Various fixes + split log_u32

libtommath_VS2008.vcproj

@@ -880,6 +880,14 @@
 			RelativePath="s_mp_karatsuba_sqr.c"
 			>
 		</File>
+		<File
+			RelativePath="s_mp_log.c"
+			>
+		</File>
+		<File
+			RelativePath="s_mp_log_d.c"
+			>
+		</File>
 		<File
 			RelativePath="s_mp_montgomery_reduce_fast.c"
 			>

makefile

@@ -46,9 +46,10 @@ mp_set_ll.o mp_set_u32.o mp_set_u64.o mp_set_ul.o mp_set_ull.o mp_shrink.o mp_si
 mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_radix.o mp_to_sbin.o \
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_balance_mul.o s_mp_exptmod.o \
 s_mp_exptmod_fast.o s_mp_get_bit.o s_mp_invmod_fast.o s_mp_invmod_slow.o s_mp_karatsuba_mul.o \
-s_mp_karatsuba_sqr.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o s_mp_mul_digs_fast.o \
-s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o s_mp_rand_jenkins.o \
-s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o s_mp_toom_mul.o s_mp_toom_sqr.o
+s_mp_karatsuba_sqr.o s_mp_log.o s_mp_log_d.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o \
+s_mp_mul_digs_fast.o s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o \
+s_mp_rand_jenkins.o s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o \
+s_mp_toom_mul.o s_mp_toom_sqr.o
 
 #END_INS
 

makefile.mingw

@@ -49,9 +49,10 @@ mp_set_ll.o mp_set_u32.o mp_set_u64.o mp_set_ul.o mp_set_ull.o mp_shrink.o mp_si
 mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_radix.o mp_to_sbin.o \
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_balance_mul.o s_mp_exptmod.o \
 s_mp_exptmod_fast.o s_mp_get_bit.o s_mp_invmod_fast.o s_mp_invmod_slow.o s_mp_karatsuba_mul.o \
-s_mp_karatsuba_sqr.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o s_mp_mul_digs_fast.o \
-s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o s_mp_rand_jenkins.o \
-s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o s_mp_toom_mul.o s_mp_toom_sqr.o
+s_mp_karatsuba_sqr.o s_mp_log.o s_mp_log_d.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o \
+s_mp_mul_digs_fast.o s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o \
+s_mp_rand_jenkins.o s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o \
+s_mp_toom_mul.o s_mp_toom_sqr.o
 
 HEADERS_PUB=tommath.h
 HEADERS=tommath_private.h tommath_class.h tommath_superclass.h tommath_cutoffs.h $(HEADERS_PUB)

makefile.msvc

@@ -41,9 +41,10 @@ mp_set_ll.obj mp_set_u32.obj mp_set_u64.obj mp_set_ul.obj mp_set_ull.obj mp_shri
 mp_sqrmod.obj mp_sqrt.obj mp_sqrtmod_prime.obj mp_sub.obj mp_sub_d.obj mp_submod.obj mp_to_radix.obj mp_to_sbin.obj \
 mp_to_ubin.obj mp_ubin_size.obj mp_unpack.obj mp_xor.obj mp_zero.obj s_mp_add.obj s_mp_balance_mul.obj s_mp_exptmod.obj \
 s_mp_exptmod_fast.obj s_mp_get_bit.obj s_mp_invmod_fast.obj s_mp_invmod_slow.obj s_mp_karatsuba_mul.obj \
-s_mp_karatsuba_sqr.obj s_mp_montgomery_reduce_fast.obj s_mp_mul_digs.obj s_mp_mul_digs_fast.obj \
-s_mp_mul_high_digs.obj s_mp_mul_high_digs_fast.obj s_mp_prime_is_divisible.obj s_mp_rand_jenkins.obj \
-s_mp_rand_platform.obj s_mp_reverse.obj s_mp_sqr.obj s_mp_sqr_fast.obj s_mp_sub.obj s_mp_toom_mul.obj s_mp_toom_sqr.obj
+s_mp_karatsuba_sqr.obj s_mp_log.obj s_mp_log_d.obj s_mp_montgomery_reduce_fast.obj s_mp_mul_digs.obj \
+s_mp_mul_digs_fast.obj s_mp_mul_high_digs.obj s_mp_mul_high_digs_fast.obj s_mp_prime_is_divisible.obj \
+s_mp_rand_jenkins.obj s_mp_rand_platform.obj s_mp_reverse.obj s_mp_sqr.obj s_mp_sqr_fast.obj s_mp_sub.obj \
+s_mp_toom_mul.obj s_mp_toom_sqr.obj
 
