feedc0de/wolfssl
1
0
Fork 0
forked from wolfSSL/wolfssl
Code Pull Requests Activity

SP: Div fixes for constant time

Browse Source 
Implement div words for different platforms.
Use implementation not using a divide instruction unless
SP_DIV_WORD_USE_DIV or we know there is a div instruction.
Fix Thumb implementation of div word.
This commit is contained in:
Sean Parkinson
2022-05-06 16:02:18 +10:00
parent a4eaa31b5e
commit 9ffc300641
3 changed files with 2490 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1092
wolfcrypt/src/sp_armthumb.c
View File

File diff suppressed because it is too large Load Diff

661
wolfcrypt/src/sp_c32.c
View File

@ -1923,6 +1923,51 @@ static WC_INLINE sp_digit sp_2048_div_word_36(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -1944,6 +1989,15 @@ static WC_INLINE sp_digit sp_2048_div_word_36(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_2048_word_div_word_36(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -2010,7 +2064,7 @@ static int sp_2048_div_36(const sp_digit* a, const sp_digit* d,
}
t1[36 - 1] += t1[36 - 2] >> 29;
t1[36 - 2] &= 0x1fffffff;
r1 = t1[36 - 1] / dv;
r1 = sp_2048_word_div_word_36(t1[36 - 1], dv);
sp_2048_mul_d_36(t2, sd, r1);
sp_2048_sub_36(t1, t1, t2);
@ -2936,6 +2990,51 @@ static WC_INLINE sp_digit sp_2048_div_word_72(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -2957,6 +3056,15 @@ static WC_INLINE sp_digit sp_2048_div_word_72(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_2048_word_div_word_72(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -3022,7 +3130,7 @@ static int sp_2048_div_72(const sp_digit* a, const sp_digit* d,
}
t1[71 - 1] += t1[71 - 2] >> 29;
t1[71 - 2] &= 0x1fffffff;
r1 = t1[71 - 1] / dv;
r1 = sp_2048_word_div_word_72(t1[71 - 1], dv);
sp_2048_mul_d_72(t2, sd, r1);
sp_2048_sub_72(t1, t1, t2);
@ -5507,6 +5615,51 @@ static WC_INLINE sp_digit sp_3072_div_word_53(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -5528,6 +5681,15 @@ static WC_INLINE sp_digit sp_3072_div_word_53(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_3072_word_div_word_53(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -5594,7 +5756,7 @@ static int sp_3072_div_53(const sp_digit* a, const sp_digit* d,
}
t1[53 - 1] += t1[53 - 2] >> 29;
t1[53 - 2] &= 0x1fffffff;
r1 = t1[53 - 1] / dv;
r1 = sp_3072_word_div_word_53(t1[53 - 1], dv);
sp_3072_mul_d_53(t2, sd, r1);
sp_3072_sub_53(t1, t1, t2);
@ -6297,6 +6459,51 @@ static WC_INLINE sp_digit sp_3072_div_word_106(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -6318,6 +6525,15 @@ static WC_INLINE sp_digit sp_3072_div_word_106(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_3072_word_div_word_106(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -6384,7 +6600,7 @@ static int sp_3072_div_106(const sp_digit* a, const sp_digit* d,
}
t1[106 - 1] += t1[106 - 2] >> 29;
t1[106 - 2] &= 0x1fffffff;
r1 = t1[106 - 1] / dv;
r1 = sp_3072_word_div_word_106(t1[106 - 1], dv);
sp_3072_mul_d_106(t2, sd, r1);
sp_3072_sub_106(t1, t1, t2);
@ -9432,6 +9648,51 @@ static WC_INLINE sp_digit sp_3072_div_word_56(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 28) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 28;
sp_digit t0 = (sp_digit)d & 0xfffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 26; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 27);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 28);
m = d - ((sp_int64)r * div);
r += (m >> 56) - (sp_digit)(d >> 56);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 28);
m = d - ((sp_int64)r * div);
r += (m >> 56) - (sp_digit)(d >> 56);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 28) + d0;
sp_digit r = 0;
@ -9453,6 +9714,15 @@ static WC_INLINE sp_digit sp_3072_div_word_56(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_3072_word_div_word_56(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -9518,7 +9788,7 @@ static int sp_3072_div_56(const sp_digit* a, const sp_digit* d,
