ((err = mp_mul(&v3, &q, &tmp)) != MP_OKAY)                              { goto _ERR; }
       if ((err = mp_sub(&u3, &tmp, &t3)) != MP_OKAY)                             { goto _ERR; }

       /* (u1,u2,u3) = (v1,v2,v3) */
       if ((err = mp_copy(&v1, &u1)) != MP_OKAY)                                  { goto _ERR; }
       if ((err = mp_copy(&v2, &u2)) != MP_OKAY)                                  { goto _ERR; }
       if ((err = mp_copy(&v3, &u3)) != MP_OKAY)                                  { goto _ERR; }

       /* (v1,v2,v3) = (t1,t2,t3) */
       if ((err = mp_copy(&t1, &v1)) != MP_OKAY)                                  { goto _ERR; }
       if ((err = mp_copy(&t2, &v2)) != MP_OKAY)                                  { goto _ERR; }
       if ((err = mp_copy(&t3, &v3)) != MP_OKAY)                                  { goto _ERR; }
   }

   /* make sure U3 >= 0 */
   if (u3.sign == MP_NEG) {
      mp_neg(&u1, &u1);
      mp_neg(&u2, &u2);
      mp_neg(&u3, &u3);
   }

   /* copy result out */
   if (U1 != NULL) { mp_exch(U1, &u1); }
   if (U2 != NULL) { mp_exch(U2, &u2); }
   if (U3 != NULL) { mp_exch(U3, &u3); }

   err = MP_OKAY;
_ERR: mp_clear_multi(&u1, &u2, &u3, &v1, &v2, &v3, &t1, &t2, &t3, &q, &tmp, NULL);
   return err;
}
#endif

/* End: bn_mp_exteuclid.c */

/* Start: bn_mp_fread.c */
#include <tommath.h>
#ifdef BN_MP_FREAD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* read a bigint from a file stream in ASCII */
int mp_fread(mp_int *a, int radix, FILE *stream)
{
   int err, ch, neg, y;
   
   /* clear a */
   mp_zero(a);
   
   /* if first digit is - then set negative */
   ch = fgetc(stream);
   if (ch == '-') {
      neg = MP_NEG;
      ch = fgetc(stream);
   } else {
      neg = MP_ZPOS;
   }
   
   for (;;) {
      /* find y in the radix map */
      for (y = 0; y < radix; y++) {
          if (mp_s_rmap[y] == ch) {
             break;
          }
      }
      if (y == radix) {
         break;
      }
      
      /* shift up and add */
      if ((err = mp_mul_d(a, radix, a)) != MP_OKAY) {
         return err;
      }
      if ((err = mp_add_d(a, y, a)) != MP_OKAY) {
         return err;
      }
      
      ch = fgetc(stream);
   }
   if (mp_cmp_d(a, 0) != MP_EQ) {
      a->sign = neg;
   }
   
   return MP_OKAY;
}

#endif

/* End: bn_mp_fread.c */

/* Start: bn_mp_fwrite.c */
#include <tommath.h>
#ifdef BN_MP_FWRITE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

int mp_fwrite(mp_int *a, int radix, FILE *stream)
{
   char *buf;
   int err, len, x;
   
   if ((err = mp_radix_size(a, radix, &len)) != MP_OKAY) {
      return err;
   }

   buf = OPT_CAST(char) XMALLOC (len);
   if (buf == NULL) {
      return MP_MEM;
   }
   
   if ((err = mp_toradix(a, buf, radix)) != MP_OKAY) {
      XFREE (buf);
      return err;
   }
   
   for (x = 0; x < len; x++) {
       if (fputc(buf[x], stream) == EOF) {
          XFREE (buf);
          return MP_VAL;
       }
   }
   
   XFREE (buf);
   return MP_OKAY;
}

#endif

/* End: bn_mp_fwrite.c */

/* Start: bn_mp_gcd.c */
#include <tommath.h>
#ifdef BN_MP_GCD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Greatest Common Divisor using the binary method */
int mp_gcd (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int  u, v;
  int     k, u_lsb, v_lsb, res;

  /* either zero than gcd is the largest */
  if (mp_iszero (a) == MP_YES) {
    return mp_abs (b, c);
  }
  if (mp_iszero (b) == MP_YES) {
    return mp_abs (a, c);
  }

  /* get copies of a and b we can modify */
  if ((res = mp_init_copy (&u, a)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init_copy (&v, b)) != MP_OKAY) {
    goto LBL_U;
  }

  /* must be positive for the remainder of the algorithm */
  u.sign = v.sign = MP_ZPOS;

  /* B1.  Find the common power of two for u and v */
  u_lsb = mp_cnt_lsb(&u);
  v_lsb = mp_cnt_lsb(&v);
  k     = MIN(u_lsb, v_lsb);

  if (k > 0) {
     /* divide the power of two out */
     if ((res = mp_div_2d(&u, k, &u, NULL)) != MP_OKAY) {
        goto LBL_V;
     }

     if ((res = mp_div_2d(&v, k, &v, NULL)) != MP_OKAY) {
        goto LBL_V;
     }
  }

  /* divide any remaining factors of two out */
  if (u_lsb != k) {
     if ((res = mp_div_2d(&u, u_lsb - k, &u, NULL)) != MP_OKAY) {
        goto LBL_V;
     }
  }

  if (v_lsb != k) {
     if ((res = mp_div_2d(&v, v_lsb - k, &v, NULL)) != MP_OKAY) {
        goto LBL_V;
     }
  }

  while (mp_iszero(&v) == 0) {
     /* make sure v is the largest */
     if (mp_cmp_mag(&u, &v) == MP_GT) {
        /* swap u and v to make sure v is >= u */
        mp_exch(&u, &v);
     }
     
     /* subtract smallest from largest */
     if ((res = s_mp_sub(&v, &u, &v)) != MP_OKAY) {
        goto LBL_V;
     }
     
     /* Divide out all factors of two */
     if ((res = mp_div_2d(&v, mp_cnt_lsb(&v), &v, NULL)) != MP_OKAY) {
        goto LBL_V;
     } 
  } 

  /* multiply by 2**k which we divided out at the beginning */
  if ((res = mp_mul_2d (&u, k, c)) != MP_OKAY) {
     goto LBL_V;
  }
  c->sign = MP_ZPOS;
  res = MP_OKAY;
LBL_V:mp_clear (&u);
LBL_U:mp_clear (&v);
  return res;
}
#endif

/* End: bn_mp_gcd.c */

/* Start: bn_mp_get_int.c */
#include <tommath.h>
#ifdef BN_MP_GET_INT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* get the lower 32-bits of an mp_int */
unsigned long mp_get_int(mp_int * a) 
{
  int i;
  unsigned long res;

  if (a->used == 0) {
     return 0;
  }

  /* get number of digits of the lsb we have to read */
  i = MIN(a->used,(int)((sizeof(unsigned long)*CHAR_BIT+DIGIT_BIT-1)/DIGIT_BIT))-1;

  /* get most significant digit of result */
  res = DIGIT(a,i);
   
  while (--i >= 0) {
    res = (res << DIGIT_BIT) | DIGIT(a,i);
  }

  /* force result to 32-bits always so it is consistent on non 32-bit platforms */
  return res & 0xFFFFFFFFUL;
}
#endif

/* End: bn_mp_get_int.c */

/* Start: bn_mp_grow.c */
#include <tommath.h>
#ifdef BN_MP_GROW_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* grow as required */
int mp_grow (mp_int * a, int size)
{
  int     i;
  mp_digit *tmp;

  /* if the alloc size is smaller alloc more ram */
  if (a->alloc < size) {
    /* ensure there are always at least MP_PREC digits extra on top */
    size += (MP_PREC * 2) - (size % MP_PREC);

    /* reallocate the array a->dp
     *
     * We store the return in a temporary variable
     * in case the operation failed we don't want
     * to overwrite the dp member of a.
     */
    tmp = OPT_CAST(mp_digit) XREALLOC (a->dp, sizeof (mp_digit) * size);
    if (tmp == NULL) {
      /* reallocation failed but "a" is still valid [can be freed] */
      return MP_MEM;
    }

    /* reallocation succeeded so set a->dp */
    a->dp = tmp;

    /* zero excess digits */
    i        = a->alloc;
    a->alloc = size;
    for (; i < a->alloc; i++) {
      a->dp[i] = 0;
    }
  }
  return MP_OKAY;
}
#endif

/* End: bn_mp_grow.c */

/* Start: bn_mp_init.c */
#include <tommath.h>
#ifdef BN_MP_INIT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* init a new mp_int */
int mp_init (mp_int * a)
{
  int i;

  /* allocate memory required and clear it */
  a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * MP_PREC);
  if (a->dp == NULL) {
    return MP_MEM;
  }

  /* set the digits to zero */
  for (i = 0; i < MP_PREC; i++) {
      a->dp[i] = 0;
  }

  /* set the used to zero, allocated digits to the default precision
   * and sign to positive */
  a->used  = 0;
  a->alloc = MP_PREC;
  a->sign  = MP_ZPOS;

  return MP_OKAY;
}
#endif

/* End: bn_mp_init.c */

/* Start: bn_mp_init_copy.c */
#include <tommath.h>
#ifdef BN_MP_INIT_COPY_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* creates "a" then copies b into it */
int mp_init_copy (mp_int * a, mp_int * b)
{
  int     res;

  if ((res = mp_init (a)) != MP_OKAY) {
    return res;
  }
  return mp_copy (b, a);
}
#endif

/* End: bn_mp_init_copy.c */

/* Start: bn_mp_init_multi.c */
#include <tommath.h>
#ifdef BN_MP_INIT_MULTI_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */
#include <stdarg.h>

int mp_init_multi(mp_int *mp, ...) 
{
    mp_err res = MP_OKAY;      /* Assume ok until proven otherwise */
    int n = 0;                 /* Number of ok inits */
    mp_int* cur_arg = mp;
    va_list args;

    va_start(args, mp);        /* init args to next argument from caller */
    while (cur_arg != NULL) {
        if (mp_init(cur_arg) != MP_OKAY) {
            /* Oops - error! Back-track and mp_clear what we already
               succeeded in init-ing, then return error.
            */
            va_list clean_args;
            
            /* end the current list */
            va_end(args);
            
            /* now start cleaning up */            
            cur_arg = mp;
            va_start(clean_args, mp);
            while (n--) {
                mp_clear(cur_arg);
                cur_arg = va_arg(clean_args, mp_int*);
            }
            va_end(clean_args);
            res = MP_MEM;
            break;
        }
        n++;
        cur_arg = va_arg(args, mp_int*);
    }
    va_end(args);
    return res;                /* Assumed ok, if error flagged above. */
}

#endif

/* End: bn_mp_init_multi.c */

/* Start: bn_mp_init_set.c */
#include <tommath.h>
#ifdef BN_MP_INIT_SET_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* initialize and set a digit */
int mp_init_set (mp_int * a, mp_digit b)
{
  int err;
  if ((err = mp_init(a)) != MP_OKAY) {
     return err;
  }
  mp_set(a, b);
  return err;
}
#endif

/* End: bn_mp_init_set.c */

/* Start: bn_mp_init_set_int.c */
#include <tommath.h>
#ifdef BN_MP_INIT_SET_INT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* initialize and set a digit */
int mp_init_set_int (mp_int * a, unsigned long b)
{
  int err;
  if ((err = mp_init(a)) != MP_OKAY) {
     return err;
  }
  return mp_set_int(a, b);
}
#endif

/* End: bn_mp_init_set_int.c */

/* Start: bn_mp_init_size.c */
#include <tommath.h>
#ifdef BN_MP_INIT_SIZE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* init an mp_init for a given size */
int mp_init_size (mp_int * a, int size)
{
  int x;

  /* pad size so there are always extra digits */
  size += (MP_PREC * 2) - (size % MP_PREC);	
  
  /* alloc mem */
  a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * size);
  if (a->dp == NULL) {
    return MP_MEM;
  }

  /* set the members */
  a->used  = 0;
  a->alloc = size;
  a->sign  = MP_ZPOS;

  /* zero the digits */
  for (x = 0; x < size; x++) {
      a->dp[x] = 0;
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_init_size.c */

/* Start: bn_mp_invmod.c */
#include <tommath.h>
#ifdef BN_MP_INVMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* hac 14.61, pp608 */
int mp_invmod (mp_int * a, mp_int * b, mp_int * c)
{
  /* b cannot be negative */
  if (b->sign == MP_NEG || mp_iszero(b) == 1) {
    return MP_VAL;
  }

#ifdef BN_FAST_MP_INVMOD_C
  /* if the modulus is odd we can use a faster routine instead */
  if (mp_isodd (b) == 1) {
    return fast_mp_invmod (a, b, c);
  }
#endif

#ifdef BN_MP_INVMOD_SLOW_C
  return mp_invmod_slow(a, b, c);
#endif

  return MP_VAL;
}
#endif

/* End: bn_mp_invmod.c */

/* Start: bn_mp_invmod_slow.c */
#include <tommath.h>
#ifdef BN_MP_INVMOD_SLOW_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* hac 14.61, pp608 */
int mp_invmod_slow (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int  x, y, u, v, A, B, C, D;
  int     res;

  /* b cannot be negative */
  if (b->sign == MP_NEG || mp_iszero(b) == 1) {
    return MP_VAL;
  }

  /* init temps */
  if ((res = mp_init_multi(&x, &y, &u, &v, 
                           &A, &B, &C, &D, NULL)) != MP_OKAY) {
     return res;
  }

  /* x = a, y = b */
  if ((res = mp_mod(a, b, &x)) != MP_OKAY) {
      goto LBL_ERR;
  }
  if ((res = mp_copy (b, &y)) != MP_OKAY) {
    goto LBL_ERR;
  }

  /* 2. [modified] if x,y are both even then return an error! */
  if (mp_iseven (&x) == 1 && mp_iseven (&y) == 1) {
    res = MP_VAL;
    goto LBL_ERR;
  }

  /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */
  if ((res = mp_copy (&x, &u)) != MP_OKAY) {
    goto LBL_ERR;
  }
  if ((res = mp_copy (&y, &v)) != MP_OKAY) {
    goto LBL_ERR;
  }
  mp_set (&A, 1);
  mp_set (&D, 1);

top:
  /* 4.  while u is even do */
  while (mp_iseven (&u) == 1) {
    /* 4.1 u = u/2 */
    if ((res = mp_div_2 (&u, &u)) != MP_OKAY) {
      goto LBL_ERR;
    }
    /* 4.2 if A or B is odd then */
    if (mp_isodd (&A) == 1 || mp_isodd (&B) == 1) {
      /* A = (A+y)/2, B = (B-x)/2 */
      if ((res = mp_add (&A, &y, &A)) != MP_OKAY) {
         goto LBL_ERR;
      }
      if ((res = mp_sub (&B, &x, &B)) != MP_OKAY) {
         goto LBL_ERR;
      }
    }
    /* A = A/2, B = B/2 */
    if ((res = mp_div_2 (&A, &A)) != MP_OKAY) {
      goto LBL_ERR;
    }
    if ((res = mp_div_2 (&B, &B)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* 5.  while v is even do */
  while (mp_iseven (&v) == 1) {
    /* 5.1 v = v/2 */
    if ((res = mp_div_2 (&v, &v)) != MP_OKAY) {
      goto LBL_ERR;
    }
    /* 5.2 if C or D is odd then */
    if (mp_isodd (&C) == 1 || mp_isodd (&D) == 1) {
      /* C = (C+y)/2, D = (D-x)/2 */
      if ((res = mp_add (&C, &y, &C)) != MP_OKAY) {
         goto LBL_ERR;
      }
      if ((res = mp_sub (&D, &x, &D)) != MP_OKAY) {
         goto LBL_ERR;
      }
    }
    /* C = C/2, D = D/2 */
    if ((res = mp_div_2 (&C, &C)) != MP_OKAY) {
      goto LBL_ERR;
    }
    if ((res = mp_div_2 (&D, &D)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* 6.  if u >= v then */
  if (mp_cmp (&u, &v) != MP_LT) {
    /* u = u - v, A = A - C, B = B - D */
    if ((res = mp_sub (&u, &v, &u)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&A, &C, &A)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&B, &D, &B)) != MP_OKAY) {
      goto LBL_ERR;
    }
  } else {
    /* v - v - u, C = C - A, D = D - B */
    if ((res = mp_sub (&v, &u, &v)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&C, &A, &C)) != MP_OKAY) {
      goto LBL_ERR;
    }

    if ((res = mp_sub (&D, &B, &D)) != MP_OKAY) {
      goto LBL_ERR;
    }
  }

  /* if not zero goto step 4 */
  if (mp_iszero (&u) == 0)
    goto top;

  /* now a = C, b = D, gcd == g*v */

  /* if v != 1 then there is no inverse */
  if (mp_cmp_d (&v, 1) != MP_EQ) {
    res = MP_VAL;
    goto LBL_ERR;
  }