 HEADERS_PUB=tommath.h
 HEADERS=tommath_private.h tommath_class.h tommath_superclass.h tommath_cutoffs.h $(HEADERS_PUB)

makefile.shared

@@ -43,9 +43,10 @@ mp_set_ll.o mp_set_u32.o mp_set_u64.o mp_set_ul.o mp_set_ull.o mp_shrink.o mp_si
 mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_radix.o mp_to_sbin.o \
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_balance_mul.o s_mp_exptmod.o \
 s_mp_exptmod_fast.o s_mp_get_bit.o s_mp_invmod_fast.o s_mp_invmod_slow.o s_mp_karatsuba_mul.o \
-s_mp_karatsuba_sqr.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o s_mp_mul_digs_fast.o \
-s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o s_mp_rand_jenkins.o \
-s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o s_mp_toom_mul.o s_mp_toom_sqr.o
+s_mp_karatsuba_sqr.o s_mp_log.o s_mp_log_d.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o \
+s_mp_mul_digs_fast.o s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o \
+s_mp_rand_jenkins.o s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o \
+s_mp_toom_mul.o s_mp_toom_sqr.o
 
 #END_INS
 

makefile.unix

@@ -50,9 +50,10 @@ mp_set_ll.o mp_set_u32.o mp_set_u64.o mp_set_ul.o mp_set_ull.o mp_shrink.o mp_si
 mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_radix.o mp_to_sbin.o \
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_balance_mul.o s_mp_exptmod.o \
 s_mp_exptmod_fast.o s_mp_get_bit.o s_mp_invmod_fast.o s_mp_invmod_slow.o s_mp_karatsuba_mul.o \
-s_mp_karatsuba_sqr.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o s_mp_mul_digs_fast.o \
-s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o s_mp_rand_jenkins.o \
-s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o s_mp_toom_mul.o s_mp_toom_sqr.o
+s_mp_karatsuba_sqr.o s_mp_log.o s_mp_log_d.o s_mp_montgomery_reduce_fast.o s_mp_mul_digs.o \
+s_mp_mul_digs_fast.o s_mp_mul_high_digs.o s_mp_mul_high_digs_fast.o s_mp_prime_is_divisible.o \
+s_mp_rand_jenkins.o s_mp_rand_platform.o s_mp_reverse.o s_mp_sqr.o s_mp_sqr_fast.o s_mp_sub.o \
+s_mp_toom_mul.o s_mp_toom_sqr.o
 
 HEADERS_PUB=tommath.h
 HEADERS=tommath_private.h tommath_class.h tommath_superclass.h tommath_cutoffs.h $(HEADERS_PUB)

mp_log_u32.c

@@ -3,82 +3,8 @@
 /* LibTomMath, multiple-precision integer library -- Tom St Denis */
 /* SPDX-License-Identifier: Unlicense */
 
-/* Compute log_{base}(a) */
-static mp_word s_pow(mp_word base, mp_word exponent)
-{
-   mp_word result = 1uLL;
-   while (exponent != 0u) {
-      if ((exponent & 1u) == 1u) {
-         result *= base;
-      }
-      exponent >>= 1;
-      base *= base;
-   }
-
-   return result;
-}
-
-static mp_digit s_digit_ilogb(mp_digit base, mp_digit n)
-{
-   mp_word bracket_low = 1uLL, bracket_mid, bracket_high, N;
-   mp_digit ret, high = 1uL, low = 0uL, mid;
-
-   if (n < base) {
-      return 0uL;
-   }
-   if (n == base) {
-      return 1uL;
-   }
-
-   bracket_high = (mp_word) base ;
-   N = (mp_word) n;
-
-   while (bracket_high < N) {
-      low = high;
-      bracket_low = bracket_high;
-      high <<= 1;
-      bracket_high *= bracket_high;
-   }
-
-   while (((mp_digit)(high - low)) > 1uL) {
-      mid = (low + high) >> 1;
-      bracket_mid = bracket_low * s_pow(base, (mp_word)(mid - low));
-
-      if (N < bracket_mid) {
-         high = mid ;
-         bracket_high = bracket_mid ;
-      }
-      if (N > bracket_mid) {
-         low = mid ;
-         bracket_low = bracket_mid ;
-      }
-      if (N == bracket_mid) {
-         return (mp_digit) mid;
-      }
-   }
-
-   if (bracket_high == N) {
-      ret = high;
-   } else {
-      ret = low;
-   }
-
-   return ret;
-}
-
-/* TODO: output could be "int" because the output of mp_radix_size is int, too,
-         as is the output of mp_bitcount.
-         With the same problem: max size is INT_MAX * MP_DIGIT not INT_MAX only!
-*/
 mp_err mp_log_u32(const mp_int *a, uint32_t base, uint32_t *c)
 {
-   mp_err err;
-   mp_ord cmp;
-   uint32_t high, low, mid;
-   mp_int bracket_low, bracket_high, bracket_mid, t, bi_base;
-
-   err = MP_OKAY;
-
    if (a->sign == MP_NEG) {
       return MP_VAL;
    }
@@ -102,79 +28,16 @@ mp_err mp_log_u32(const mp_int *a, uint32_t base, uint32_t *c)
       return MP_OKAY;
    }
 
