lift comba limit for s_mp_mul_comba

this is how it is done in tfm

Author: minad

demo/test.c

@@ -1895,7 +1895,7 @@ static int test_s_mp_mul_balance(void)
    return EXIT_FAILURE;
 }
 
-#define s_mp_mul_full(a, b, c) s_mp_mul(a, b, c, (a)->used + (b)->used + 1)
+#define s_mp_mul_full(a, b, c) s_mp_mul_comba(a, b, c, (a)->used + (b)->used + 1)
 static int test_s_mp_mul_karatsuba(void)
 {
    mp_int a, b, c, d;

etc/tune.c

@@ -60,11 +60,7 @@ static int s_offset = 1;
 
 static mp_err s_mul_full(const mp_int *a, const mp_int *b, mp_int *c)
 {
-   if (MP_HAS(S_MP_MUL_HIGH_COMBA)
-       && (MP_MIN(a->used, b->used) < MP_MAX_COMBA)) {
-      return s_mp_mul_comba(a, b, c, a->used + b->used + 1);
-   }
-   return s_mp_mul(a, b, c, a->used + b->used + 1);
+   return s_mp_mul_comba(a, b, c, a->used + b->used + 1);
 }
 
 static uint64_t s_time_mul(int size)

libtommath_VS2008.vcproj

@@ -868,10 +868,6 @@
 			RelativePath="s_mp_montgomery_reduce_comba.c"
 			>
 		</File>
-		<File
-			RelativePath="s_mp_mul.c"
-			>
-		</File>
 		<File
 			RelativePath="s_mp_mul_balance.c"
 			>

makefile

@@ -46,7 +46,7 @@ mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_r
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_copy_digs.o s_mp_div_3.o \
 s_mp_div_recursive.o s_mp_div_school.o s_mp_div_small.o s_mp_exptmod.o s_mp_exptmod_fast.o s_mp_get_bit.o \
 s_mp_invmod.o s_mp_invmod_odd.o s_mp_log.o s_mp_log_d.o s_mp_log_pow2.o s_mp_montgomery_reduce_comba.o \
-s_mp_mul.o s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
+s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
 s_mp_mul_toom.o s_mp_prime_is_divisible.o s_mp_prime_tab.o s_mp_radix_map.o s_mp_rand_jenkins.o \
 s_mp_rand_platform.o s_mp_sqr.o s_mp_sqr_comba.o s_mp_sqr_karatsuba.o s_mp_sqr_toom.o s_mp_sub.o \
 s_mp_zero_buf.o s_mp_zero_digs.o

makefile.mingw

@@ -48,7 +48,7 @@ mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_r
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_copy_digs.o s_mp_div_3.o \
 s_mp_div_recursive.o s_mp_div_school.o s_mp_div_small.o s_mp_exptmod.o s_mp_exptmod_fast.o s_mp_get_bit.o \
 s_mp_invmod.o s_mp_invmod_odd.o s_mp_log.o s_mp_log_d.o s_mp_log_pow2.o s_mp_montgomery_reduce_comba.o \
-s_mp_mul.o s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
+s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
 s_mp_mul_toom.o s_mp_prime_is_divisible.o s_mp_prime_tab.o s_mp_radix_map.o s_mp_rand_jenkins.o \
 s_mp_rand_platform.o s_mp_sqr.o s_mp_sqr_comba.o s_mp_sqr_karatsuba.o s_mp_sqr_toom.o s_mp_sub.o \
 s_mp_zero_buf.o s_mp_zero_digs.o

makefile.msvc

@@ -41,7 +41,7 @@ mp_sqrmod.obj mp_sqrt.obj mp_sqrtmod_prime.obj mp_sub.obj mp_sub_d.obj mp_submod
 mp_to_ubin.obj mp_ubin_size.obj mp_unpack.obj mp_xor.obj mp_zero.obj s_mp_add.obj s_mp_copy_digs.obj s_mp_div_3.obj \
 s_mp_div_recursive.obj s_mp_div_school.obj s_mp_div_small.obj s_mp_exptmod.obj s_mp_exptmod_fast.obj s_mp_get_bit.obj \
 s_mp_invmod.obj s_mp_invmod_odd.obj s_mp_log.obj s_mp_log_d.obj s_mp_log_pow2.obj s_mp_montgomery_reduce_comba.obj \
-s_mp_mul.obj s_mp_mul_balance.obj s_mp_mul_comba.obj s_mp_mul_high.obj s_mp_mul_high_comba.obj s_mp_mul_karatsuba.obj \
+s_mp_mul_balance.obj s_mp_mul_comba.obj s_mp_mul_high.obj s_mp_mul_high_comba.obj s_mp_mul_karatsuba.obj \
 s_mp_mul_toom.obj s_mp_prime_is_divisible.obj s_mp_prime_tab.obj s_mp_radix_map.obj s_mp_rand_jenkins.obj \
 s_mp_rand_platform.obj s_mp_sqr.obj s_mp_sqr_comba.obj s_mp_sqr_karatsuba.obj s_mp_sqr_toom.obj s_mp_sub.obj \
 s_mp_zero_buf.obj s_mp_zero_digs.obj