}
t1[55 - 1] += t1[55 - 2] >> 28;
t1[55 - 2] &= 0xfffffff;
r1 = t1[55 - 1] / dv;
r1 = sp_3072_word_div_word_56(t1[55 - 1], dv);
sp_3072_mul_d_56(t2, sd, r1);
sp_3072_sub_56(t1, t1, t2);
@ -10301,6 +10571,51 @@ static WC_INLINE sp_digit sp_3072_div_word_112(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 28) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 28;
sp_digit t0 = (sp_digit)d & 0xfffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 26; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 27);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 28);
m = d - ((sp_int64)r * div);
r += (m >> 56) - (sp_digit)(d >> 56);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 28);
m = d - ((sp_int64)r * div);
r += (m >> 56) - (sp_digit)(d >> 56);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 28) + d0;
sp_digit r = 0;
@ -10322,6 +10637,15 @@ static WC_INLINE sp_digit sp_3072_div_word_112(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_3072_word_div_word_112(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -10387,7 +10711,7 @@ static int sp_3072_div_112(const sp_digit* a, const sp_digit* d,
}
t1[110 - 1] += t1[110 - 2] >> 28;
t1[110 - 2] &= 0xfffffff;
r1 = t1[110 - 1] / dv;
r1 = sp_3072_word_div_word_112(t1[110 - 1], dv);
sp_3072_mul_d_112(t2, sd, r1);
sp_3072_sub_112(t1, t1, t2);
@ -12954,6 +13278,51 @@ static WC_INLINE sp_digit sp_4096_div_word_71(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -12975,6 +13344,15 @@ static WC_INLINE sp_digit sp_4096_div_word_71(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_4096_word_div_word_71(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -13041,7 +13419,7 @@ static int sp_4096_div_71(const sp_digit* a, const sp_digit* d,
}
t1[71 - 1] += t1[71 - 2] >> 29;
t1[71 - 2] &= 0x1fffffff;
r1 = t1[71 - 1] / dv;
r1 = sp_4096_word_div_word_71(t1[71 - 1], dv);
sp_4096_mul_d_71(t2, sd, r1);
sp_4096_sub_71(t1, t1, t2);
@ -13745,6 +14123,51 @@ static WC_INLINE sp_digit sp_4096_div_word_142(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 29;
sp_digit t0 = (sp_digit)d & 0x1fffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 27; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 28);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 29);
m = d - ((sp_int64)r * div);
r += (m >> 58) - (sp_digit)(d >> 58);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 29) + d0;
sp_digit r = 0;
@ -13766,6 +14189,15 @@ static WC_INLINE sp_digit sp_4096_div_word_142(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_4096_word_div_word_142(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -13832,7 +14264,7 @@ static int sp_4096_div_142(const sp_digit* a, const sp_digit* d,
}
t1[142 - 1] += t1[142 - 2] >> 29;
t1[142 - 2] &= 0x1fffffff;
r1 = t1[142 - 1] / dv;
r1 = sp_4096_word_div_word_142(t1[142 - 1], dv);
sp_4096_mul_d_142(t2, sd, r1);
sp_4096_sub_142(t1, t1, t2);
@ -16759,6 +17191,51 @@ static WC_INLINE sp_digit sp_4096_div_word_81(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 26) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 26;
sp_digit t0 = (sp_digit)d & 0x3ffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 24; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 25);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 26);
m = d - ((sp_int64)r * div);
r += (m >> 52) - (sp_digit)(d >> 52);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 26);
m = d - ((sp_int64)r * div);
r += (m >> 52) - (sp_digit)(d >> 52);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 26) + d0;
sp_digit r = 0;
@ -16780,6 +17257,15 @@ static WC_INLINE sp_digit sp_4096_div_word_81(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_4096_word_div_word_81(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -16845,7 +17331,7 @@ static int sp_4096_div_81(const sp_digit* a, const sp_digit* d,
}
t1[79 - 1] += t1[79 - 2] >> 26;
t1[79 - 2] &= 0x3ffffff;
r1 = t1[79 - 1] / dv;