-   if (a->used == 1) {
-      *c = (uint32_t)s_digit_ilogb(base, a->dp[0]);
-      return err;
-   }
-
-   cmp = mp_cmp_d(a, base);
-   if ((cmp == MP_LT) || (cmp == MP_EQ)) {
-      *c = cmp == MP_EQ;
-      return err;
-   }
-
-   if ((err =
-           mp_init_multi(&bracket_low, &bracket_high,
-                         &bracket_mid, &t, &bi_base, NULL)) != MP_OKAY) {
-      return err;
-   }
-
-   low = 0u;
-   mp_set(&bracket_low, 1uL);
-   high = 1u;
-
-   mp_set(&bracket_high, base);
-
-   /*
-       A kind of Giant-step/baby-step algorithm.
-       Idea shamelessly stolen from https://programmingpraxis.com/2010/05/07/integer-logarithms/2/
-       The effect is asymptotic, hence needs benchmarks to test if the Giant-step should be skipped
-       for small n.
-    */
-   while (mp_cmp(&bracket_high, a) == MP_LT) {
-      low = high;
-      if ((err = mp_copy(&bracket_high, &bracket_low)) != MP_OKAY) {
-         goto LBL_ERR;
-      }
-      high <<= 1;
-      if ((err = mp_sqr(&bracket_high, &bracket_high)) != MP_OKAY) {
-         goto LBL_ERR;
-      }
+   if (MP_HAS(S_MP_LOG_D) && a->used == 1) {
+      *c = (uint32_t)s_mp_log_d(base, a->dp[0]);
+      return MP_OKAY;
    }
-   mp_set(&bi_base, base);
 
-   while ((high - low) > 1u) {
-      mid = (high + low) >> 1;
-
-      if ((err = mp_expt_u32(&bi_base, (uint32_t)(mid - low), &t)) != MP_OKAY) {
-         goto LBL_ERR;
-      }
-      if ((err = mp_mul(&bracket_low, &t, &bracket_mid)) != MP_OKAY) {
-         goto LBL_ERR;
-      }
-      cmp = mp_cmp(a, &bracket_mid);
-      if (cmp == MP_LT) {
-         high = mid;
-         mp_exch(&bracket_mid, &bracket_high);
-      }
-      if (cmp == MP_GT) {
-         low = mid;
-         mp_exch(&bracket_mid, &bracket_low);
-      }
-      if (cmp == MP_EQ) {
-         *c = mid;
-         goto LBL_END;
-      }
+   if (MP_HAS(S_MP_LOG)) {
+      return s_mp_log(a, base, c);
    }
 
-   *c = (mp_cmp(&bracket_high, a) == MP_EQ) ? high : low;
-
-LBL_END:
-LBL_ERR:
-   mp_clear_multi(&bracket_low, &bracket_high, &bracket_mid,
-                  &t, &bi_base, NULL);
-   return err;
+   return MP_VAL;
 }
 
-
 #endif

mp_prime_is_prime.c

@@ -100,11 +100,6 @@ mp_err mp_prime_is_prime(const mp_int *a, int t, mp_bool *result)
     */
 #ifndef LTM_USE_ONLY_MR
    if (t >= 0) {
-      /*
-       * Use a Frobenius-Underwood test instead of the Lucas-Selfridge test for
-       * MP_8BIT (It is unknown if the Lucas-Selfridge test works with 16-bit
-       * integers but the necesssary analysis is on the todo-list).
-       */
 #ifdef LTM_USE_FROBENIUS_TEST
       err = mp_prime_frobenius_underwood(a, &res);
       if ((err != MP_OKAY) && (err != MP_ITER)) {