  /* if its too low */
  while (mp_cmp_d(&C, 0) == MP_LT) {
      if ((res = mp_add(&C, b, &C)) != MP_OKAY) {
         goto LBL_ERR;
      }
  }
  
  /* too big */
  while (mp_cmp_mag(&C, b) != MP_LT) {
      if ((res = mp_sub(&C, b, &C)) != MP_OKAY) {
         goto LBL_ERR;
      }
  }
  
  /* C is now the inverse */
  mp_exch (&C, c);
  res = MP_OKAY;
LBL_ERR:mp_clear_multi (&x, &y, &u, &v, &A, &B, &C, &D, NULL);
  return res;
}
#endif

/* End: bn_mp_invmod_slow.c */

/* Start: bn_mp_is_square.c */
#include <tommath.h>
#ifdef BN_MP_IS_SQUARE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Check if remainders are possible squares - fast exclude non-squares */
static const char rem_128[128] = {
 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1
};

static const char rem_105[105] = {
 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1,
 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1,
 0, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1,
 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1,
 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 1, 1,
 1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1
};

/* Store non-zero to ret if arg is square, and zero if not */
int mp_is_square(mp_int *arg,int *ret) 
{
  int           res;
  mp_digit      c;
  mp_int        t;
  unsigned long r;

  /* Default to Non-square :) */
  *ret = MP_NO; 

  if (arg->sign == MP_NEG) {
    return MP_VAL;
  }

  /* digits used?  (TSD) */
  if (arg->used == 0) {
     return MP_OKAY;
  }

  /* First check mod 128 (suppose that DIGIT_BIT is at least 7) */
  if (rem_128[127 & DIGIT(arg,0)] == 1) {
     return MP_OKAY;
  }

  /* Next check mod 105 (3*5*7) */
  if ((res = mp_mod_d(arg,105,&c)) != MP_OKAY) {
     return res;
  }
  if (rem_105[c] == 1) {
     return MP_OKAY;
  }


  if ((res = mp_init_set_int(&t,11L*13L*17L*19L*23L*29L*31L)) != MP_OKAY) {
     return res;
  }
  if ((res = mp_mod(arg,&t,&t)) != MP_OKAY) {
     goto ERR;
  }
  r = mp_get_int(&t);
  /* Check for other prime modules, note it's not an ERROR but we must
   * free "t" so the easiest way is to goto ERR.  We know that res
   * is already equal to MP_OKAY from the mp_mod call 
   */ 
  if ( (1L<<(r%11)) & 0x5C4L )             goto ERR;
  if ( (1L<<(r%13)) & 0x9E4L )             goto ERR;
  if ( (1L<<(r%17)) & 0x5CE8L )            goto ERR;
  if ( (1L<<(r%19)) & 0x4F50CL )           goto ERR;
  if ( (1L<<(r%23)) & 0x7ACCA0L )          goto ERR;
  if ( (1L<<(r%29)) & 0xC2EDD0CL )         goto ERR;
  if ( (1L<<(r%31)) & 0x6DE2B848L )        goto ERR;

  /* Final check - is sqr(sqrt(arg)) == arg ? */
  if ((res = mp_sqrt(arg,&t)) != MP_OKAY) {
     goto ERR;
  }
  if ((res = mp_sqr(&t,&t)) != MP_OKAY) {
     goto ERR;
  }

  *ret = (mp_cmp_mag(&t,arg) == MP_EQ) ? MP_YES : MP_NO;
ERR:mp_clear(&t);
  return res;
}
#endif

/* End: bn_mp_is_square.c */

/* Start: bn_mp_jacobi.c */
#include <tommath.h>
#ifdef BN_MP_JACOBI_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes the jacobi c = (a | n) (or Legendre if n is prime)
 * HAC pp. 73 Algorithm 2.149
 */
int mp_jacobi (mp_int * a, mp_int * p, int *c)
{
  mp_int  a1, p1;
  int     k, s, r, res;
  mp_digit residue;

  /* if p <= 0 return MP_VAL */
  if (mp_cmp_d(p, 0) != MP_GT) {
     return MP_VAL;
  }

  /* step 1.  if a == 0, return 0 */
  if (mp_iszero (a) == 1) {
    *c = 0;
    return MP_OKAY;
  }

  /* step 2.  if a == 1, return 1 */
  if (mp_cmp_d (a, 1) == MP_EQ) {
    *c = 1;
    return MP_OKAY;
  }

  /* default */
  s = 0;

  /* step 3.  write a = a1 * 2**k  */
  if ((res = mp_init_copy (&a1, a)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init (&p1)) != MP_OKAY) {
    goto LBL_A1;
  }

  /* divide out larger power of two */
  k = mp_cnt_lsb(&a1);
  if ((res = mp_div_2d(&a1, k, &a1, NULL)) != MP_OKAY) {
     goto LBL_P1;
  }

  /* step 4.  if e is even set s=1 */
  if ((k & 1) == 0) {
    s = 1;
  } else {
    /* else set s=1 if p = 1/7 (mod 8) or s=-1 if p = 3/5 (mod 8) */
    residue = p->dp[0] & 7;

    if (residue == 1 || residue == 7) {
      s = 1;
    } else if (residue == 3 || residue == 5) {
      s = -1;
    }
  }

  /* step 5.  if p == 3 (mod 4) *and* a1 == 3 (mod 4) then s = -s */
  if ( ((p->dp[0] & 3) == 3) && ((a1.dp[0] & 3) == 3)) {
    s = -s;
  }

  /* if a1 == 1 we're done */
  if (mp_cmp_d (&a1, 1) == MP_EQ) {
    *c = s;
  } else {
    /* n1 = n mod a1 */
    if ((res = mp_mod (p, &a1, &p1)) != MP_OKAY) {
      goto LBL_P1;
    }
    if ((res = mp_jacobi (&p1, &a1, &r)) != MP_OKAY) {
      goto LBL_P1;
    }
    *c = s * r;
  }

  /* done */
  res = MP_OKAY;
LBL_P1:mp_clear (&p1);
LBL_A1:mp_clear (&a1);
  return res;
}
#endif

/* End: bn_mp_jacobi.c */

/* Start: bn_mp_karatsuba_mul.c */
#include <tommath.h>
#ifdef BN_MP_KARATSUBA_MUL_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* c = |a| * |b| using Karatsuba Multiplication using 
 * three half size multiplications
 *
 * Let B represent the radix [e.g. 2**DIGIT_BIT] and 
 * let n represent half of the number of digits in 
 * the min(a,b)
 *
 * a = a1 * B**n + a0
 * b = b1 * B**n + b0
 *
 * Then, a * b => 
   a1b1 * B**2n + ((a1 + a0)(b1 + b0) - (a0b0 + a1b1)) * B + a0b0
 *
 * Note that a1b1 and a0b0 are used twice and only need to be 
 * computed once.  So in total three half size (half # of 
 * digit) multiplications are performed, a0b0, a1b1 and 
 * (a1+b1)(a0+b0)
 *
 * Note that a multiplication of half the digits requires
 * 1/4th the number of single precision multiplications so in 
 * total after one call 25% of the single precision multiplications 
 * are saved.  Note also that the call to mp_mul can end up back 
 * in this function if the a0, a1, b0, or b1 are above the threshold.  
 * This is known as divide-and-conquer and leads to the famous 
 * O(N**lg(3)) or O(N**1.584) work which is asymptopically lower than 
 * the standard O(N**2) that the baseline/comba methods use.  
 * Generally though the overhead of this method doesn't pay off 
 * until a certain size (N ~ 80) is reached.
 */
int mp_karatsuba_mul (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int  x0, x1, y0, y1, t1, x0y0, x1y1;
  int     B, err;

  /* default the return code to an error */
  err = MP_MEM;

  /* min # of digits */
  B = MIN (a->used, b->used);

  /* now divide in two */
  B = B >> 1;

  /* init copy all the temps */
  if (mp_init_size (&x0, B) != MP_OKAY)
    goto ERR;
  if (mp_init_size (&x1, a->used - B) != MP_OKAY)
    goto X0;
  if (mp_init_size (&y0, B) != MP_OKAY)
    goto X1;
  if (mp_init_size (&y1, b->used - B) != MP_OKAY)
    goto Y0;

  /* init temps */
  if (mp_init_size (&t1, B * 2) != MP_OKAY)
    goto Y1;
  if (mp_init_size (&x0y0, B * 2) != MP_OKAY)
    goto T1;
  if (mp_init_size (&x1y1, B * 2) != MP_OKAY)
    goto X0Y0;

  /* now shift the digits */
  x0.used = y0.used = B;
  x1.used = a->used - B;
  y1.used = b->used - B;

  {
    register int x;
    register mp_digit *tmpa, *tmpb, *tmpx, *tmpy;

    /* we copy the digits directly instead of using higher level functions
     * since we also need to shift the digits
     */
    tmpa = a->dp;
    tmpb = b->dp;

    tmpx = x0.dp;
    tmpy = y0.dp;
    for (x = 0; x < B; x++) {
      *tmpx++ = *tmpa++;
      *tmpy++ = *tmpb++;
    }

    tmpx = x1.dp;
    for (x = B; x < a->used; x++) {
      *tmpx++ = *tmpa++;
    }

    tmpy = y1.dp;
    for (x = B; x < b->used; x++) {
      *tmpy++ = *tmpb++;
    }
  }

  /* only need to clamp the lower words since by definition the 
   * upper words x1/y1 must have a known number of digits
   */
  mp_clamp (&x0);
  mp_clamp (&y0);

  /* now calc the products x0y0 and x1y1 */
  /* after this x0 is no longer required, free temp [x0==t2]! */
  if (mp_mul (&x0, &y0, &x0y0) != MP_OKAY)  
    goto X1Y1;          /* x0y0 = x0*y0 */
  if (mp_mul (&x1, &y1, &x1y1) != MP_OKAY)
    goto X1Y1;          /* x1y1 = x1*y1 */

  /* now calc x1+x0 and y1+y0 */
  if (s_mp_add (&x1, &x0, &t1) != MP_OKAY)
    goto X1Y1;          /* t1 = x1 - x0 */
  if (s_mp_add (&y1, &y0, &x0) != MP_OKAY)
    goto X1Y1;          /* t2 = y1 - y0 */
  if (mp_mul (&t1, &x0, &t1) != MP_OKAY)
    goto X1Y1;          /* t1 = (x1 + x0) * (y1 + y0) */

  /* add x0y0 */
  if (mp_add (&x0y0, &x1y1, &x0) != MP_OKAY)
    goto X1Y1;          /* t2 = x0y0 + x1y1 */
  if (s_mp_sub (&t1, &x0, &t1) != MP_OKAY)
    goto X1Y1;          /* t1 = (x1+x0)*(y1+y0) - (x1y1 + x0y0) */

  /* shift by B */
  if (mp_lshd (&t1, B) != MP_OKAY)
    goto X1Y1;          /* t1 = (x0y0 + x1y1 - (x1-x0)*(y1-y0))<<B */
  if (mp_lshd (&x1y1, B * 2) != MP_OKAY)
    goto X1Y1;          /* x1y1 = x1y1 << 2*B */

  if (mp_add (&x0y0, &t1, &t1) != MP_OKAY)
    goto X1Y1;          /* t1 = x0y0 + t1 */
  if (mp_add (&t1, &x1y1, c) != MP_OKAY)
    goto X1Y1;          /* t1 = x0y0 + t1 + x1y1 */

  /* Algorithm succeeded set the return code to MP_OKAY */
  err = MP_OKAY;

X1Y1:mp_clear (&x1y1);
X0Y0:mp_clear (&x0y0);
T1:mp_clear (&t1);
Y1:mp_clear (&y1);
Y0:mp_clear (&y0);
X1:mp_clear (&x1);
X0:mp_clear (&x0);
ERR:
  return err;
}
#endif

/* End: bn_mp_karatsuba_mul.c */

/* Start: bn_mp_karatsuba_sqr.c */
#include <tommath.h>
#ifdef BN_MP_KARATSUBA_SQR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Karatsuba squaring, computes b = a*a using three 
 * half size squarings
 *
 * See comments of karatsuba_mul for details.  It 
 * is essentially the same algorithm but merely 
 * tuned to perform recursive squarings.
 */
int mp_karatsuba_sqr (mp_int * a, mp_int * b)
{
  mp_int  x0, x1, t1, t2, x0x0, x1x1;
  int     B, err;

  err = MP_MEM;

  /* min # of digits */
  B = a->used;

  /* now divide in two */
  B = B >> 1;

  /* init copy all the temps */
  if (mp_init_size (&x0, B) != MP_OKAY)
    goto ERR;
  if (mp_init_size (&x1, a->used - B) != MP_OKAY)
    goto X0;

  /* init temps */
  if (mp_init_size (&t1, a->used * 2) != MP_OKAY)
    goto X1;
  if (mp_init_size (&t2, a->used * 2) != MP_OKAY)
    goto T1;
  if (mp_init_size (&x0x0, B * 2) != MP_OKAY)
    goto T2;
  if (mp_init_size (&x1x1, (a->used - B) * 2) != MP_OKAY)
    goto X0X0;

  {
    register int x;
    register mp_digit *dst, *src;

    src = a->dp;

    /* now shift the digits */
    dst = x0.dp;
    for (x = 0; x < B; x++) {
      *dst++ = *src++;
    }

    dst = x1.dp;
    for (x = B; x < a->used; x++) {
      *dst++ = *src++;
    }
  }

  x0.used = B;
  x1.used = a->used - B;

  mp_clamp (&x0);

  /* now calc the products x0*x0 and x1*x1 */
  if (mp_sqr (&x0, &x0x0) != MP_OKAY)
    goto X1X1;           /* x0x0 = x0*x0 */
  if (mp_sqr (&x1, &x1x1) != MP_OKAY)
    goto X1X1;           /* x1x1 = x1*x1 */

  /* now calc (x1+x0)**2 */
  if (s_mp_add (&x1, &x0, &t1) != MP_OKAY)
    goto X1X1;           /* t1 = x1 - x0 */
  if (mp_sqr (&t1, &t1) != MP_OKAY)
    goto X1X1;           /* t1 = (x1 - x0) * (x1 - x0) */

  /* add x0y0 */
  if (s_mp_add (&x0x0, &x1x1, &t2) != MP_OKAY)
    goto X1X1;           /* t2 = x0x0 + x1x1 */
  if (s_mp_sub (&t1, &t2, &t1) != MP_OKAY)
    goto X1X1;           /* t1 = (x1+x0)**2 - (x0x0 + x1x1) */

  /* shift by B */
  if (mp_lshd (&t1, B) != MP_OKAY)
    goto X1X1;           /* t1 = (x0x0 + x1x1 - (x1-x0)*(x1-x0))<<B */
  if (mp_lshd (&x1x1, B * 2) != MP_OKAY)
    goto X1X1;           /* x1x1 = x1x1 << 2*B */

  if (mp_add (&x0x0, &t1, &t1) != MP_OKAY)
    goto X1X1;           /* t1 = x0x0 + t1 */
  if (mp_add (&t1, &x1x1, b) != MP_OKAY)
    goto X1X1;           /* t1 = x0x0 + t1 + x1x1 */

  err = MP_OKAY;

X1X1:mp_clear (&x1x1);
X0X0:mp_clear (&x0x0);
T2:mp_clear (&t2);
T1:mp_clear (&t1);
X1:mp_clear (&x1);
X0:mp_clear (&x0);
ERR:
  return err;
}
#endif

/* End: bn_mp_karatsuba_sqr.c */

/* Start: bn_mp_lcm.c */
#include <tommath.h>
#ifdef BN_MP_LCM_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes least common multiple as |a*b|/(a, b) */
int mp_lcm (mp_int * a, mp_int * b, mp_int * c)
{
  int     res;
  mp_int  t1, t2;


  if ((res = mp_init_multi (&t1, &t2, NULL)) != MP_OKAY) {
    return res;
  }

  /* t1 = get the GCD of the two inputs */
  if ((res = mp_gcd (a, b, &t1)) != MP_OKAY) {
    goto LBL_T;
  }

  /* divide the smallest by the GCD */
  if (mp_cmp_mag(a, b) == MP_LT) {
     /* store quotient in t2 such that t2 * b is the LCM */
     if ((res = mp_div(a, &t1, &t2, NULL)) != MP_OKAY) {
        goto LBL_T;
     }
     res = mp_mul(b, &t2, c);
  } else {
     /* store quotient in t2 such that t2 * a is the LCM */
     if ((res = mp_div(b, &t1, &t2, NULL)) != MP_OKAY) {
        goto LBL_T;
     }
     res = mp_mul(a, &t2, c);
  }

  /* fix the sign to positive */
  c->sign = MP_ZPOS;