r1 = sp_4096_word_div_word_81(t1[79 - 1], dv);
sp_4096_mul_d_81(t2, sd, r1);
sp_4096_sub_81(t1, t1, t2);
@ -17614,6 +18100,51 @@ static WC_INLINE sp_digit sp_4096_div_word_162(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 26) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 26;
sp_digit t0 = (sp_digit)d & 0x3ffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 24; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 25);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 26);
m = d - ((sp_int64)r * div);
r += (m >> 52) - (sp_digit)(d >> 52);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 26);
m = d - ((sp_int64)r * div);
r += (m >> 52) - (sp_digit)(d >> 52);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 26) + d0;
sp_digit r = 0;
@ -17635,6 +18166,15 @@ static WC_INLINE sp_digit sp_4096_div_word_162(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_4096_word_div_word_162(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -17700,7 +18240,7 @@ static int sp_4096_div_162(const sp_digit* a, const sp_digit* d,
}
t1[158 - 1] += t1[158 - 2] >> 26;
t1[158 - 2] &= 0x3ffffff;
r1 = t1[158 - 1] / dv;
r1 = sp_4096_word_div_word_162(t1[158 - 1], dv);
sp_4096_mul_d_162(t2, sd, r1);
sp_4096_sub_162(t1, t1, t2);
@ -40524,6 +41064,51 @@ static WC_INLINE sp_digit sp_521_div_word_21(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 25) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 25;
sp_digit t0 = (sp_digit)d & 0x1ffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 23; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 24);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 25);
m = d - ((sp_int64)r * div);
r += (m >> 50) - (sp_digit)(d >> 50);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 25);
m = d - ((sp_int64)r * div);
r += (m >> 50) - (sp_digit)(d >> 50);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 25) + d0;
sp_digit r = 0;
@ -43142,6 +43727,51 @@ static WC_INLINE sp_digit sp_1024_div_word_42(sp_digit d1, sp_digit d0,
);
return (sp_digit)lo;
#elif !defined(__aarch64__) && !defined(SP_DIV_WORD_USE_DIV)
sp_int64 d = ((sp_int64)d1 << 25) + d0;
sp_digit dv = (div >> 1) + 1;
sp_digit t1 = d >> 25;
sp_digit t0 = (sp_digit)d & 0x1ffffff;
sp_digit t2;
sp_digit sign;
sp_digit r;
int i;
sp_int64 m;
r = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
t1 -= dv & (0 - r);
for (i = 23; i >= 1; i--) {
t1 += t1 + ((sp_uint32)t0 >> 24);
t0 <<= 1;
t2 = (sp_digit)(((sp_uint32)(dv - t1)) >> 31);
r += r + t2;
t1 -= dv & (0 - t2);
t1 += t2;
}
r += r + 1;
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 25);
m = d - ((sp_int64)r * div);
r += (m >> 50) - (sp_digit)(d >> 50);
m = d - ((sp_int64)r * div);
r += (sp_digit)(m >> 25);
m = d - ((sp_int64)r * div);
r += (m >> 50) - (sp_digit)(d >> 50);
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
m = d - ((sp_int64)r * div);
sign = (sp_digit)(0 - ((sp_uint32)m >> 31)) * 2 + 1;
m *= sign;
t2 = (sp_digit)(((sp_uint32)(div - m)) >> 31);
r += sign * t2;
return r;
#else
sp_int64 d = ((sp_int64)d1 << 25) + d0;
sp_digit r = 0;
@ -43163,6 +43793,15 @@ static WC_INLINE sp_digit sp_1024_div_word_42(sp_digit d1, sp_digit d0,
return r;
#endif
}
static WC_INLINE sp_digit sp_1024_word_div_word_42(sp_digit d, sp_digit div)
{
#if defined(__x86_64__) || defined(__i386__) || defined(__aarch64__) || \
defined(SP_DIV_WORD_USE_DIV)
return d / div;
#else
return (sp_digit)((sp_uint32)(div - d) >> 31);
#endif
}
/* Divide d in a and put remainder into r (m*d + r = a)
* m is not calculated as it is not needed at this time.
*
@ -43228,7 +43867,7 @@ static int sp_1024_div_42(const sp_digit* a, const sp_digit* d,
}
t1[41 - 1] += t1[41 - 2] >> 25;
t1[41 - 2] &= 0x1ffffff;
r1 = t1[41 - 1] / dv;
r1 = sp_1024_word_div_word_42(t1[41 - 1], dv);
sp_1024_mul_d_42(t2, sd, r1);
sp_1024_sub_42(t1, t1, t2);

773
wolfcrypt/src/sp_c64.c
View File

File diff suppressed because it is too large Load Diff