addition of FFT multiplication/squaring

Author: czurnieden

bn_mp_mul.c

@@ -43,54 +43,60 @@ int mp_mul(const mp_int *a, const mp_int *b, mp_int *c)
 GO_ON:
 #endif
 
+#ifdef BN_S_MP_FFT_MUL_C
+   if ((MP_MIN(a->used, b->used) >= FFT_MUL_LOWER_CO) && (MP_MAX(a->used, b->used) < FFT_MUL_UPPER_CO)) {
+      res = s_mp_fft_mul(a, b, c);
+   } else
+#endif
+
 #ifdef BN_S_MP_TOOM_COOK_5_MUL_C
 
-   if (MP_MIN(a->used, b->used) >= TOOM_COOK_5_MUL_CO) {
-      res = s_mp_toom_cook_5_mul(a, b, c);
-   } else
+      if (MP_MIN(a->used, b->used) >= TOOM_COOK_5_MUL_CO) {
+         res = s_mp_toom_cook_5_mul(a, b, c);
+      } else
 #endif
 
 #ifdef BN_S_MP_TOOM_COOK_4_MUL_C
-      if (MP_MIN(a->used, b->used) >= TOOM_COOK_4_MUL_CO) {
-         res = s_mp_toom_cook_4_mul(a, b, c);
-      } else
+         if (MP_MIN(a->used, b->used) >= TOOM_COOK_4_MUL_CO) {
+            res = s_mp_toom_cook_4_mul(a, b, c);
+         } else
 #endif
-         /* use Toom-Cook? */
+            /* use Toom-Cook? */
 #ifdef BN_S_MP_TOOM_MUL_C
-         if (MP_MIN(a->used, b->used) >= TOOM_MUL_CUTOFF) {
-            res = s_mp_toom_mul(a, b, c);
-         } else
+            if (MP_MIN(a->used, b->used) >= TOOM_MUL_CUTOFF) {
+               res = s_mp_toom_mul(a, b, c);
+            } else
 #endif
 #ifdef BN_S_MP_KARATSUBA_MUL_C
-            /* use Karatsuba? */
-            if (MP_MIN(a->used, b->used) >= KARATSUBA_MUL_CUTOFF) {
-               res = s_mp_karatsuba_mul(a, b, c);
-            } else
+               /* use Karatsuba? */
+               if (MP_MIN(a->used, b->used) >= KARATSUBA_MUL_CUTOFF) {
+                  res = s_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;
+               {
+                  /* 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_S_MP_MUL_DIGS_FAST_C
-               if ((digs < (int)MP_WARRAY) &&
-                   (MP_MIN(a->used, b->used) <=
-                    (int)(1u << ((CHAR_BIT * sizeof(mp_word)) - (2u * (size_t)DIGIT_BIT))))) {
-                  res = s_mp_mul_digs_fast(a, b, c, digs);
-               } else
+                  if ((digs < (int)MP_WARRAY) &&
+                      (MP_MIN(a->used, b->used) <=
+                       (int)(1u << ((CHAR_BIT * sizeof(mp_word)) - (2u * (size_t)MP_DIGIT_BIT))))) {
+                     res = s_mp_mul_digs_fast(a, b, c, digs);
+                  } else
 #endif
-               {
+                  {
 #ifdef BN_S_MP_MUL_DIGS_C
-                  res = s_mp_mul_digs(a, b, c, a->used + b->used + 1);
+                     res = s_mp_mul_digs(a, b, c, a->used + b->used + 1);
 #else
-                  res = MP_VAL;
+                     res = MP_VAL;
 #endif
+                  }
                }
-            }
 END:
    c->sign = (c->used > 0) ? neg : MP_ZPOS;
    return res;

bn_mp_sqr.c

@@ -8,47 +8,53 @@ int mp_sqr(const mp_int *a, mp_int *b)
 {
    int     res;
 
-#ifdef BN_S_MP_TOOM_COOK_5_SQR_C
-   if (a->used >= TOOM_COOK_5_SQR_CO) {
-      res = s_mp_toom_cook_5_sqr(a, b);
+#ifdef BN_S_MP_FFT_SQR_C
+   if ((a->used >= FFT_SQR_LOWER_CO) && (a->used < FFT_SQR_UPPER_CO)) {
+      res = s_mp_fft_sqr(a, b);
    } else
 #endif
 
-#ifdef BN_S_MP_TOOM_COOK_4_SQR_C
-      if (a->used >= TOOM_COOK_4_SQR_CO) {
-         res = s_mp_toom_cook_4_sqr(a, b);
+#ifdef BN_S_MP_TOOM_COOK_5_SQR_C
+      if (a->used >= TOOM_COOK_5_SQR_CO) {
+         res = s_mp_toom_cook_5_sqr(a, b);
       } else
 #endif
 
-#ifdef BN_S_MP_TOOM_SQR_C
-         /* use Toom-Cook? */
-         if (a->used >= TOOM_SQR_CUTOFF) {
-            res = s_mp_toom_sqr(a, b);
-            /* Karatsuba? */
+#ifdef BN_S_MP_TOOM_COOK_4_SQR_C
+         if (a->used >= TOOM_COOK_4_SQR_CO) {
+            res = s_mp_toom_cook_4_sqr(a, b);
          } else
 #endif
-#ifdef BN_S_MP_KARATSUBA_SQR_C
-            if (a->used >= KARATSUBA_SQR_CUTOFF) {
-               res = s_mp_karatsuba_sqr(a, b);
+
+#ifdef BN_S_MP_TOOM_SQR_C
+            /* use Toom-Cook? */
+            if (a->used >= TOOM_SQR_CUTOFF) {
+               res = s_mp_toom_sqr(a, b);
+               /* Karatsuba? */
             } else
 #endif
-            {
-#ifdef BN_S_MP_SQR_FAST_C
-               /* can we use the fast comba multiplier? */
-               if ((((a->used * 2) + 1) < (int)MP_WARRAY) &&
-                   (a->used <
-                    (int)(1u << (((CHAR_BIT * sizeof(mp_word)) - (2u * (size_t)DIGIT_BIT)) - 1u)))) {
-                  res = s_mp_sqr_fast(a, b);
+#ifdef BN_S_MP_KARATSUBA_SQR_C
+               if (a->used >= KARATSUBA_SQR_CUTOFF) {
+                  res = s_mp_karatsuba_sqr(a, b);
                } else
 #endif
                {
+#ifdef BN_S_MP_SQR_FAST_C
+                  /* can we use the fast comba multiplier? */
+                  if ((((a->used * 2) + 1) < (int)MP_WARRAY) &&
+                      (a->used <
+                       (int)(1u << (((CHAR_BIT * sizeof(mp_word)) - (2u * (size_t)MP_DIGIT_BIT)) - 1u)))) {
+                     res = s_mp_sqr_fast(a, b);
+                  } else
+#endif
+                  {
 #ifdef BN_S_MP_SQR_C
-                  res = s_mp_sqr(a, b);
+                     res = s_mp_sqr(a, b);
 #else
-                  res = MP_VAL;
+                     res = MP_VAL;
 #endif
+                  }
                }
-            }
    b->sign = MP_ZPOS;
    return res;
 }