LBL_T:
  mp_clear_multi (&t1, &t2, NULL);
  return res;
}
#endif

/* End: bn_mp_lcm.c */

/* Start: bn_mp_lshd.c */
#include <tommath.h>
#ifdef BN_MP_LSHD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* shift left a certain amount of digits */
int mp_lshd (mp_int * a, int b)
{
  int     x, res;

  /* if its less than zero return */
  if (b <= 0) {
    return MP_OKAY;
  }

  /* grow to fit the new digits */
  if (a->alloc < a->used + b) {
     if ((res = mp_grow (a, a->used + b)) != MP_OKAY) {
       return res;
     }
  }

  {
    register mp_digit *top, *bottom;

    /* increment the used by the shift amount then copy upwards */
    a->used += b;

    /* top */
    top = a->dp + a->used - 1;

    /* base */
    bottom = a->dp + a->used - 1 - b;

    /* much like mp_rshd this is implemented using a sliding window
     * except the window goes the otherway around.  Copying from
     * the bottom to the top.  see bn_mp_rshd.c for more info.
     */
    for (x = a->used - 1; x >= b; x--) {
      *top-- = *bottom--;
    }

    /* zero the lower digits */
    top = a->dp;
    for (x = 0; x < b; x++) {
      *top++ = 0;
    }
  }
  return MP_OKAY;
}
#endif

/* End: bn_mp_lshd.c */

/* Start: bn_mp_mod.c */
#include <tommath.h>
#ifdef BN_MP_MOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* c = a mod b, 0 <= c < b */
int
mp_mod (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int  t;
  int     res;

  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_div (a, b, NULL, &t)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }

  if (t.sign != b->sign) {
    res = mp_add (b, &t, c);
  } else {
    res = MP_OKAY;
    mp_exch (&t, c);
  }

  mp_clear (&t);
  return res;
}
#endif

/* End: bn_mp_mod.c */

/* Start: bn_mp_mod_2d.c */
#include <tommath.h>
#ifdef BN_MP_MOD_2D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* calc a value mod 2**b */
int
mp_mod_2d (mp_int * a, int b, mp_int * c)
{
  int     x, res;

  /* if b is <= 0 then zero the int */
  if (b <= 0) {
    mp_zero (c);
    return MP_OKAY;
  }

  /* if the modulus is larger than the value than return */
  if (b >= (int) (a->used * DIGIT_BIT)) {
    res = mp_copy (a, c);
    return res;
  }

  /* copy */
  if ((res = mp_copy (a, c)) != MP_OKAY) {
    return res;
  }

  /* zero digits above the last digit of the modulus */
  for (x = (b / DIGIT_BIT) + ((b % DIGIT_BIT) == 0 ? 0 : 1); x < c->used; x++) {
    c->dp[x] = 0;
  }
  /* clear the digit that is not completely outside/inside the modulus */
  c->dp[b / DIGIT_BIT] &=
    (mp_digit) ((((mp_digit) 1) << (((mp_digit) b) % DIGIT_BIT)) - ((mp_digit) 1));
  mp_clamp (c);
  return MP_OKAY;
}
#endif

/* End: bn_mp_mod_2d.c */

/* Start: bn_mp_mod_d.c */
#include <tommath.h>
#ifdef BN_MP_MOD_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

int
mp_mod_d (mp_int * a, mp_digit b, mp_digit * c)
{
  return mp_div_d(a, b, NULL, c);
}
#endif

/* End: bn_mp_mod_d.c */

/* Start: bn_mp_montgomery_calc_normalization.c */
#include <tommath.h>
#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/*
 * shifts with subtractions when the result is greater than b.
 *
 * The method is slightly modified to shift B unconditionally upto just under
 * the leading bit of b.  This saves alot of multiple precision shifting.
 */
int mp_montgomery_calc_normalization (mp_int * a, mp_int * b)
{
  int     x, bits, res;

  /* how many bits of last digit does b use */
  bits = mp_count_bits (b) % DIGIT_BIT;

  if (b->used > 1) {
     if ((res = mp_2expt (a, (b->used - 1) * DIGIT_BIT + bits - 1)) != MP_OKAY) {
        return res;
     }
  } else {
     mp_set(a, 1);
     bits = 1;
  }


  /* now compute C = A * B mod b */
  for (x = bits - 1; x < (int)DIGIT_BIT; x++) {
    if ((res = mp_mul_2 (a, a)) != MP_OKAY) {
      return res;
    }
    if (mp_cmp_mag (a, b) != MP_LT) {
      if ((res = s_mp_sub (a, b, a)) != MP_OKAY) {
        return res;
      }
    }
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_montgomery_calc_normalization.c */

/* Start: bn_mp_montgomery_reduce.c */
#include <tommath.h>
#ifdef BN_MP_MONTGOMERY_REDUCE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes xR**-1 == x (mod N) via Montgomery Reduction */
int
mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho)
{
  int     ix, res, digs;
  mp_digit mu;

  /* can the fast reduction [comba] method be used?
   *
   * Note that unlike in mul you're safely allowed *less*
   * than the available columns [255 per default] since carries
   * are fixed up in the inner loop.
   */
  digs = n->used * 2 + 1;
  if ((digs < MP_WARRAY) &&
      n->used <
      (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
    return fast_mp_montgomery_reduce (x, n, rho);
  }

  /* grow the input as required */
  if (x->alloc < digs) {
    if ((res = mp_grow (x, digs)) != MP_OKAY) {
      return res;
    }
  }
  x->used = digs;

  for (ix = 0; ix < n->used; ix++) {
    /* mu = ai * rho mod b
     *
     * The value of rho must be precalculated via
     * montgomery_setup() such that
     * it equals -1/n0 mod b this allows the
     * following inner loop to reduce the
     * input one digit at a time
     */
    mu = (mp_digit) (((mp_word)x->dp[ix]) * ((mp_word)rho) & MP_MASK);

    /* a = a + mu * m * b**i */
    {
      register int iy;
      register mp_digit *tmpn, *tmpx, u;
      register mp_word r;

      /* alias for digits of the modulus */
      tmpn = n->dp;

      /* alias for the digits of x [the input] */
      tmpx = x->dp + ix;

      /* set the carry to zero */
      u = 0;

      /* Multiply and add in place */
      for (iy = 0; iy < n->used; iy++) {
        /* compute product and sum */
        r       = ((mp_word)mu) * ((mp_word)*tmpn++) +
                  ((mp_word) u) + ((mp_word) * tmpx);

        /* get carry */
        u       = (mp_digit)(r >> ((mp_word) DIGIT_BIT));

        /* fix digit */
        *tmpx++ = (mp_digit)(r & ((mp_word) MP_MASK));
      }
      /* At this point the ix'th digit of x should be zero */


      /* propagate carries upwards as required*/
      while (u) {
        *tmpx   += u;
        u        = *tmpx >> DIGIT_BIT;
        *tmpx++ &= MP_MASK;
      }
    }
  }

  /* at this point the n.used'th least
   * significant digits of x are all zero
   * which means we can shift x to the
   * right by n.used digits and the
   * residue is unchanged.
   */

  /* x = x/b**n.used */
  mp_clamp(x);
  mp_rshd (x, n->used);

  /* if x >= n then x = x - n */
  if (mp_cmp_mag (x, n) != MP_LT) {
    return s_mp_sub (x, n, x);
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_montgomery_reduce.c */

/* Start: bn_mp_montgomery_setup.c */
#include <tommath.h>
#ifdef BN_MP_MONTGOMERY_SETUP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* setups the montgomery reduction stuff */
int
mp_montgomery_setup (mp_int * n, mp_digit * rho)
{
  mp_digit x, b;

/* fast inversion mod 2**k
 *
 * Based on the fact that
 *
 * XA = 1 (mod 2**n)  =>  (X(2-XA)) A = 1 (mod 2**2n)
 *                    =>  2*X*A - X*X*A*A = 1
 *                    =>  2*(1) - (1)     = 1
 */
  b = n->dp[0];

  if ((b & 1) == 0) {
    return MP_VAL;
  }

  x = (((b + 2) & 4) << 1) + b; /* here x*a==1 mod 2**4 */
  x *= 2 - b * x;               /* here x*a==1 mod 2**8 */
#if !defined(MP_8BIT)
  x *= 2 - b * x;               /* here x*a==1 mod 2**16 */
#endif
#if defined(MP_64BIT) || !(defined(MP_8BIT) || defined(MP_16BIT))
  x *= 2 - b * x;               /* here x*a==1 mod 2**32 */
#endif
#ifdef MP_64BIT
  x *= 2 - b * x;               /* here x*a==1 mod 2**64 */
#endif

  /* rho = -1/m mod b */
  *rho = (unsigned long)(((mp_word)1 << ((mp_word) DIGIT_BIT)) - x) & MP_MASK;

  return MP_OKAY;
}
#endif

/* End: bn_mp_montgomery_setup.c */

/* Start: bn_mp_mul.c */
#include <tommath.h>
#ifdef BN_MP_MUL_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* high level multiplication (handles sign) */
int mp_mul (mp_int * a, mp_int * b, mp_int * c)
{
  int     res, neg;
  neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG;

  /* use Toom-Cook? */
#ifdef BN_MP_TOOM_MUL_C
  if (MIN (a->used, b->used) >= TOOM_MUL_CUTOFF) {
    res = mp_toom_mul(a, b, c);
  } else 
#endif
#ifdef BN_MP_KARATSUBA_MUL_C
  /* use Karatsuba? */
  if (MIN (a->used, b->used) >= KARATSUBA_MUL_CUTOFF) {
    res = mp_karatsuba_mul (a, b, c);
  } else 
#endif
  {
    /* can we use the fast multiplier?
     *
     * The fast multiplier can be used if the output will 
     * have less than MP_WARRAY digits and the number of 
     * digits won't affect carry propagation
     */
    int     digs = a->used + b->used + 1;

#ifdef BN_FAST_S_MP_MUL_DIGS_C
    if ((digs < MP_WARRAY) &&
        MIN(a->used, b->used) <= 
        (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
      res = fast_s_mp_mul_digs (a, b, c, digs);
    } else 
#endif
#ifdef BN_S_MP_MUL_DIGS_C
      res = s_mp_mul (a, b, c); /* uses s_mp_mul_digs */
#else
      res = MP_VAL;
#endif

  }
  c->sign = (c->used > 0) ? neg : MP_ZPOS;
  return res;
}
#endif

/* End: bn_mp_mul.c */

/* Start: bn_mp_mul_2.c */
#include <tommath.h>
#ifdef BN_MP_MUL_2_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* b = a*2 */
int mp_mul_2(mp_int * a, mp_int * b)
{
  int     x, res, oldused;

  /* grow to accomodate result */
  if (b->alloc < a->used + 1) {
    if ((res = mp_grow (b, a->used + 1)) != MP_OKAY) {
      return res;
    }
  }

  oldused = b->used;
  b->used = a->used;

  {
    register mp_digit r, rr, *tmpa, *tmpb;

    /* alias for source */
    tmpa = a->dp;
    
    /* alias for dest */
    tmpb = b->dp;

    /* carry */
    r = 0;
    for (x = 0; x < a->used; x++) {
    
      /* get what will be the *next* carry bit from the 
       * MSB of the current digit 
       */
      rr = *tmpa >> ((mp_digit)(DIGIT_BIT - 1));
      
      /* now shift up this digit, add in the carry [from the previous] */
      *tmpb++ = ((*tmpa++ << ((mp_digit)1)) | r) & MP_MASK;
      
      /* copy the carry that would be from the source 
       * digit into the next iteration 
       */
      r = rr;
    }

    /* new leading digit? */
    if (r != 0) {
      /* add a MSB which is always 1 at this point */
      *tmpb = 1;
      ++(b->used);
    }

    /* now zero any excess digits on the destination 
     * that we didn't write to 
     */
    tmpb = b->dp + b->used;
    for (x = b->used; x < oldused; x++) {
      *tmpb++ = 0;
    }
  }
  b->sign = a->sign;
  return MP_OKAY;
}
#endif

/* End: bn_mp_mul_2.c */

/* Start: bn_mp_mul_2d.c */
#include <tommath.h>
#ifdef BN_MP_MUL_2D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* shift left by a certain bit count */
int mp_mul_2d (mp_int * a, int b, mp_int * c)
{
  mp_digit d;
  int      res;

  /* copy */
  if (a != c) {
     if ((res = mp_copy (a, c)) != MP_OKAY) {
       return res;
     }
  }

  if (c->alloc < (int)(c->used + b/DIGIT_BIT + 1)) {
     if ((res = mp_grow (c, c->used + b / DIGIT_BIT + 1)) != MP_OKAY) {
       return res;
     }
  }

  /* shift by as many digits in the bit count */
  if (b >= (int)DIGIT_BIT) {
    if ((res = mp_lshd (c, b / DIGIT_BIT)) != MP_OKAY) {
      return res;
    }
  }

  /* shift any bit count < DIGIT_BIT */
  d = (mp_digit) (b % DIGIT_BIT);
  if (d != 0) {
    register mp_digit *tmpc, shift, mask, r, rr;
    register int x;

    /* bitmask for carries */
    mask = (((mp_digit)1) << d) - 1;

    /* shift for msbs */
    shift = DIGIT_BIT - d;

    /* alias */
    tmpc = c->dp;

    /* carry */
    r    = 0;
    for (x = 0; x < c->used; x++) {
      /* get the higher bits of the current word */
      rr = (*tmpc >> shift) & mask;

      /* shift the current word and OR in the carry */
      *tmpc = ((*tmpc << d) | r) & MP_MASK;
      ++tmpc;

      /* set the carry to the carry bits of the current word */
      r = rr;
    }
    
    /* set final carry */
    if (r != 0) {
       c->dp[(c->used)++] = r;
    }
  }
  mp_clamp (c);
  return MP_OKAY;
}
#endif

/* End: bn_mp_mul_2d.c */

/* Start: bn_mp_mul_d.c */
#include <tommath.h>
#ifdef BN_MP_MUL_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* multiply by a digit */
int
mp_mul_d (mp_int * a, mp_digit b, mp_int * c)
{
  mp_digit u, *tmpa, *tmpc;
  mp_word  r;
  int      ix, res, olduse;

  /* make sure c is big enough to hold a*b */
  if (c->alloc < a->used + 1) {
    if ((res = mp_grow (c, a->used + 1)) != MP_OKAY) {
      return res;
    }
  }

  /* get the original destinations used count */
  olduse = c->used;

  /* set the sign */
  c->sign = a->sign;

  /* alias for a->dp [source] */
  tmpa = a->dp;

  /* alias for c->dp [dest] */
  tmpc = c->dp;

  /* zero carry */
  u = 0;

  /* compute columns */
  for (ix = 0; ix < a->used; ix++) {
    /* compute product and carry sum for this term */
    r       = ((mp_word) u) + ((mp_word)*tmpa++) * ((mp_word)b);

    /* mask off higher bits to get a single digit */
    *tmpc++ = (mp_digit) (r & ((mp_word) MP_MASK));

    /* send carry into next iteration */
    u       = (mp_digit) (r >> ((mp_word) DIGIT_BIT));
  }

  /* store final carry [if any] and increment ix offset  */
  *tmpc++ = u;
  ++ix;

  /* now zero digits above the top */
  while (ix++ < olduse) {
     *tmpc++ = 0;
  }

  /* set used count */
  c->used = a->used + 1;
  mp_clamp(c);

  return MP_OKAY;
}
#endif

/* End: bn_mp_mul_d.c */

/* Start: bn_mp_mulmod.c */
#include <tommath.h>
#ifdef BN_MP_MULMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* d = a * b (mod c) */
int mp_mulmod (mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
  int     res;
  mp_int  t;

  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_mul (a, b, &t)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }
  res = mp_mod (&t, c, d);
  mp_clear (&t);
  return res;
}
#endif

/* End: bn_mp_mulmod.c */

/* Start: bn_mp_n_root.c */
#include <tommath.h>
#ifdef BN_MP_N_ROOT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* find the n'th root of an integer 
 *
 * Result found such that (c)**b <= a and (c+1)**b > a 
 *
 * This algorithm uses Newton's approximation 
 * x[i+1] = x[i] - f(x[i])/f'(x[i]) 
 * which will find the root in log(N) time where 
 * each step involves a fair bit.  This is not meant to 
 * find huge roots [square and cube, etc].
 */
int mp_n_root (mp_int * a, mp_digit b, mp_int * c)
{
  mp_int  t1, t2, t3;
  int     res, neg;

  /* input must be positive if b is even */
  if ((b & 1) == 0 && a->sign == MP_NEG) {
    return MP_VAL;
  }

  if ((res = mp_init (&t1)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init (&t2)) != MP_OKAY) {
    goto LBL_T1;
  }

  if ((res = mp_init (&t3)) != MP_OKAY) {
    goto LBL_T2;
  }

  /* if a is negative fudge the sign but keep track */
  neg     = a->sign;
  a->sign = MP_ZPOS;