makefile.shared

@@ -43,7 +43,7 @@ mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_r
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_copy_digs.o s_mp_div_3.o \
 s_mp_div_recursive.o s_mp_div_school.o s_mp_div_small.o s_mp_exptmod.o s_mp_exptmod_fast.o s_mp_get_bit.o \
 s_mp_invmod.o s_mp_invmod_odd.o s_mp_log.o s_mp_log_d.o s_mp_log_pow2.o s_mp_montgomery_reduce_comba.o \
-s_mp_mul.o s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
+s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
 s_mp_mul_toom.o s_mp_prime_is_divisible.o s_mp_prime_tab.o s_mp_radix_map.o s_mp_rand_jenkins.o \
 s_mp_rand_platform.o s_mp_sqr.o s_mp_sqr_comba.o s_mp_sqr_karatsuba.o s_mp_sqr_toom.o s_mp_sub.o \
 s_mp_zero_buf.o s_mp_zero_digs.o

makefile.unix

@@ -49,7 +49,7 @@ mp_sqrmod.o mp_sqrt.o mp_sqrtmod_prime.o mp_sub.o mp_sub_d.o mp_submod.o mp_to_r
 mp_to_ubin.o mp_ubin_size.o mp_unpack.o mp_xor.o mp_zero.o s_mp_add.o s_mp_copy_digs.o s_mp_div_3.o \
 s_mp_div_recursive.o s_mp_div_school.o s_mp_div_small.o s_mp_exptmod.o s_mp_exptmod_fast.o s_mp_get_bit.o \
 s_mp_invmod.o s_mp_invmod_odd.o s_mp_log.o s_mp_log_d.o s_mp_log_pow2.o s_mp_montgomery_reduce_comba.o \
-s_mp_mul.o s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
+s_mp_mul_balance.o s_mp_mul_comba.o s_mp_mul_high.o s_mp_mul_high_comba.o s_mp_mul_karatsuba.o \
 s_mp_mul_toom.o s_mp_prime_is_divisible.o s_mp_prime_tab.o s_mp_radix_map.o s_mp_rand_jenkins.o \
 s_mp_rand_platform.o s_mp_sqr.o s_mp_sqr_comba.o s_mp_sqr_karatsuba.o s_mp_sqr_toom.o s_mp_sub.o \
 s_mp_zero_buf.o s_mp_zero_digs.o

mp_mul.c

@@ -46,11 +46,8 @@ mp_err mp_mul(const mp_int *a, const mp_int *b, mp_int *c)
    } else if (MP_HAS(S_MP_MUL_KARATSUBA) &&
               (min >= MP_MUL_KARATSUBA_CUTOFF)) {
       err = s_mp_mul_karatsuba(a, b, c);
-   } else if (MP_HAS(S_MP_MUL_COMBA) && /* can we use the fast multiplier? */
-              (min <= MP_MAX_COMBA)) {
+   } else if (MP_HAS(S_MP_MUL_COMBA)) {
       err = s_mp_mul_comba(a, b, c, digs);
-   } else if (MP_HAS(S_MP_MUL)) {
-      err = s_mp_mul(a, b, c, digs);
    } else {
       err = MP_VAL;
    }

mp_reduce.c

@@ -40,12 +40,7 @@ mp_err mp_reduce(mp_int *x, const mp_int *m, const mp_int *mu)
    if ((err = mp_mod_2d(x, MP_DIGIT_BIT * (um + 1), x)) != MP_OKAY) goto LBL_ERR;
 
    /* q = q * m mod b**(k+1), quick (no division) */
-   if (MP_HAS(S_MP_MUL_COMBA)
-       && (MP_MIN(q.used, m->used) < MP_MAX_COMBA)) {
-      if ((err = s_mp_mul_comba(&q, m, &q, um + 1)) != MP_OKAY)     goto LBL_ERR;
-   } else {
-      if ((err = s_mp_mul(&q, m, &q, um + 1)) != MP_OKAY)           goto LBL_ERR;
-   }
+   if ((err = s_mp_mul_comba(&q, m, &q, um + 1)) != MP_OKAY)        goto LBL_ERR;
 
    /* x = x - q */
    if ((err = mp_sub(x, &q, x)) != MP_OKAY)                         goto LBL_ERR;