mp_prime_strong_lucas_selfridge.c

@@ -52,7 +52,6 @@ mp_err mp_prime_strong_lucas_selfridge(const mp_int *a, mp_bool *result)
 {
    /* CZ TODO: choose better variable names! */
    mp_int Dz, gcd, Np1, Uz, Vz, U2mz, V2mz, Qmz, Q2mz, Qkdz, T1z, T2z, T3z, T4z, Q2kdz;
-   /* CZ TODO: Some of them need the full 32 bit, hence the (temporary) exclusion of MP_8BIT */
    int32_t D, Ds, J, sign, P, Q, r, s, u, Nbits;
    mp_err err;
    mp_bool oddness;

mp_sqrtmod_prime.c

@@ -59,7 +59,7 @@ mp_err mp_sqrtmod_prime(const mp_int *n, const mp_int *prime, mp_int *ret)
    }
 
    /* find a Z such that the Legendre symbol (Z|prime) == -1 */
-   mp_set_u32(&Z, 2u);
+   mp_set(&Z, 2u);
    /* Z = 2 */
    for (;;) {
       if ((err = mp_kronecker(&Z, prime, &legendre)) != MP_OKAY)     goto cleanup;
@@ -79,7 +79,7 @@ mp_err mp_sqrtmod_prime(const mp_int *n, const mp_int *prime, mp_int *ret)
    /* T = n ^ Q mod prime */
    if ((err = mp_copy(&S, &M)) != MP_OKAY)                       goto cleanup;
    /* M = S */
-   mp_set_u32(&two, 2u);
+   mp_set(&two, 2u);
 
    for (;;) {
       if ((err = mp_copy(&T, &t1)) != MP_OKAY)                    goto cleanup;

s_mp_log.c

@@ -0,0 +1,82 @@
+#include "tommath_private.h"
+#ifdef S_MP_LOG_C
+/* LibTomMath, multiple-precision integer library -- Tom St Denis */
+/* SPDX-License-Identifier: Unlicense */
+
+mp_err s_mp_log(const mp_int *a, uint32_t base, uint32_t *c)
+{
+   mp_err err;
+   mp_ord cmp;
+   uint32_t high, low, mid;
+   mp_int bracket_low, bracket_high, bracket_mid, t, bi_base;
+
+   cmp = mp_cmp_d(a, base);
+   if ((cmp == MP_LT) || (cmp == MP_EQ)) {
+      *c = cmp == MP_EQ;
+      return MP_OKAY;
+   }
+
+   if ((err =
+           mp_init_multi(&bracket_low, &bracket_high,
+                         &bracket_mid, &t, &bi_base, NULL)) != MP_OKAY) {
+      return err;
+   }
+
+   low = 0u;
+   mp_set(&bracket_low, 1uL);
+   high = 1u;
+
+   mp_set(&bracket_high, base);
+
+   /*
+       A kind of Giant-step/baby-step algorithm.
+       Idea shamelessly stolen from https://programmingpraxis.com/2010/05/07/integer-logarithms/2/
+       The effect is asymptotic, hence needs benchmarks to test if the Giant-step should be skipped
+       for small n.
+    */
+   while (mp_cmp(&bracket_high, a) == MP_LT) {
+      low = high;
+      if ((err = mp_copy(&bracket_high, &bracket_low)) != MP_OKAY) {
+         goto LBL_END;
+      }
+      high <<= 1;
+      if ((err = mp_sqr(&bracket_high, &bracket_high)) != MP_OKAY) {
+         goto LBL_END;
+      }
+   }
+   mp_set(&bi_base, base);
+
+   while ((high - low) > 1u) {
+      mid = (high + low) >> 1;
+
+      if ((err = mp_expt_u32(&bi_base, (uint32_t)(mid - low), &t)) != MP_OKAY) {
+         goto LBL_END;
+      }
+      if ((err = mp_mul(&bracket_low, &t, &bracket_mid)) != MP_OKAY) {
+         goto LBL_END;
+      }
+      cmp = mp_cmp(a, &bracket_mid);
+      if (cmp == MP_LT) {
+         high = mid;
+         mp_exch(&bracket_mid, &bracket_high);
+      }
+      if (cmp == MP_GT) {
+         low = mid;
+         mp_exch(&bracket_mid, &bracket_low);
+      }
+      if (cmp == MP_EQ) {
+         *c = mid;
+         goto LBL_END;
+      }
+   }
+
+   *c = (mp_cmp(&bracket_high, a) == MP_EQ) ? high : low;
+
+LBL_END:
+   mp_clear_multi(&bracket_low, &bracket_high, &bracket_mid,
+                  &t, &bi_base, NULL);
+   return err;
+}
+
+
+#endif