  /* t2 = 2 */
  mp_set (&t2, 2);

  do {
    /* t1 = t2 */
    if ((res = mp_copy (&t2, &t1)) != MP_OKAY) {
      goto LBL_T3;
    }

    /* t2 = t1 - ((t1**b - a) / (b * t1**(b-1))) */
    
    /* t3 = t1**(b-1) */
    if ((res = mp_expt_d (&t1, b - 1, &t3)) != MP_OKAY) {   
      goto LBL_T3;
    }

    /* numerator */
    /* t2 = t1**b */
    if ((res = mp_mul (&t3, &t1, &t2)) != MP_OKAY) {    
      goto LBL_T3;
    }

    /* t2 = t1**b - a */
    if ((res = mp_sub (&t2, a, &t2)) != MP_OKAY) {  
      goto LBL_T3;
    }

    /* denominator */
    /* t3 = t1**(b-1) * b  */
    if ((res = mp_mul_d (&t3, b, &t3)) != MP_OKAY) {    
      goto LBL_T3;
    }

    /* t3 = (t1**b - a)/(b * t1**(b-1)) */
    if ((res = mp_div (&t2, &t3, &t3, NULL)) != MP_OKAY) {  
      goto LBL_T3;
    }

    if ((res = mp_sub (&t1, &t3, &t2)) != MP_OKAY) {
      goto LBL_T3;
    }
  }  while (mp_cmp (&t1, &t2) != MP_EQ);

  /* result can be off by a few so check */
  for (;;) {
    if ((res = mp_expt_d (&t1, b, &t2)) != MP_OKAY) {
      goto LBL_T3;
    }

    if (mp_cmp (&t2, a) == MP_GT) {
      if ((res = mp_sub_d (&t1, 1, &t1)) != MP_OKAY) {
         goto LBL_T3;
      }
    } else {
      break;
    }
  }

  /* reset the sign of a first */
  a->sign = neg;

  /* set the result */
  mp_exch (&t1, c);

  /* set the sign of the result */
  c->sign = neg;

  res = MP_OKAY;

LBL_T3:mp_clear (&t3);
LBL_T2:mp_clear (&t2);
LBL_T1:mp_clear (&t1);
  return res;
}
#endif

/* End: bn_mp_n_root.c */

/* Start: bn_mp_neg.c */
#include <tommath.h>
#ifdef BN_MP_NEG_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* b = -a */
int mp_neg (mp_int * a, mp_int * b)
{
  int     res;
  if (a != b) {
     if ((res = mp_copy (a, b)) != MP_OKAY) {
        return res;
     }
  }

  if (mp_iszero(b) != MP_YES) {
     b->sign = (a->sign == MP_ZPOS) ? MP_NEG : MP_ZPOS;
  } else {
     b->sign = MP_ZPOS;
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_neg.c */

/* Start: bn_mp_or.c */
#include <tommath.h>
#ifdef BN_MP_OR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* OR two ints together */
int mp_or (mp_int * a, mp_int * b, mp_int * c)
{
  int     res, ix, px;
  mp_int  t, *x;

  if (a->used > b->used) {
    if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
      return res;
    }
    px = b->used;
    x = b;
  } else {
    if ((res = mp_init_copy (&t, b)) != MP_OKAY) {
      return res;
    }
    px = a->used;
    x = a;
  }

  for (ix = 0; ix < px; ix++) {
    t.dp[ix] |= x->dp[ix];
  }
  mp_clamp (&t);
  mp_exch (c, &t);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_mp_or.c */

/* Start: bn_mp_prime_fermat.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_FERMAT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* performs one Fermat test.
 * 
 * If "a" were prime then b**a == b (mod a) since the order of
 * the multiplicative sub-group would be phi(a) = a-1.  That means
 * it would be the same as b**(a mod (a-1)) == b**1 == b (mod a).
 *
 * Sets result to 1 if the congruence holds, or zero otherwise.
 */
int mp_prime_fermat (mp_int * a, mp_int * b, int *result)
{
  mp_int  t;
  int     err;

  /* default to composite  */
  *result = MP_NO;

  /* ensure b > 1 */
  if (mp_cmp_d(b, 1) != MP_GT) {
     return MP_VAL;
  }

  /* init t */
  if ((err = mp_init (&t)) != MP_OKAY) {
    return err;
  }

  /* compute t = b**a mod a */
  if ((err = mp_exptmod (b, a, a, &t)) != MP_OKAY) {
    goto LBL_T;
  }

  /* is it equal to b? */
  if (mp_cmp (&t, b) == MP_EQ) {
    *result = MP_YES;
  }

  err = MP_OKAY;
LBL_T:mp_clear (&t);
  return err;
}
#endif

/* End: bn_mp_prime_fermat.c */

/* Start: bn_mp_prime_is_divisible.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_IS_DIVISIBLE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, tomstden[email protected], http://math.libtomcrypt.com
 */

/* determines if an integers is divisible by one 
 * of the first PRIME_SIZE primes or not
 *
 * sets result to 0 if not, 1 if yes
 */
int mp_prime_is_divisible (mp_int * a, int *result)
{
  int     err, ix;
  mp_digit res;

  /* default to not */
  *result = MP_NO;

  for (ix = 0; ix < PRIME_SIZE; ix++) {
    /* what is a mod LBL_prime_tab[ix] */
    if ((err = mp_mod_d (a, ltm_prime_tab[ix], &res)) != MP_OKAY) {
      return err;
    }

    /* is the residue zero? */
    if (res == 0) {
      *result = MP_YES;
      return MP_OKAY;
    }
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_prime_is_divisible.c */

/* Start: bn_mp_prime_is_prime.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_IS_PRIME_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* performs a variable number of rounds of Miller-Rabin
 *
 * Probability of error after t rounds is no more than

 *
 * Sets result to 1 if probably prime, 0 otherwise
 */
int mp_prime_is_prime (mp_int * a, int t, int *result)
{
  mp_int  b;
  int     ix, err, res;

  /* default to no */
  *result = MP_NO;

  /* valid value of t? */
  if (t <= 0 || t > PRIME_SIZE) {
    return MP_VAL;
  }

  /* is the input equal to one of the primes in the table? */
  for (ix = 0; ix < PRIME_SIZE; ix++) {
      if (mp_cmp_d(a, ltm_prime_tab[ix]) == MP_EQ) {
         *result = 1;
         return MP_OKAY;
      }
  }

  /* first perform trial division */
  if ((err = mp_prime_is_divisible (a, &res)) != MP_OKAY) {
    return err;
  }

  /* return if it was trivially divisible */
  if (res == MP_YES) {
    return MP_OKAY;
  }

  /* now perform the miller-rabin rounds */
  if ((err = mp_init (&b)) != MP_OKAY) {
    return err;
  }

  for (ix = 0; ix < t; ix++) {
    /* set the prime */
    mp_set (&b, ltm_prime_tab[ix]);

    if ((err = mp_prime_miller_rabin (a, &b, &res)) != MP_OKAY) {
      goto LBL_B;
    }

    if (res == MP_NO) {
      goto LBL_B;
    }
  }

  /* passed the test */
  *result = MP_YES;
LBL_B:mp_clear (&b);
  return err;
}
#endif

/* End: bn_mp_prime_is_prime.c */

/* Start: bn_mp_prime_miller_rabin.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_MILLER_RABIN_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Miller-Rabin test of "a" to the base of "b" as described in 
 * HAC pp. 139 Algorithm 4.24
 *
 * Sets result to 0 if definitely composite or 1 if probably prime.
 * Randomly the chance of error is no more than 1/4 and often 
 * very much lower.
 */
int mp_prime_miller_rabin (mp_int * a, mp_int * b, int *result)
{
  mp_int  n1, y, r;
  int     s, j, err;

  /* default */
  *result = MP_NO;

  /* ensure b > 1 */
  if (mp_cmp_d(b, 1) != MP_GT) {
     return MP_VAL;
  }     

  /* get n1 = a - 1 */
  if ((err = mp_init_copy (&n1, a)) != MP_OKAY) {
    return err;
  }
  if ((err = mp_sub_d (&n1, 1, &n1)) != MP_OKAY) {
    goto LBL_N1;
  }

  /* set 2**s * r = n1 */
  if ((err = mp_init_copy (&r, &n1)) != MP_OKAY) {
    goto LBL_N1;
  }

  /* count the number of least significant bits
   * which are zero
   */
  s = mp_cnt_lsb(&r);

  /* now divide n - 1 by 2**s */
  if ((err = mp_div_2d (&r, s, &r, NULL)) != MP_OKAY) {
    goto LBL_R;
  }

  /* compute y = b**r mod a */
  if ((err = mp_init (&y)) != MP_OKAY) {
    goto LBL_R;
  }
  if ((err = mp_exptmod (b, &r, a, &y)) != MP_OKAY) {
    goto LBL_Y;
  }

  /* if y != 1 and y != n1 do */
  if (mp_cmp_d (&y, 1) != MP_EQ && mp_cmp (&y, &n1) != MP_EQ) {
    j = 1;
    /* while j <= s-1 and y != n1 */
    while ((j <= (s - 1)) && mp_cmp (&y, &n1) != MP_EQ) {
      if ((err = mp_sqrmod (&y, a, &y)) != MP_OKAY) {
         goto LBL_Y;
      }

      /* if y == 1 then composite */
      if (mp_cmp_d (&y, 1) == MP_EQ) {
         goto LBL_Y;
      }

      ++j;
    }

    /* if y != n1 then composite */
    if (mp_cmp (&y, &n1) != MP_EQ) {
      goto LBL_Y;
    }
  }

  /* probably prime now */
  *result = MP_YES;
LBL_Y:mp_clear (&y);
LBL_R:mp_clear (&r);
LBL_N1:mp_clear (&n1);
  return err;
}
#endif

/* End: bn_mp_prime_miller_rabin.c */

/* Start: bn_mp_prime_next_prime.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_NEXT_PRIME_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* finds the next prime after the number "a" using "t" trials
 * of Miller-Rabin.
 *
 * bbs_style = 1 means the prime must be congruent to 3 mod 4
 */
int mp_prime_next_prime(mp_int *a, int t, int bbs_style)
{
   int      err, res, x, y;
   mp_digit res_tab[PRIME_SIZE], step, kstep;
   mp_int   b;

   /* ensure t is valid */
   if (t <= 0 || t > PRIME_SIZE) {
      return MP_VAL;
   }

   /* force positive */
   a->sign = MP_ZPOS;

   /* simple algo if a is less than the largest prime in the table */
   if (mp_cmp_d(a, ltm_prime_tab[PRIME_SIZE-1]) == MP_LT) {
      /* find which prime it is bigger than */
      for (x = PRIME_SIZE - 2; x >= 0; x--) {
          if (mp_cmp_d(a, ltm_prime_tab[x]) != MP_LT) {
             if (bbs_style == 1) {
                /* ok we found a prime smaller or
                 * equal [so the next is larger]
                 *
                 * however, the prime must be
                 * congruent to 3 mod 4
                 */
                if ((ltm_prime_tab[x + 1] & 3) != 3) {
                   /* scan upwards for a prime congruent to 3 mod 4 */
                   for (y = x + 1; y < PRIME_SIZE; y++) {
                       if ((ltm_prime_tab[y] & 3) == 3) {
                          mp_set(a, ltm_prime_tab[y]);
                          return MP_OKAY;
                       }
                   }
                }
             } else {
                mp_set(a, ltm_prime_tab[x + 1]);
                return MP_OKAY;
             }
          }
      }
      /* at this point a maybe 1 */
      if (mp_cmp_d(a, 1) == MP_EQ) {
         mp_set(a, 2);
         return MP_OKAY;
      }
      /* fall through to the sieve */
   }

   /* generate a prime congruent to 3 mod 4 or 1/3 mod 4? */
   if (bbs_style == 1) {
      kstep   = 4;
   } else {
      kstep   = 2;
   }

   /* at this point we will use a combination of a sieve and Miller-Rabin */

   if (bbs_style == 1) {
      /* if a mod 4 != 3 subtract the correct value to make it so */
      if ((a->dp[0] & 3) != 3) {
         if ((err = mp_sub_d(a, (a->dp[0] & 3) + 1, a)) != MP_OKAY) { return err; };
      }
   } else {
      if (mp_iseven(a) == 1) {
         /* force odd */
         if ((err = mp_sub_d(a, 1, a)) != MP_OKAY) {
            return err;
         }
      }
   }

   /* generate the restable */
   for (x = 1; x < PRIME_SIZE; x++) {
      if ((err = mp_mod_d(a, ltm_prime_tab[x], res_tab + x)) != MP_OKAY) {
         return err;
      }
   }

   /* init temp used for Miller-Rabin Testing */
   if ((err = mp_init(&b)) != MP_OKAY) {
      return err;
   }

   for (;;) {
      /* skip to the next non-trivially divisible candidate */
      step = 0;
      do {
         /* y == 1 if any residue was zero [e.g. cannot be prime] */
         y     =  0;

         /* increase step to next candidate */
         step += kstep;

         /* compute the new residue without using division */
         for (x = 1; x < PRIME_SIZE; x++) {
             /* add the step to each residue */
             res_tab[x] += kstep;

             /* subtract the modulus [instead of using division] */
             if (res_tab[x] >= ltm_prime_tab[x]) {
                res_tab[x]  -= ltm_prime_tab[x];
             }

             /* set flag if zero */
             if (res_tab[x] == 0) {
                y = 1;
             }
         }
      } while (y == 1 && step < ((((mp_digit)1)<<DIGIT_BIT) - kstep));

      /* add the step */
      if ((err = mp_add_d(a, step, a)) != MP_OKAY) {
         goto LBL_ERR;
      }

      /* if didn't pass sieve and step == MAX then skip test */
      if (y == 1 && step >= ((((mp_digit)1)<<DIGIT_BIT) - kstep)) {
         continue;
      }

      /* is this prime? */
      for (x = 0; x < t; x++) {
          mp_set(&b, ltm_prime_tab[x]);
          if ((err = mp_prime_miller_rabin(a, &b, &res)) != MP_OKAY) {
             goto LBL_ERR;
          }
          if (res == MP_NO) {
             break;
          }
      }

      if (res == MP_YES) {
         break;
      }
   }

   err = MP_OKAY;
LBL_ERR:
   mp_clear(&b);
   return err;
}

#endif

/* End: bn_mp_prime_next_prime.c */

/* Start: bn_mp_prime_rabin_miller_trials.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_RABIN_MILLER_TRIALS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */


static const struct {
   int k, t;
} sizes[] = {
{   128,    28 },
{   256,    16 },
{   384,    10 },
{   512,     7 },
{   640,     6 },
{   768,     5 },
{   896,     4 },
{  1024,     4 }
};

/* returns # of RM trials required for a given bit size */
int mp_prime_rabin_miller_trials(int size)
{
   int x;

   for (x = 0; x < (int)(sizeof(sizes)/(sizeof(sizes[0]))); x++) {
       if (sizes[x].k == size) {
          return sizes[x].t;
       } else if (sizes[x].k > size) {
          return (x == 0) ? sizes[0].t : sizes[x - 1].t;
       }
   }
   return sizes[x-1].t + 1;
}


#endif

/* End: bn_mp_prime_rabin_miller_trials.c */

/* Start: bn_mp_prime_random_ex.c */
#include <tommath.h>
#ifdef BN_MP_PRIME_RANDOM_EX_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* makes a truly random prime of a given size (bits),
 *
 * Flags are as follows:
 * 
 *   LTM_PRIME_BBS      - make prime congruent to 3 mod 4
 *   LTM_PRIME_SAFE     - make sure (p-1)/2 is prime as well (implies LTM_PRIME_BBS)
 *   LTM_PRIME_2MSB_OFF - make the 2nd highest bit zero
 *   LTM_PRIME_2MSB_ON  - make the 2nd highest bit one
 *
 * You have to supply a callback which fills in a buffer with random bytes.  "dat" is a parameter you can
 * have passed to the callback (e.g. a state or something).  This function doesn't use "dat" itself
 * so it can be NULL
 *
 */

/* This is possibly the mother of all prime generation functions, muahahahahaha! */
int mp_prime_random_ex(mp_int *a, int t, int size, int flags, ltm_prime_callback cb, void *dat)
{
   unsigned char *tmp, maskAND, maskOR_msb, maskOR_lsb;
   int res, err, bsize, maskOR_msb_offset;

   /* sanity check the input */
   if (size <= 1 || t <= 0) {
      return MP_VAL;
   }

   /* LTM_PRIME_SAFE implies LTM_PRIME_BBS */
   if (flags & LTM_PRIME_SAFE) {
      flags |= LTM_PRIME_BBS;
   }

   /* calc the byte size */
   bsize = (size>>3) + ((size&7)?1:0);

   /* we need a buffer of bsize bytes */
   tmp = OPT_CAST(unsigned char) XMALLOC(bsize);
   if (tmp == NULL) {
      return MP_MEM;
   }

   /* calc the maskAND value for the MSbyte*/
   maskAND = ((size&7) == 0) ? 0xFF : (0xFF >> (8 - (size & 7)));

   /* calc the maskOR_msb */
   maskOR_msb        = 0;
   maskOR_msb_offset = ((size & 7) == 1) ? 1 : 0;
   if (flags & LTM_PRIME_2MSB_ON) {
      maskOR_msb       |= 0x80 >> ((9 - size) & 7);
   }  

   /* get the maskOR_lsb */
   maskOR_lsb         = 1;
   if (flags & LTM_PRIME_BBS) {
      maskOR_lsb     |= 3;
   }

   do {
      /* read the bytes */
      if (cb(tmp, bsize, dat) != bsize) {
         err = MP_VAL;
         goto error;
      }
 
      /* work over the MSbyte */
      tmp[0]    &= maskAND;
      tmp[0]    |= 1 << ((size - 1) & 7);

      /* mix in the maskORs */
      tmp[maskOR_msb_offset]   |= maskOR_msb;
      tmp[bsize-1]             |= maskOR_lsb;

      /* read it in */
      if ((err = mp_read_unsigned_bin(a, tmp, bsize)) != MP_OKAY)     { goto error; }

      /* is it prime? */
      if ((err = mp_prime_is_prime(a, t, &res)) != MP_OKAY)           { goto error; }
      if (res == MP_NO) {  
         continue;
      }

      if (flags & LTM_PRIME_SAFE) {
         /* see if (a-1)/2 is prime */
         if ((err = mp_sub_d(a, 1, a)) != MP_OKAY)                    { goto error; }
         if ((err = mp_div_2(a, a)) != MP_OKAY)                       { goto error; }
 
         /* is it prime? */
         if ((err = mp_prime_is_prime(a, t, &res)) != MP_OKAY)        { goto error; }
      }
   } while (res == MP_NO);

   if (flags & LTM_PRIME_SAFE) {
      /* restore a to the original value */
      if ((err = mp_mul_2(a, a)) != MP_OKAY)                          { goto error; }
      if ((err = mp_add_d(a, 1, a)) != MP_OKAY)                       { goto error; }
   }

   err = MP_OKAY;
error:
   XFREE(tmp);
   return err;
}


#endif

/* End: bn_mp_prime_random_ex.c */

/* Start: bn_mp_radix_size.c */
#include <tommath.h>
#ifdef BN_MP_RADIX_SIZE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* returns size of ASCII reprensentation */
int mp_radix_size (mp_int * a, int radix, int *size)
{
  int     res, digs;
  mp_int  t;
  mp_digit d;

  *size = 0;

  /* special case for binary */
  if (radix == 2) {
    *size = mp_count_bits (a) + (a->sign == MP_NEG ? 1 : 0) + 1;
    return MP_OKAY;
  }

  /* make sure the radix is in range */
  if (radix < 2 || radix > 64) {
    return MP_VAL;
  }

  if (mp_iszero(a) == MP_YES) {
    *size = 2;
    return MP_OKAY;
  }

  /* digs is the digit count */
  digs = 0;

  /* if it's negative add one for the sign */
  if (a->sign == MP_NEG) {
    ++digs;
  }

  /* init a copy of the input */
  if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
    return res;
  }

  /* force temp to positive */
  t.sign = MP_ZPOS; 

  /* fetch out all of the digits */
  while (mp_iszero (&t) == MP_NO) {
    if ((res = mp_div_d (&t, (mp_digit) radix, &t, &d)) != MP_OKAY) {
      mp_clear (&t);
      return res;
    }
    ++digs;
  }
  mp_clear (&t);

  /* return digs + 1, the 1 is for the NULL byte that would be required. */
  *size = digs + 1;
  return MP_OKAY;
}

#endif

/* End: bn_mp_radix_size.c */

/* Start: bn_mp_radix_smap.c */
#include <tommath.h>
#ifdef BN_MP_RADIX_SMAP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* chars used in radix conversions */
const char *mp_s_rmap = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz+/";
#endif

/* End: bn_mp_radix_smap.c */

/* Start: bn_mp_rand.c */
#include <tommath.h>
#ifdef BN_MP_RAND_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* makes a pseudo-random int of a given size */
int
mp_rand (mp_int * a, int digits)
{
  int     res;
  mp_digit d;

  mp_zero (a);
  if (digits <= 0) {
    return MP_OKAY;
  }

  /* first place a random non-zero digit */
  do {
    d = ((mp_digit) abs (rand ())) & MP_MASK;
  } while (d == 0);

  if ((res = mp_add_d (a, d, a)) != MP_OKAY) {
    return res;
  }

  while (--digits > 0) {
    if ((res = mp_lshd (a, 1)) != MP_OKAY) {
      return res;
    }

    if ((res = mp_add_d (a, ((mp_digit) abs (rand ())), a)) != MP_OKAY) {
      return res;
    }
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_rand.c */

/* Start: bn_mp_read_radix.c */
#include <tommath.h>
#ifdef BN_MP_READ_RADIX_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* read a string [ASCII] in a given radix */
int mp_read_radix (mp_int * a, const char *str, int radix)
{
  int     y, res, neg;
  char    ch;

  /* zero the digit bignum */
  mp_zero(a);

  /* make sure the radix is ok */
  if (radix < 2 || radix > 64) {
    return MP_VAL;
  }

  /* if the leading digit is a 
   * minus set the sign to negative. 
   */
  if (*str == '-') {
    ++str;
    neg = MP_NEG;
  } else {
    neg = MP_ZPOS;
  }

  /* set the integer to the default of zero */
  mp_zero (a);
  
  /* process each digit of the string */
  while (*str) {
    /* if the radix < 36 the conversion is case insensitive
     * this allows numbers like 1AB and 1ab to represent the same  value
     * [e.g. in hex]
     */
    ch = (char) ((radix < 36) ? toupper (*str) : *str);
    for (y = 0; y < 64; y++) {
      if (ch == mp_s_rmap[y]) {
         break;
      }
    }

    /* if the char was found in the map 
     * and is less than the given radix add it
     * to the number, otherwise exit the loop. 
     */
    if (y < radix) {
      if ((res = mp_mul_d (a, (mp_digit) radix, a)) != MP_OKAY) {
         return res;
      }
      if ((res = mp_add_d (a, (mp_digit) y, a)) != MP_OKAY) {
         return res;
      }
    } else {
      break;
    }
    ++str;
  }
  
  /* set the sign only if a != 0 */
  if (mp_iszero(a) != 1) {
     a->sign = neg;
  }
  return MP_OKAY;
}
#endif

/* End: bn_mp_read_radix.c */

/* Start: bn_mp_read_signed_bin.c */
#include <tommath.h>
#ifdef BN_MP_READ_SIGNED_BIN_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* read signed bin, big endian, first byte is 0==positive or 1==negative */
int mp_read_signed_bin (mp_int * a, const unsigned char *b, int c)
{
  int     res;

  /* read magnitude */
  if ((res = mp_read_unsigned_bin (a, b + 1, c - 1)) != MP_OKAY) {
    return res;
  }

  /* first byte is 0 for positive, non-zero for negative */
  if (b[0] == 0) {
     a->sign = MP_ZPOS;
  } else {
     a->sign = MP_NEG;
  }

  return MP_OKAY;
}
#endif

/* End: bn_mp_read_signed_bin.c */

/* Start: bn_mp_read_unsigned_bin.c */
#include <tommath.h>
#ifdef BN_MP_READ_UNSIGNED_BIN_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reads a unsigned char array, assumes the msb is stored first [big endian] */
int mp_read_unsigned_bin (mp_int * a, const unsigned char *b, int c)
{
  int     res;

  /* make sure there are at least two digits */
  if (a->alloc < 2) {
     if ((res = mp_grow(a, 2)) != MP_OKAY) {
        return res;
     }
  }

  /* zero the int */
  mp_zero (a);

  /* read the bytes in */
  while (c-- > 0) {
    if ((res = mp_mul_2d (a, 8, a)) != MP_OKAY) {
      return res;
    }

#ifndef MP_8BIT
      a->dp[0] |= *b++;
      a->used += 1;
#else
      a->dp[0] = (*b & MP_MASK);
      a->dp[1] |= ((*b++ >> 7U) & 1);
      a->used += 2;
#endif
  }
  mp_clamp (a);
  return MP_OKAY;
}
#endif

/* End: bn_mp_read_unsigned_bin.c */

/* Start: bn_mp_reduce.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reduces x mod m, assumes 0 < x < m**2, mu is 
 * precomputed via mp_reduce_setup.
 * From HAC pp.604 Algorithm 14.42
 */
int mp_reduce (mp_int * x, mp_int * m, mp_int * mu)
{
  mp_int  q;
  int     res, um = m->used;

  /* q = x */
  if ((res = mp_init_copy (&q, x)) != MP_OKAY) {
    return res;
  }

  /* q1 = x / b**(k-1)  */
  mp_rshd (&q, um - 1);         

  /* according to HAC this optimization is ok */
  if (((unsigned long) um) > (((mp_digit)1) << (DIGIT_BIT - 1))) {
    if ((res = mp_mul (&q, mu, &q)) != MP_OKAY) {
      goto CLEANUP;
    }
  } else {
#ifdef BN_S_MP_MUL_HIGH_DIGS_C
    if ((res = s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) {
      goto CLEANUP;
    }
#elif defined(BN_FAST_S_MP_MUL_HIGH_DIGS_C)
    if ((res = fast_s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) {
      goto CLEANUP;
    }
#else 
    { 
      res = MP_VAL;
      goto CLEANUP;
    }
#endif
  }

  /* q3 = q2 / b**(k+1) */
  mp_rshd (&q, um + 1);         

  /* x = x mod b**(k+1), quick (no division) */
  if ((res = mp_mod_2d (x, DIGIT_BIT * (um + 1), x)) != MP_OKAY) {
    goto CLEANUP;
  }

  /* q = q * m mod b**(k+1), quick (no division) */
  if ((res = s_mp_mul_digs (&q, m, &q, um + 1)) != MP_OKAY) {
    goto CLEANUP;
  }

  /* x = x - q */
  if ((res = mp_sub (x, &q, x)) != MP_OKAY) {
    goto CLEANUP;
  }

  /* If x < 0, add b**(k+1) to it */
  if (mp_cmp_d (x, 0) == MP_LT) {
    mp_set (&q, 1);
    if ((res = mp_lshd (&q, um + 1)) != MP_OKAY)
      goto CLEANUP;
    if ((res = mp_add (x, &q, x)) != MP_OKAY)
      goto CLEANUP;
  }

  /* Back off if it's too big */
  while (mp_cmp (x, m) != MP_LT) {
    if ((res = s_mp_sub (x, m, x)) != MP_OKAY) {
      goto CLEANUP;
    }
  }
  
CLEANUP:
  mp_clear (&q);

  return res;
}
#endif

/* End: bn_mp_reduce.c */

/* Start: bn_mp_reduce_2k.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_2K_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reduces a modulo n where n is of the form 2**p - d */
int mp_reduce_2k(mp_int *a, mp_int *n, mp_digit d)
{
   mp_int q;
   int    p, res;
   
   if ((res = mp_init(&q)) != MP_OKAY) {
      return res;
   }
   
   p = mp_count_bits(n);    
top:
   /* q = a/2**p, a = a mod 2**p */
   if ((res = mp_div_2d(a, p, &q, a)) != MP_OKAY) {
      goto ERR;
   }
   
   if (d != 1) {
      /* q = q * d */
      if ((res = mp_mul_d(&q, d, &q)) != MP_OKAY) { 
         goto ERR;
      }
   }
   
   /* a = a + q */
   if ((res = s_mp_add(a, &q, a)) != MP_OKAY) {
      goto ERR;
   }
   
   if (mp_cmp_mag(a, n) != MP_LT) {
      s_mp_sub(a, n, a);
      goto top;
   }
   
ERR:
   mp_clear(&q);
   return res;
}

#endif

/* End: bn_mp_reduce_2k.c */

/* Start: bn_mp_reduce_2k_l.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_2K_L_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reduces a modulo n where n is of the form 2**p - d 
   This differs from reduce_2k since "d" can be larger
   than a single digit.
*/
int mp_reduce_2k_l(mp_int *a, mp_int *n, mp_int *d)
{
   mp_int q;
   int    p, res;
   
   if ((res = mp_init(&q)) != MP_OKAY) {
      return res;
   }
   
   p = mp_count_bits(n);    
top:
   /* q = a/2**p, a = a mod 2**p */
   if ((res = mp_div_2d(a, p, &q, a)) != MP_OKAY) {
      goto ERR;
   }
   
   /* q = q * d */
   if ((res = mp_mul(&q, d, &q)) != MP_OKAY) { 
      goto ERR;
   }
   
   /* a = a + q */
   if ((res = s_mp_add(a, &q, a)) != MP_OKAY) {
      goto ERR;
   }
   
   if (mp_cmp_mag(a, n) != MP_LT) {
      s_mp_sub(a, n, a);
      goto top;
   }
   
ERR:
   mp_clear(&q);
   return res;
}

#endif

/* End: bn_mp_reduce_2k_l.c */

/* Start: bn_mp_reduce_2k_setup.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_2K_SETUP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines the setup value */
int mp_reduce_2k_setup(mp_int *a, mp_digit *d)
{
   int res, p;
   mp_int tmp;
   
   if ((res = mp_init(&tmp)) != MP_OKAY) {
      return res;
   }
   
   p = mp_count_bits(a);
   if ((res = mp_2expt(&tmp, p)) != MP_OKAY) {
      mp_clear(&tmp);
      return res;
   }
   
   if ((res = s_mp_sub(&tmp, a, &tmp)) != MP_OKAY) {
      mp_clear(&tmp);
      return res;
   }
   
   *d = tmp.dp[0];
   mp_clear(&tmp);
   return MP_OKAY;
}
#endif

/* End: bn_mp_reduce_2k_setup.c */

/* Start: bn_mp_reduce_2k_setup_l.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_2K_SETUP_L_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines the setup value */
int mp_reduce_2k_setup_l(mp_int *a, mp_int *d)
{
   int    res;
   mp_int tmp;
   
   if ((res = mp_init(&tmp)) != MP_OKAY) {
      return res;
   }
   
   if ((res = mp_2expt(&tmp, mp_count_bits(a))) != MP_OKAY) {
      goto ERR;
   }
   
   if ((res = s_mp_sub(&tmp, a, d)) != MP_OKAY) {
      goto ERR;
   }
   
ERR:
   mp_clear(&tmp);
   return res;
}
#endif

/* End: bn_mp_reduce_2k_setup_l.c */

/* Start: bn_mp_reduce_is_2k.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_IS_2K_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines if mp_reduce_2k can be used */
int mp_reduce_is_2k(mp_int *a)
{
   int ix, iy, iw;
   mp_digit iz;
   
   if (a->used == 0) {
      return MP_NO;
   } else if (a->used == 1) {
      return MP_YES;
   } else if (a->used > 1) {
      iy = mp_count_bits(a);
      iz = 1;
      iw = 1;
    
      /* Test every bit from the second digit up, must be 1 */
      for (ix = DIGIT_BIT; ix < iy; ix++) {
          if ((a->dp[iw] & iz) == 0) {
             return MP_NO;
          }
          iz <<= 1;
          if (iz > (mp_digit)MP_MASK) {
             ++iw;
             iz = 1;
          }
      }
   }
   return MP_YES;
}

#endif

/* End: bn_mp_reduce_is_2k.c */

/* Start: bn_mp_reduce_is_2k_l.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_IS_2K_L_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* determines if reduce_2k_l can be used */
int mp_reduce_is_2k_l(mp_int *a)
{
   int ix, iy;
   
   if (a->used == 0) {
      return MP_NO;
   } else if (a->used == 1) {
      return MP_YES;
   } else if (a->used > 1) {
      /* if more than half of the digits are -1 we're sold */
      for (iy = ix = 0; ix < a->used; ix++) {
          if (a->dp[ix] == MP_MASK) {
              ++iy;
          }
      }
      return (iy >= (a->used/2)) ? MP_YES : MP_NO;
      
   }
   return MP_NO;
}

#endif

/* End: bn_mp_reduce_is_2k_l.c */

/* Start: bn_mp_reduce_setup.c */
#include <tommath.h>
#ifdef BN_MP_REDUCE_SETUP_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* pre-calculate the value required for Barrett reduction
 * For a given modulus "b" it calulates the value required in "a"
 */
int mp_reduce_setup (mp_int * a, mp_int * b)
{
  int     res;
  
  if ((res = mp_2expt (a, b->used * 2 * DIGIT_BIT)) != MP_OKAY) {
    return res;
  }
  return mp_div (a, b, a, NULL);
}
#endif

/* End: bn_mp_reduce_setup.c */

/* Start: bn_mp_rshd.c */
#include <tommath.h>
#ifdef BN_MP_RSHD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* shift right a certain amount of digits */
void mp_rshd (mp_int * a, int b)
{
  int     x;

  /* if b <= 0 then ignore it */
  if (b <= 0) {
    return;
  }

  /* if b > used then simply zero it and return */
  if (a->used <= b) {
    mp_zero (a);
    return;
  }

  {
    register mp_digit *bottom, *top;

    /* shift the digits down */

    /* bottom */
    bottom = a->dp;

    /* top [offset into digits] */
    top = a->dp + b;

    /* this is implemented as a sliding window where 
     * the window is b-digits long and digits from 
     * the top of the window are copied to the bottom
     *
     * e.g.

     b-2 | b-1 | b0 | b1 | b2 | ... | bb |   ---->
                 /\                   |      ---->
                  \-------------------/      ---->
     */
    for (x = 0; x < (a->used - b); x++) {
      *bottom++ = *top++;
    }

    /* zero the top digits */
    for (; x < a->used; x++) {
      *bottom++ = 0;
    }
  }
  
  /* remove excess digits */
  a->used -= b;
}
#endif

/* End: bn_mp_rshd.c */

/* Start: bn_mp_set.c */
#include <tommath.h>
#ifdef BN_MP_SET_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* set to a digit */
void mp_set (mp_int * a, mp_digit b)
{
  mp_zero (a);
  a->dp[0] = b & MP_MASK;
  a->used  = (a->dp[0] != 0) ? 1 : 0;
}
#endif

/* End: bn_mp_set.c */

/* Start: bn_mp_set_int.c */
#include <tommath.h>
#ifdef BN_MP_SET_INT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* set a 32-bit const */
int mp_set_int (mp_int * a, unsigned long b)
{
  int     x, res;

  mp_zero (a);
  
  /* set four bits at a time */
  for (x = 0; x < 8; x++) {
    /* shift the number up four bits */
    if ((res = mp_mul_2d (a, 4, a)) != MP_OKAY) {
      return res;
    }

    /* OR in the top four bits of the source */
    a->dp[0] |= (b >> 28) & 15;

    /* shift the source up to the next four bits */
    b <<= 4;

    /* ensure that digits are not clamped off */
    a->used += 1;
  }
  mp_clamp (a);
  return MP_OKAY;
}
#endif

/* End: bn_mp_set_int.c */

/* Start: bn_mp_shrink.c */
#include <tommath.h>
#ifdef BN_MP_SHRINK_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* shrink a bignum */
int mp_shrink (mp_int * a)
{
  mp_digit *tmp;
  int used = 1;
  
  if(a->used > 0)
    used = a->used;
  
  if (a->alloc != used) {
    if ((tmp = OPT_CAST(mp_digit) XREALLOC (a->dp, sizeof (mp_digit) * used)) == NULL) {
      return MP_MEM;
    }
    a->dp    = tmp;
    a->alloc = used;
  }
  return MP_OKAY;
}
#endif

/* End: bn_mp_shrink.c */

/* Start: bn_mp_signed_bin_size.c */
#include <tommath.h>
#ifdef BN_MP_SIGNED_BIN_SIZE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* get the size for an signed equivalent */
int mp_signed_bin_size (mp_int * a)
{
  return 1 + mp_unsigned_bin_size (a);
}
#endif

/* End: bn_mp_signed_bin_size.c */

/* Start: bn_mp_sqr.c */
#include <tommath.h>
#ifdef BN_MP_SQR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* computes b = a*a */
int
mp_sqr (mp_int * a, mp_int * b)
{
  int     res;

#ifdef BN_MP_TOOM_SQR_C
  /* use Toom-Cook? */
  if (a->used >= TOOM_SQR_CUTOFF) {
    res = mp_toom_sqr(a, b);
  /* Karatsuba? */
  } else 
#endif
#ifdef BN_MP_KARATSUBA_SQR_C
if (a->used >= KARATSUBA_SQR_CUTOFF) {
    res = mp_karatsuba_sqr (a, b);
  } else 
#endif
  {
#ifdef BN_FAST_S_MP_SQR_C
    /* can we use the fast comba multiplier? */
    if ((a->used * 2 + 1) < MP_WARRAY && 
         a->used < 
         (1 << (sizeof(mp_word) * CHAR_BIT - 2*DIGIT_BIT - 1))) {
      res = fast_s_mp_sqr (a, b);
    } else
#endif
#ifdef BN_S_MP_SQR_C
      res = s_mp_sqr (a, b);
#else
      res = MP_VAL;
#endif
  }
  b->sign = MP_ZPOS;
  return res;
}
#endif

/* End: bn_mp_sqr.c */

/* Start: bn_mp_sqrmod.c */
#include <tommath.h>
#ifdef BN_MP_SQRMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* c = a * a (mod b) */
int
mp_sqrmod (mp_int * a, mp_int * b, mp_int * c)
{
  int     res;
  mp_int  t;

  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_sqr (a, &t)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }
  res = mp_mod (&t, b, c);
  mp_clear (&t);
  return res;
}
#endif

/* End: bn_mp_sqrmod.c */

/* Start: bn_mp_sqrt.c */
#include <tommath.h>

#ifdef BN_MP_SQRT_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* this function is less generic than mp_n_root, simpler and faster */
int mp_sqrt(mp_int *arg, mp_int *ret) 
{
  int res;
  mp_int t1,t2;

  /* must be positive */
  if (arg->sign == MP_NEG) {
    return MP_VAL;
  }

  /* easy out */
  if (mp_iszero(arg) == MP_YES) {
    mp_zero(ret);
    return MP_OKAY;
  }

  if ((res = mp_init_copy(&t1, arg)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_init(&t2)) != MP_OKAY) {
    goto E2;
  }

  /* First approx. (not very bad for large arg) */
  mp_rshd (&t1,t1.used/2);

  /* t1 > 0  */ 
  if ((res = mp_div(arg,&t1,&t2,NULL)) != MP_OKAY) {
    goto E1;
  }
  if ((res = mp_add(&t1,&t2,&t1)) != MP_OKAY) {
    goto E1;
  }
  if ((res = mp_div_2(&t1,&t1)) != MP_OKAY) {
    goto E1;
  }
  /* And now t1 > sqrt(arg) */
  do { 
    if ((res = mp_div(arg,&t1,&t2,NULL)) != MP_OKAY) {
      goto E1;
    }
    if ((res = mp_add(&t1,&t2,&t1)) != MP_OKAY) {
      goto E1;
    }
    if ((res = mp_div_2(&t1,&t1)) != MP_OKAY) {
      goto E1;
    }
    /* t1 >= sqrt(arg) >= t2 at this point */
  } while (mp_cmp_mag(&t1,&t2) == MP_GT);

  mp_exch(&t1,ret);

E1: mp_clear(&t2);
E2: mp_clear(&t1);
  return res;
}

#endif

/* End: bn_mp_sqrt.c */

/* Start: bn_mp_sub.c */
#include <tommath.h>
#ifdef BN_MP_SUB_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* high level subtraction (handles signs) */
int
mp_sub (mp_int * a, mp_int * b, mp_int * c)
{
  int     sa, sb, res;

  sa = a->sign;
  sb = b->sign;

  if (sa != sb) {
    /* subtract a negative from a positive, OR */
    /* subtract a positive from a negative. */
    /* In either case, ADD their magnitudes, */
    /* and use the sign of the first number. */
    c->sign = sa;
    res = s_mp_add (a, b, c);
  } else {
    /* subtract a positive from a positive, OR */
    /* subtract a negative from a negative. */
    /* First, take the difference between their */
    /* magnitudes, then... */
    if (mp_cmp_mag (a, b) != MP_LT) {
      /* Copy the sign from the first */
      c->sign = sa;
      /* The first has a larger or equal magnitude */
      res = s_mp_sub (a, b, c);
    } else {
      /* The result has the *opposite* sign from */
      /* the first number. */
      c->sign = (sa == MP_ZPOS) ? MP_NEG : MP_ZPOS;
      /* The second has a larger magnitude */
      res = s_mp_sub (b, a, c);
    }
  }
  return res;
}

#endif

/* End: bn_mp_sub.c */

/* Start: bn_mp_sub_d.c */
#include <tommath.h>
#ifdef BN_MP_SUB_D_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* single digit subtraction */
int
mp_sub_d (mp_int * a, mp_digit b, mp_int * c)
{
  mp_digit *tmpa, *tmpc, mu;
  int       res, ix, oldused;

  /* grow c as required */
  if (c->alloc < a->used + 1) {
     if ((res = mp_grow(c, a->used + 1)) != MP_OKAY) {
        return res;
     }
  }

  /* if a is negative just do an unsigned
   * addition [with fudged signs]
   */
  if (a->sign == MP_NEG) {
     a->sign = MP_ZPOS;
     res     = mp_add_d(a, b, c);
     a->sign = c->sign = MP_NEG;

     /* clamp */
     mp_clamp(c);

     return res;
  }

  /* setup regs */
  oldused = c->used;
  tmpa    = a->dp;
  tmpc    = c->dp;

  /* if a <= b simply fix the single digit */
  if ((a->used == 1 && a->dp[0] <= b) || a->used == 0) {
     if (a->used == 1) {
        *tmpc++ = b - *tmpa;
     } else {
        *tmpc++ = b;
     }
     ix      = 1;

     /* negative/1digit */
     c->sign = MP_NEG;
     c->used = 1;
  } else {
     /* positive/size */
     c->sign = MP_ZPOS;
     c->used = a->used;

     /* subtract first digit */
     *tmpc    = *tmpa++ - b;
     mu       = *tmpc >> (sizeof(mp_digit) * CHAR_BIT - 1);
     *tmpc++ &= MP_MASK;

     /* handle rest of the digits */
     for (ix = 1; ix < a->used; ix++) {
        *tmpc    = *tmpa++ - mu;
        mu       = *tmpc >> (sizeof(mp_digit) * CHAR_BIT - 1);
        *tmpc++ &= MP_MASK;
     }
  }

  /* zero excess digits */
  while (ix++ < oldused) {
     *tmpc++ = 0;
  }
  mp_clamp(c);
  return MP_OKAY;
}

#endif

/* End: bn_mp_sub_d.c */

/* Start: bn_mp_submod.c */
#include <tommath.h>
#ifdef BN_MP_SUBMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* d = a - b (mod c) */
int
mp_submod (mp_int * a, mp_int * b, mp_int * c, mp_int * d)
{
  int     res;
  mp_int  t;


  if ((res = mp_init (&t)) != MP_OKAY) {
    return res;
  }

  if ((res = mp_sub (a, b, &t)) != MP_OKAY) {
    mp_clear (&t);
    return res;
  }
  res = mp_mod (&t, c, d);
  mp_clear (&t);
  return res;
}
#endif

/* End: bn_mp_submod.c */

/* Start: bn_mp_to_signed_bin.c */
#include <tommath.h>
#ifdef BN_MP_TO_SIGNED_BIN_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* store in signed [big endian] format */
int mp_to_signed_bin (mp_int * a, unsigned char *b)
{
  int     res;

  if ((res = mp_to_unsigned_bin (a, b + 1)) != MP_OKAY) {
    return res;
  }
  b[0] = (unsigned char) ((a->sign == MP_ZPOS) ? 0 : 1);
  return MP_OKAY;
}
#endif

/* End: bn_mp_to_signed_bin.c */

/* Start: bn_mp_to_signed_bin_n.c */
#include <tommath.h>
#ifdef BN_MP_TO_SIGNED_BIN_N_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* store in signed [big endian] format */
int mp_to_signed_bin_n (mp_int * a, unsigned char *b, unsigned long *outlen)
{
   if (*outlen < (unsigned long)mp_signed_bin_size(a)) {
      return MP_VAL;
   }
   *outlen = mp_signed_bin_size(a);
   return mp_to_signed_bin(a, b);
}
#endif

/* End: bn_mp_to_signed_bin_n.c */

/* Start: bn_mp_to_unsigned_bin.c */
#include <tommath.h>
#ifdef BN_MP_TO_UNSIGNED_BIN_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* store in unsigned [big endian] format */
int mp_to_unsigned_bin (mp_int * a, unsigned char *b)
{
  int     x, res;
  mp_int  t;

  if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
    return res;
  }

  x = 0;
  while (mp_iszero (&t) == 0) {
#ifndef MP_8BIT
      b[x++] = (unsigned char) (t.dp[0] & 255);
#else
      b[x++] = (unsigned char) (t.dp[0] | ((t.dp[1] & 0x01) << 7));
#endif
    if ((res = mp_div_2d (&t, 8, &t, NULL)) != MP_OKAY) {
      mp_clear (&t);
      return res;
    }
  }
  bn_reverse (b, x);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_mp_to_unsigned_bin.c */

/* Start: bn_mp_to_unsigned_bin_n.c */
#include <tommath.h>
#ifdef BN_MP_TO_UNSIGNED_BIN_N_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* store in unsigned [big endian] format */
int mp_to_unsigned_bin_n (mp_int * a, unsigned char *b, unsigned long *outlen)
{
   if (*outlen < (unsigned long)mp_unsigned_bin_size(a)) {
      return MP_VAL;
   }
   *outlen = mp_unsigned_bin_size(a);
   return mp_to_unsigned_bin(a, b);
}
#endif

/* End: bn_mp_to_unsigned_bin_n.c */

/* Start: bn_mp_toom_mul.c */
#include <tommath.h>
#ifdef BN_MP_TOOM_MUL_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* multiplication using the Toom-Cook 3-way algorithm 
 *
 * Much more complicated than Karatsuba but has a lower 
 * asymptotic running time of O(N**1.464).  This algorithm is 
 * only particularly useful on VERY large inputs 
 * (we're talking 1000s of digits here...).
*/
int mp_toom_mul(mp_int *a, mp_int *b, mp_int *c)
{
    mp_int w0, w1, w2, w3, w4, tmp1, tmp2, a0, a1, a2, b0, b1, b2;
    int res, B;
        
    /* init temps */
    if ((res = mp_init_multi(&w0, &w1, &w2, &w3, &w4, 
                             &a0, &a1, &a2, &b0, &b1, 
                             &b2, &tmp1, &tmp2, NULL)) != MP_OKAY) {
       return res;
    }
    
    /* B */
    B = MIN(a->used, b->used) / 3;
    
    /* a = a2 * B**2 + a1 * B + a0 */
    if ((res = mp_mod_2d(a, DIGIT_BIT * B, &a0)) != MP_OKAY) {
       goto ERR;
    }

    if ((res = mp_copy(a, &a1)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&a1, B);
    mp_mod_2d(&a1, DIGIT_BIT * B, &a1);

    if ((res = mp_copy(a, &a2)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&a2, B*2);
    
    /* b = b2 * B**2 + b1 * B + b0 */
    if ((res = mp_mod_2d(b, DIGIT_BIT * B, &b0)) != MP_OKAY) {
       goto ERR;
    }

    if ((res = mp_copy(b, &b1)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&b1, B);
    mp_mod_2d(&b1, DIGIT_BIT * B, &b1);

    if ((res = mp_copy(b, &b2)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&b2, B*2);
    
    /* w0 = a0*b0 */
    if ((res = mp_mul(&a0, &b0, &w0)) != MP_OKAY) {
       goto ERR;
    }
    
    /* w4 = a2 * b2 */
    if ((res = mp_mul(&a2, &b2, &w4)) != MP_OKAY) {
       goto ERR;
    }
    
    /* w1 = (a2 + 2(a1 + 2a0))(b2 + 2(b1 + 2b0)) */
    if ((res = mp_mul_2(&a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a2, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    
    if ((res = mp_mul_2(&b0, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp2, &b1, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp2, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp2, &b2, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    
    if ((res = mp_mul(&tmp1, &tmp2, &w1)) != MP_OKAY) {
       goto ERR;
    }
    
    /* w3 = (a0 + 2(a1 + 2a2))(b0 + 2(b1 + 2b2)) */
    if ((res = mp_mul_2(&a2, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    
    if ((res = mp_mul_2(&b2, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp2, &b1, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp2, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp2, &b0, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    
    if ((res = mp_mul(&tmp1, &tmp2, &w3)) != MP_OKAY) {
       goto ERR;
    }
    

    /* w2 = (a2 + a1 + a0)(b2 + b1 + b0) */
    if ((res = mp_add(&a2, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&b2, &b1, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp2, &b0, &tmp2)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul(&tmp1, &tmp2, &w2)) != MP_OKAY) {
       goto ERR;
    }
    
    /* now solve the matrix 
    
       0  0  0  0  1
       1  2  4  8  16
       1  1  1  1  1
       16 8  4  2  1
       1  0  0  0  0
       
       using 12 subtractions, 4 shifts, 
              2 small divisions and 1 small multiplication 
     */
     
     /* r1 - r4 */
     if ((res = mp_sub(&w1, &w4, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r0 */
     if ((res = mp_sub(&w3, &w0, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1/2 */
     if ((res = mp_div_2(&w1, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3/2 */
     if ((res = mp_div_2(&w3, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r2 - r0 - r4 */
     if ((res = mp_sub(&w2, &w0, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w4, &w2)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - r2 */
     if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r2 */
     if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - 8r0 */
     if ((res = mp_mul_2d(&w0, 3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w1, &tmp1, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - 8r4 */
     if ((res = mp_mul_2d(&w4, 3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w3, &tmp1, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* 3r2 - r1 - r3 */
     if ((res = mp_mul_d(&w2, 3, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w1, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w3, &w2)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - r2 */
     if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r2 */
     if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1/3 */
     if ((res = mp_div_3(&w1, &w1, NULL)) != MP_OKAY) {
        goto ERR;
     }
     /* r3/3 */
     if ((res = mp_div_3(&w3, &w3, NULL)) != MP_OKAY) {
        goto ERR;
     }
     
     /* at this point shift W[n] by B*n */
     if ((res = mp_lshd(&w1, 1*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w2, 2*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w3, 3*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w4, 4*B)) != MP_OKAY) {
        goto ERR;
     }     
     
     if ((res = mp_add(&w0, &w1, c)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&w2, &w3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&w4, &tmp1, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&tmp1, c, c)) != MP_OKAY) {
        goto ERR;
     }     
     
ERR:
     mp_clear_multi(&w0, &w1, &w2, &w3, &w4, 
                    &a0, &a1, &a2, &b0, &b1, 
                    &b2, &tmp1, &tmp2, NULL);
     return res;
}     
     
#endif

/* End: bn_mp_toom_mul.c */

/* Start: bn_mp_toom_sqr.c */
#include <tommath.h>
#ifdef BN_MP_TOOM_SQR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* squaring using Toom-Cook 3-way algorithm */
int
mp_toom_sqr(mp_int *a, mp_int *b)
{
    mp_int w0, w1, w2, w3, w4, tmp1, a0, a1, a2;
    int res, B;

    /* init temps */
    if ((res = mp_init_multi(&w0, &w1, &w2, &w3, &w4, &a0, &a1, &a2, &tmp1, NULL)) != MP_OKAY) {
       return res;
    }

    /* B */
    B = a->used / 3;

    /* a = a2 * B**2 + a1 * B + a0 */
    if ((res = mp_mod_2d(a, DIGIT_BIT * B, &a0)) != MP_OKAY) {
       goto ERR;
    }

    if ((res = mp_copy(a, &a1)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&a1, B);
    mp_mod_2d(&a1, DIGIT_BIT * B, &a1);

    if ((res = mp_copy(a, &a2)) != MP_OKAY) {
       goto ERR;
    }
    mp_rshd(&a2, B*2);

    /* w0 = a0*a0 */
    if ((res = mp_sqr(&a0, &w0)) != MP_OKAY) {
       goto ERR;
    }

    /* w4 = a2 * a2 */
    if ((res = mp_sqr(&a2, &w4)) != MP_OKAY) {
       goto ERR;
    }

    /* w1 = (a2 + 2(a1 + 2a0))**2 */
    if ((res = mp_mul_2(&a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a2, &tmp1)) != MP_OKAY) {
       goto ERR;
    }

    if ((res = mp_sqr(&tmp1, &w1)) != MP_OKAY) {
       goto ERR;
    }

    /* w3 = (a0 + 2(a1 + 2a2))**2 */
    if ((res = mp_mul_2(&a2, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_mul_2(&tmp1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }

    if ((res = mp_sqr(&tmp1, &w3)) != MP_OKAY) {
       goto ERR;
    }


    /* w2 = (a2 + a1 + a0)**2 */
    if ((res = mp_add(&a2, &a1, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_add(&tmp1, &a0, &tmp1)) != MP_OKAY) {
       goto ERR;
    }
    if ((res = mp_sqr(&tmp1, &w2)) != MP_OKAY) {
       goto ERR;
    }

    /* now solve the matrix

       0  0  0  0  1
       1  2  4  8  16
       1  1  1  1  1
       16 8  4  2  1
       1  0  0  0  0

       using 12 subtractions, 4 shifts, 2 small divisions and 1 small multiplication.
     */

     /* r1 - r4 */
     if ((res = mp_sub(&w1, &w4, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r0 */
     if ((res = mp_sub(&w3, &w0, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1/2 */
     if ((res = mp_div_2(&w1, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3/2 */
     if ((res = mp_div_2(&w3, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r2 - r0 - r4 */
     if ((res = mp_sub(&w2, &w0, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w4, &w2)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - r2 */
     if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r2 */
     if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - 8r0 */
     if ((res = mp_mul_2d(&w0, 3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w1, &tmp1, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - 8r4 */
     if ((res = mp_mul_2d(&w4, 3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w3, &tmp1, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* 3r2 - r1 - r3 */
     if ((res = mp_mul_d(&w2, 3, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w1, &w2)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_sub(&w2, &w3, &w2)) != MP_OKAY) {
        goto ERR;
     }
     /* r1 - r2 */
     if ((res = mp_sub(&w1, &w2, &w1)) != MP_OKAY) {
        goto ERR;
     }
     /* r3 - r2 */
     if ((res = mp_sub(&w3, &w2, &w3)) != MP_OKAY) {
        goto ERR;
     }
     /* r1/3 */
     if ((res = mp_div_3(&w1, &w1, NULL)) != MP_OKAY) {
        goto ERR;
     }
     /* r3/3 */
     if ((res = mp_div_3(&w3, &w3, NULL)) != MP_OKAY) {
        goto ERR;
     }

     /* at this point shift W[n] by B*n */
     if ((res = mp_lshd(&w1, 1*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w2, 2*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w3, 3*B)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_lshd(&w4, 4*B)) != MP_OKAY) {
        goto ERR;
     }

     if ((res = mp_add(&w0, &w1, b)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&w2, &w3, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&w4, &tmp1, &tmp1)) != MP_OKAY) {
        goto ERR;
     }
     if ((res = mp_add(&tmp1, b, b)) != MP_OKAY) {
        goto ERR;
     }

ERR:
     mp_clear_multi(&w0, &w1, &w2, &w3, &w4, &a0, &a1, &a2, &tmp1, NULL);
     return res;
}

#endif

/* End: bn_mp_toom_sqr.c */

/* Start: bn_mp_toradix.c */
#include <tommath.h>
#ifdef BN_MP_TORADIX_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* stores a bignum as a ASCII string in a given radix (2..64) */
int mp_toradix (mp_int * a, char *str, int radix)
{
  int     res, digs;
  mp_int  t;
  mp_digit d;
  char   *_s = str;

  /* check range of the radix */
  if (radix < 2 || radix > 64) {
    return MP_VAL;
  }

  /* quick out if its zero */
  if (mp_iszero(a) == 1) {
     *str++ = '0';
     *str = '\0';
     return MP_OKAY;
  }

  if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
    return res;
  }

  /* if it is negative output a - */
  if (t.sign == MP_NEG) {
    ++_s;
    *str++ = '-';
    t.sign = MP_ZPOS;
  }

  digs = 0;
  while (mp_iszero (&t) == 0) {
    if ((res = mp_div_d (&t, (mp_digit) radix, &t, &d)) != MP_OKAY) {
      mp_clear (&t);
      return res;
    }
    *str++ = mp_s_rmap[d];
    ++digs;
  }

  /* reverse the digits of the string.  In this case _s points
   * to the first digit [exluding the sign] of the number]
   */
  bn_reverse ((unsigned char *)_s, digs);

  /* append a NULL so the string is properly terminated */
  *str = '\0';

  mp_clear (&t);
  return MP_OKAY;
}

#endif

/* End: bn_mp_toradix.c */

/* Start: bn_mp_toradix_n.c */
#include <tommath.h>
#ifdef BN_MP_TORADIX_N_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* stores a bignum as a ASCII string in a given radix (2..64) 
 *
 * Stores upto maxlen-1 chars and always a NULL byte 
 */
int mp_toradix_n(mp_int * a, char *str, int radix, int maxlen)
{
  int     res, digs;
  mp_int  t;
  mp_digit d;
  char   *_s = str;

  /* check range of the maxlen, radix */
  if (maxlen < 2 || radix < 2 || radix > 64) {
    return MP_VAL;
  }

  /* quick out if its zero */
  if (mp_iszero(a) == MP_YES) {
     *str++ = '0';
     *str = '\0';
     return MP_OKAY;
  }

  if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
    return res;
  }

  /* if it is negative output a - */
  if (t.sign == MP_NEG) {
    /* we have to reverse our digits later... but not the - sign!! */
    ++_s;

    /* store the flag and mark the number as positive */
    *str++ = '-';
    t.sign = MP_ZPOS;
 
    /* subtract a char */
    --maxlen;
  }

  digs = 0;
  while (mp_iszero (&t) == 0) {
    if (--maxlen < 1) {
       /* no more room */
       break;
    }
    if ((res = mp_div_d (&t, (mp_digit) radix, &t, &d)) != MP_OKAY) {
      mp_clear (&t);
      return res;
    }
    *str++ = mp_s_rmap[d];
    ++digs;
  }

  /* reverse the digits of the string.  In this case _s points
   * to the first digit [exluding the sign] of the number
   */
  bn_reverse ((unsigned char *)_s, digs);

  /* append a NULL so the string is properly terminated */
  *str = '\0';

  mp_clear (&t);
  return MP_OKAY;
}

#endif

/* End: bn_mp_toradix_n.c */

/* Start: bn_mp_unsigned_bin_size.c */
#include <tommath.h>
#ifdef BN_MP_UNSIGNED_BIN_SIZE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* get the size for an unsigned equivalent */
int mp_unsigned_bin_size (mp_int * a)
{
  int     size = mp_count_bits (a);
  return (size / 8 + ((size & 7) != 0 ? 1 : 0));
}
#endif

/* End: bn_mp_unsigned_bin_size.c */

/* Start: bn_mp_xor.c */
#include <tommath.h>
#ifdef BN_MP_XOR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* XOR two ints together */
int
mp_xor (mp_int * a, mp_int * b, mp_int * c)
{
  int     res, ix, px;
  mp_int  t, *x;

  if (a->used > b->used) {
    if ((res = mp_init_copy (&t, a)) != MP_OKAY) {
      return res;
    }
    px = b->used;
    x = b;
  } else {
    if ((res = mp_init_copy (&t, b)) != MP_OKAY) {
      return res;
    }
    px = a->used;
    x = a;
  }

  for (ix = 0; ix < px; ix++) {
     t.dp[ix] ^= x->dp[ix];
  }
  mp_clamp (&t);
  mp_exch (c, &t);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_mp_xor.c */

/* Start: bn_mp_zero.c */
#include <tommath.h>
#ifdef BN_MP_ZERO_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* set to zero */
void mp_zero (mp_int * a)
{
  int       n;
  mp_digit *tmp;

  a->sign = MP_ZPOS;
  a->used = 0;

  tmp = a->dp;
  for (n = 0; n < a->alloc; n++) {
     *tmp++ = 0;
  }
}
#endif

/* End: bn_mp_zero.c */

/* Start: bn_prime_tab.c */
#include <tommath.h>
#ifdef BN_PRIME_TAB_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */
const mp_digit ltm_prime_tab[] = {
  0x0002, 0x0003, 0x0005, 0x0007, 0x000B, 0x000D, 0x0011, 0x0013,
  0x0017, 0x001D, 0x001F, 0x0025, 0x0029, 0x002B, 0x002F, 0x0035,
  0x003B, 0x003D, 0x0043, 0x0047, 0x0049, 0x004F, 0x0053, 0x0059,
  0x0061, 0x0065, 0x0067, 0x006B, 0x006D, 0x0071, 0x007F,
#ifndef MP_8BIT
  0x0083,
  0x0089, 0x008B, 0x0095, 0x0097, 0x009D, 0x00A3, 0x00A7, 0x00AD,
  0x00B3, 0x00B5, 0x00BF, 0x00C1, 0x00C5, 0x00C7, 0x00D3, 0x00DF,
  0x00E3, 0x00E5, 0x00E9, 0x00EF, 0x00F1, 0x00FB, 0x0101, 0x0107,
  0x010D, 0x010F, 0x0115, 0x0119, 0x011B, 0x0125, 0x0133, 0x0137,

  0x0139, 0x013D, 0x014B, 0x0151, 0x015B, 0x015D, 0x0161, 0x0167,
  0x016F, 0x0175, 0x017B, 0x017F, 0x0185, 0x018D, 0x0191, 0x0199,
  0x01A3, 0x01A5, 0x01AF, 0x01B1, 0x01B7, 0x01BB, 0x01C1, 0x01C9,
  0x01CD, 0x01CF, 0x01D3, 0x01DF, 0x01E7, 0x01EB, 0x01F3, 0x01F7,
  0x01FD, 0x0209, 0x020B, 0x021D, 0x0223, 0x022D, 0x0233, 0x0239,
  0x023B, 0x0241, 0x024B, 0x0251, 0x0257, 0x0259, 0x025F, 0x0265,
  0x0269, 0x026B, 0x0277, 0x0281, 0x0283, 0x0287, 0x028D, 0x0293,
  0x0295, 0x02A1, 0x02A5, 0x02AB, 0x02B3, 0x02BD, 0x02C5, 0x02CF,

  0x02D7, 0x02DD, 0x02E3, 0x02E7, 0x02EF, 0x02F5, 0x02F9, 0x0301,
  0x0305, 0x0313, 0x031D, 0x0329, 0x032B, 0x0335, 0x0337, 0x033B,
  0x033D, 0x0347, 0x0355, 0x0359, 0x035B, 0x035F, 0x036D, 0x0371,
  0x0373, 0x0377, 0x038B, 0x038F, 0x0397, 0x03A1, 0x03A9, 0x03AD,
  0x03B3, 0x03B9, 0x03C7, 0x03CB, 0x03D1, 0x03D7, 0x03DF, 0x03E5,
  0x03F1, 0x03F5, 0x03FB, 0x03FD, 0x0407, 0x0409, 0x040F, 0x0419,
  0x041B, 0x0425, 0x0427, 0x042D, 0x043F, 0x0443, 0x0445, 0x0449,
  0x044F, 0x0455, 0x045D, 0x0463, 0x0469, 0x047F, 0x0481, 0x048B,

  0x0493, 0x049D, 0x04A3, 0x04A9, 0x04B1, 0x04BD, 0x04C1, 0x04C7,
  0x04CD, 0x04CF, 0x04D5, 0x04E1, 0x04EB, 0x04FD, 0x04FF, 0x0503,
  0x0509, 0x050B, 0x0511, 0x0515, 0x0517, 0x051B, 0x0527, 0x0529,
  0x052F, 0x0551, 0x0557, 0x055D, 0x0565, 0x0577, 0x0581, 0x058F,
  0x0593, 0x0595, 0x0599, 0x059F, 0x05A7, 0x05AB, 0x05AD, 0x05B3,
  0x05BF, 0x05C9, 0x05CB, 0x05CF, 0x05D1, 0x05D5, 0x05DB, 0x05E7,
  0x05F3, 0x05FB, 0x0607, 0x060D, 0x0611, 0x0617, 0x061F, 0x0623,
  0x062B, 0x062F, 0x063D, 0x0641, 0x0647, 0x0649, 0x064D, 0x0653
#endif
};
#endif

/* End: bn_prime_tab.c */

/* Start: bn_reverse.c */
#include <tommath.h>
#ifdef BN_REVERSE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* reverse an array, used for radix code */
void
bn_reverse (unsigned char *s, int len)
{
  int     ix, iy;
  unsigned char t;

  ix = 0;
  iy = len - 1;
  while (ix < iy) {
    t     = s[ix];
    s[ix] = s[iy];
    s[iy] = t;
    ++ix;
    --iy;
  }
}
#endif

/* End: bn_reverse.c */

/* Start: bn_s_mp_add.c */
#include <tommath.h>
#ifdef BN_S_MP_ADD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* low level addition, based on HAC pp.594, Algorithm 14.7 */
int
s_mp_add (mp_int * a, mp_int * b, mp_int * c)
{
  mp_int *x;
  int     olduse, res, min, max;

  /* find sizes, we let |a| <= |b| which means we have to sort
   * them.  "x" will point to the input with the most digits
   */
  if (a->used > b->used) {
    min = b->used;
    max = a->used;
    x = a;
  } else {
    min = a->used;
    max = b->used;
    x = b;
  }

  /* init result */
  if (c->alloc < max + 1) {
    if ((res = mp_grow (c, max + 1)) != MP_OKAY) {
      return res;
    }
  }

  /* get old used digit count and set new one */
  olduse = c->used;
  c->used = max + 1;

  {
    register mp_digit u, *tmpa, *tmpb, *tmpc;
    register int i;

    /* alias for digit pointers */

    /* first input */
    tmpa = a->dp;

    /* second input */
    tmpb = b->dp;

    /* destination */
    tmpc = c->dp;

    /* zero the carry */
    u = 0;
    for (i = 0; i < min; i++) {
      /* Compute the sum at one digit, T[i] = A[i] + B[i] + U */
      *tmpc = *tmpa++ + *tmpb++ + u;

      /* U = carry bit of T[i] */
      u = *tmpc >> ((mp_digit)DIGIT_BIT);

      /* take away carry bit from T[i] */
      *tmpc++ &= MP_MASK;
    }

    /* now copy higher words if any, that is in A+B 
     * if A or B has more digits add those in 
     */
    if (min != max) {
      for (; i < max; i++) {
        /* T[i] = X[i] + U */
        *tmpc = x->dp[i] + u;

        /* U = carry bit of T[i] */
        u = *tmpc >> ((mp_digit)DIGIT_BIT);

        /* take away carry bit from T[i] */
        *tmpc++ &= MP_MASK;
      }
    }

    /* add carry */
    *tmpc++ = u;

    /* clear digits above oldused */
    for (i = c->used; i < olduse; i++) {
      *tmpc++ = 0;
    }
  }

  mp_clamp (c);
  return MP_OKAY;
}
#endif

/* End: bn_s_mp_add.c */

/* Start: bn_s_mp_exptmod.c */
#include <tommath.h>
#ifdef BN_S_MP_EXPTMOD_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */
#ifdef MP_LOW_MEM
   #define TAB_SIZE 32
#else
   #define TAB_SIZE 256
#endif

int s_mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode)
{
  mp_int  M[TAB_SIZE], res, mu;
  mp_digit buf;
  int     err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize;
  int (*redux)(mp_int*,mp_int*,mp_int*);

  /* find window size */
  x = mp_count_bits (X);
  if (x <= 7) {
    winsize = 2;
  } else if (x <= 36) {
    winsize = 3;
  } else if (x <= 140) {
    winsize = 4;
  } else if (x <= 450) {
    winsize = 5;
  } else if (x <= 1303) {
    winsize = 6;
  } else if (x <= 3529) {
    winsize = 7;
  } else {
    winsize = 8;
  }

#ifdef MP_LOW_MEM
    if (winsize > 5) {
       winsize = 5;
    }
#endif

  /* init M array */
  /* init first cell */
  if ((err = mp_init(&M[1])) != MP_OKAY) {
     return err; 
  }

  /* now init the second half of the array */
  for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
    if ((err = mp_init(&M[x])) != MP_OKAY) {
      for (y = 1<<(winsize-1); y < x; y++) {
        mp_clear (&M[y]);
      }
      mp_clear(&M[1]);
      return err;
    }
  }

  /* create mu, used for Barrett reduction */
  if ((err = mp_init (&mu)) != MP_OKAY) {
    goto LBL_M;
  }
  
  if (redmode == 0) {
     if ((err = mp_reduce_setup (&mu, P)) != MP_OKAY) {
        goto LBL_MU;
     }
     redux = mp_reduce;
  } else {
     if ((err = mp_reduce_2k_setup_l (P, &mu)) != MP_OKAY) {
        goto LBL_MU;
     }
     redux = mp_reduce_2k_l;
  }    

  /* create M table
   *
   * The M table contains powers of the base, 
   * e.g. M[x] = G**x mod P
   *
   * The first half of the table is not 
   * computed though accept for M[0] and M[1]
   */
  if ((err = mp_mod (G, P, &M[1])) != MP_OKAY) {
    goto LBL_MU;
  }

  /* compute the value at M[1<<(winsize-1)] by squaring 
   * M[1] (winsize-1) times 
   */
  if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) {
    goto LBL_MU;
  }

  for (x = 0; x < (winsize - 1); x++) {
    /* square it */
    if ((err = mp_sqr (&M[1 << (winsize - 1)], 
                       &M[1 << (winsize - 1)])) != MP_OKAY) {
      goto LBL_MU;
    }

    /* reduce modulo P */
    if ((err = redux (&M[1 << (winsize - 1)], P, &mu)) != MP_OKAY) {
      goto LBL_MU;
    }
  }

  /* create upper table, that is M[x] = M[x-1] * M[1] (mod P)
   * for x = (2**(winsize - 1) + 1) to (2**winsize - 1)
   */
  for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) {
    if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) {
      goto LBL_MU;
    }
    if ((err = redux (&M[x], P, &mu)) != MP_OKAY) {
      goto LBL_MU;
    }
  }

  /* setup result */
  if ((err = mp_init (&res)) != MP_OKAY) {
    goto LBL_MU;
  }
  mp_set (&res, 1);

  /* set initial mode and bit cnt */
  mode   = 0;
  bitcnt = 1;
  buf    = 0;
  digidx = X->used - 1;
  bitcpy = 0;
  bitbuf = 0;

  for (;;) {
    /* grab next digit as required */
    if (--bitcnt == 0) {
      /* if digidx == -1 we are out of digits */
      if (digidx == -1) {
        break;
      }
      /* read next digit and reset the bitcnt */
      buf    = X->dp[digidx--];
      bitcnt = (int) DIGIT_BIT;
    }

    /* grab the next msb from the exponent */
    y     = (buf >> (mp_digit)(DIGIT_BIT - 1)) & 1;
    buf <<= (mp_digit)1;

    /* if the bit is zero and mode == 0 then we ignore it
     * These represent the leading zero bits before the first 1 bit
     * in the exponent.  Technically this opt is not required but it
     * does lower the # of trivial squaring/reductions used
     */
    if (mode == 0 && y == 0) {
      continue;
    }

    /* if the bit is zero and mode == 1 then we square */
    if (mode == 1 && y == 0) {
      if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, &mu)) != MP_OKAY) {
        goto LBL_RES;
      }
      continue;
    }

    /* else we add it to the window */
    bitbuf |= (y << (winsize - ++bitcpy));
    mode    = 2;

    if (bitcpy == winsize) {
      /* ok window is filled so square as required and multiply  */
      /* square first */
      for (x = 0; x < winsize; x++) {
        if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
          goto LBL_RES;
        }
        if ((err = redux (&res, P, &mu)) != MP_OKAY) {
          goto LBL_RES;
        }
      }

      /* then multiply */
      if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, &mu)) != MP_OKAY) {
        goto LBL_RES;
      }

      /* empty window and reset */
      bitcpy = 0;
      bitbuf = 0;
      mode   = 1;
    }
  }

  /* if bits remain then square/multiply */
  if (mode == 2 && bitcpy > 0) {
    /* square then multiply if the bit is set */
    for (x = 0; x < bitcpy; x++) {
      if ((err = mp_sqr (&res, &res)) != MP_OKAY) {
        goto LBL_RES;
      }
      if ((err = redux (&res, P, &mu)) != MP_OKAY) {
        goto LBL_RES;
      }

      bitbuf <<= 1;
      if ((bitbuf & (1 << winsize)) != 0) {
        /* then multiply */
        if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) {
          goto LBL_RES;
        }
        if ((err = redux (&res, P, &mu)) != MP_OKAY) {
          goto LBL_RES;
        }
      }
    }
  }

  mp_exch (&res, Y);
  err = MP_OKAY;
LBL_RES:mp_clear (&res);
LBL_MU:mp_clear (&mu);
LBL_M:
  mp_clear(&M[1]);
  for (x = 1<<(winsize-1); x < (1 << winsize); x++) {
    mp_clear (&M[x]);
  }
  return err;
}
#endif

/* End: bn_s_mp_exptmod.c */

/* Start: bn_s_mp_mul_digs.c */
#include <tommath.h>
#ifdef BN_S_MP_MUL_DIGS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* multiplies |a| * |b| and only computes upto digs digits of result
 * HAC pp. 595, Algorithm 14.12  Modified so you can control how 
 * many digits of output are created.
 */
int s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs)
{
  mp_int  t;
  int     res, pa, pb, ix, iy;
  mp_digit u;
  mp_word r;
  mp_digit tmpx, *tmpt, *tmpy;

  /* can we use the fast multiplier? */
  if (((digs) < MP_WARRAY) &&
      MIN (a->used, b->used) < 
          (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
    return fast_s_mp_mul_digs (a, b, c, digs);
  }

  if ((res = mp_init_size (&t, digs)) != MP_OKAY) {
    return res;
  }
  t.used = digs;

  /* compute the digits of the product directly */
  pa = a->used;
  for (ix = 0; ix < pa; ix++) {
    /* set the carry to zero */
    u = 0;

    /* limit ourselves to making digs digits of output */
    pb = MIN (b->used, digs - ix);

    /* setup some aliases */
    /* copy of the digit from a used within the nested loop */
    tmpx = a->dp[ix];
    
    /* an alias for the destination shifted ix places */
    tmpt = t.dp + ix;
    
    /* an alias for the digits of b */
    tmpy = b->dp;

    /* compute the columns of the output and propagate the carry */
    for (iy = 0; iy < pb; iy++) {
      /* compute the column as a mp_word */
      r       = ((mp_word)*tmpt) +
                ((mp_word)tmpx) * ((mp_word)*tmpy++) +
                ((mp_word) u);

      /* the new column is the lower part of the result */
      *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));

      /* get the carry word from the result */
      u       = (mp_digit) (r >> ((mp_word) DIGIT_BIT));
    }
    /* set carry if it is placed below digs */
    if (ix + iy < digs) {
      *tmpt = u;
    }
  }

  mp_clamp (&t);
  mp_exch (&t, c);

  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_s_mp_mul_digs.c */

/* Start: bn_s_mp_mul_high_digs.c */
#include <tommath.h>
#ifdef BN_S_MP_MUL_HIGH_DIGS_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* multiplies |a| * |b| and does not compute the lower digs digits
 * [meant to get the higher part of the product]
 */
int
s_mp_mul_high_digs (mp_int * a, mp_int * b, mp_int * c, int digs)
{
  mp_int  t;
  int     res, pa, pb, ix, iy;
  mp_digit u;
  mp_word r;
  mp_digit tmpx, *tmpt, *tmpy;

  /* can we use the fast multiplier? */
#ifdef BN_FAST_S_MP_MUL_HIGH_DIGS_C
  if (((a->used + b->used + 1) < MP_WARRAY)
      && MIN (a->used, b->used) < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) {
    return fast_s_mp_mul_high_digs (a, b, c, digs);
  }
#endif

  if ((res = mp_init_size (&t, a->used + b->used + 1)) != MP_OKAY) {
    return res;
  }
  t.used = a->used + b->used + 1;

  pa = a->used;
  pb = b->used;
  for (ix = 0; ix < pa; ix++) {
    /* clear the carry */
    u = 0;

    /* left hand side of A[ix] * B[iy] */
    tmpx = a->dp[ix];

    /* alias to the address of where the digits will be stored */
    tmpt = &(t.dp[digs]);

    /* alias for where to read the right hand side from */
    tmpy = b->dp + (digs - ix);

    for (iy = digs - ix; iy < pb; iy++) {
      /* calculate the double precision result */
      r       = ((mp_word)*tmpt) +
                ((mp_word)tmpx) * ((mp_word)*tmpy++) +
                ((mp_word) u);

      /* get the lower part */
      *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));

      /* carry the carry */
      u       = (mp_digit) (r >> ((mp_word) DIGIT_BIT));
    }
    *tmpt = u;
  }
  mp_clamp (&t);
  mp_exch (&t, c);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_s_mp_mul_high_digs.c */

/* Start: bn_s_mp_sqr.c */
#include <tommath.h>
#ifdef BN_S_MP_SQR_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* low level squaring, b = a*a, HAC pp.596-597, Algorithm 14.16 */
int s_mp_sqr (mp_int * a, mp_int * b)
{
  mp_int  t;
  int     res, ix, iy, pa;
  mp_word r;
  mp_digit u, tmpx, *tmpt;

  pa = a->used;
  if ((res = mp_init_size (&t, 2*pa + 1)) != MP_OKAY) {
    return res;
  }

  /* default used is maximum possible size */
  t.used = 2*pa + 1;

  for (ix = 0; ix < pa; ix++) {
    /* first calculate the digit at 2*ix */
    /* calculate double precision result */
    r = ((mp_word) t.dp[2*ix]) +
        ((mp_word)a->dp[ix])*((mp_word)a->dp[ix]);

    /* store lower part in result */
    t.dp[ix+ix] = (mp_digit) (r & ((mp_word) MP_MASK));

    /* get the carry */
    u           = (mp_digit)(r >> ((mp_word) DIGIT_BIT));

    /* left hand side of A[ix] * A[iy] */
    tmpx        = a->dp[ix];

    /* alias for where to store the results */
    tmpt        = t.dp + (2*ix + 1);
    
    for (iy = ix + 1; iy < pa; iy++) {
      /* first calculate the product */
      r       = ((mp_word)tmpx) * ((mp_word)a->dp[iy]);

      /* now calculate the double precision result, note we use
       * addition instead of *2 since it's easier to optimize
       */
      r       = ((mp_word) *tmpt) + r + r + ((mp_word) u);

      /* store lower part */
      *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));

      /* get carry */
      u       = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
    }
    /* propagate upwards */
    while (u != ((mp_digit) 0)) {
      r       = ((mp_word) *tmpt) + ((mp_word) u);
      *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK));
      u       = (mp_digit)(r >> ((mp_word) DIGIT_BIT));
    }
  }

  mp_clamp (&t);
  mp_exch (&t, b);
  mp_clear (&t);
  return MP_OKAY;
}
#endif

/* End: bn_s_mp_sqr.c */

/* Start: bn_s_mp_sub.c */
#include <tommath.h>
#ifdef BN_S_MP_SUB_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* low level subtraction (assumes |a| > |b|), HAC pp.595 Algorithm 14.9 */
int
s_mp_sub (mp_int * a, mp_int * b, mp_int * c)
{
  int     olduse, res, min, max;

  /* find sizes */
  min = b->used;
  max = a->used;

  /* init result */
  if (c->alloc < max) {
    if ((res = mp_grow (c, max)) != MP_OKAY) {
      return res;
    }
  }
  olduse = c->used;
  c->used = max;

  {
    register mp_digit u, *tmpa, *tmpb, *tmpc;
    register int i;

    /* alias for digit pointers */
    tmpa = a->dp;
    tmpb = b->dp;
    tmpc = c->dp;

    /* set carry to zero */
    u = 0;
    for (i = 0; i < min; i++) {
      /* T[i] = A[i] - B[i] - U */
      *tmpc = *tmpa++ - *tmpb++ - u;

      /* U = carry bit of T[i]
       * Note this saves performing an AND operation since
       * if a carry does occur it will propagate all the way to the
       * MSB.  As a result a single shift is enough to get the carry
       */
      u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1));

      /* Clear carry from T[i] */
      *tmpc++ &= MP_MASK;
    }

    /* now copy higher words if any, e.g. if A has more digits than B  */
    for (; i < max; i++) {
      /* T[i] = A[i] - U */
      *tmpc = *tmpa++ - u;

      /* U = carry bit of T[i] */
      u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1));

      /* Clear carry from T[i] */
      *tmpc++ &= MP_MASK;
    }

    /* clear digits above used (since we may not have grown result above) */
    for (i = c->used; i < olduse; i++) {
      *tmpc++ = 0;
    }
  }

  mp_clamp (c);
  return MP_OKAY;
}

#endif

/* End: bn_s_mp_sub.c */

/* Start: bncore.c */
#include <tommath.h>
#ifdef BNCORE_C
/* LibTomMath, multiple-precision integer library -- Tom St Denis
 *
 * LibTomMath is a library that provides multiple-precision
 * integer arithmetic as well as number theoretic functionality.
 *
 * The library was designed directly after the MPI library by
 * Michael Fromberger but has been written from scratch with
 * additional optimizations in place.
 *
 * The library is free for all purposes without any express
 * guarantee it works.
 *
 * Tom St Denis, [email protected], http://math.libtomcrypt.com
 */

/* Known optimal configurations

 CPU                    /Compiler     /MUL CUTOFF/SQR CUTOFF
-------------------------------------------------------------
 Intel P4 Northwood     /GCC v3.4.1   /        88/       128/LTM 0.32 ;-)
 AMD Athlon64           /GCC v3.4.4   /        80/       120/LTM 0.35
 
*/

int     KARATSUBA_MUL_CUTOFF = 80,      /* Min. number of digits before Karatsuba multiplication is used. */
        KARATSUBA_SQR_CUTOFF = 120,     /* Min. number of digits before Karatsuba squaring is used. */
        
        TOOM_MUL_CUTOFF      = 350,      /* no optimal values of these are known yet so set em high */
        TOOM_SQR_CUTOFF      = 400; 
#endif

/* End: bncore.c */


/* EOF */