diff options
Diffstat (limited to 'src/scalar_4x64_impl.h')
-rw-r--r-- | src/scalar_4x64_impl.h | 81 |
1 files changed, 55 insertions, 26 deletions
diff --git a/src/scalar_4x64_impl.h b/src/scalar_4x64_impl.h index ff365292f8..aa2703dd23 100644 --- a/src/scalar_4x64_impl.h +++ b/src/scalar_4x64_impl.h @@ -24,26 +24,26 @@ #define SECP256K1_N_H_2 ((uint64_t)0xFFFFFFFFFFFFFFFFULL) #define SECP256K1_N_H_3 ((uint64_t)0x7FFFFFFFFFFFFFFFULL) -SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar_t *r) { +SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar *r) { r->d[0] = 0; r->d[1] = 0; r->d[2] = 0; r->d[3] = 0; } -SECP256K1_INLINE static void secp256k1_scalar_set_int(secp256k1_scalar_t *r, unsigned int v) { +SECP256K1_INLINE static void secp256k1_scalar_set_int(secp256k1_scalar *r, unsigned int v) { r->d[0] = v; r->d[1] = 0; r->d[2] = 0; r->d[3] = 0; } -SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits(const secp256k1_scalar_t *a, unsigned int offset, unsigned int count) { +SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits(const secp256k1_scalar *a, unsigned int offset, unsigned int count) { VERIFY_CHECK((offset + count - 1) >> 6 == offset >> 6); return (a->d[offset >> 6] >> (offset & 0x3F)) & ((((uint64_t)1) << count) - 1); } -SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits_var(const secp256k1_scalar_t *a, unsigned int offset, unsigned int count) { +SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits_var(const secp256k1_scalar *a, unsigned int offset, unsigned int count) { VERIFY_CHECK(count < 32); VERIFY_CHECK(offset + count <= 256); if ((offset + count - 1) >> 6 == offset >> 6) { @@ -54,7 +54,7 @@ SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits_var(const secp256 } } -SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar_t *a) { +SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar *a) { int yes = 0; int no = 0; no |= (a->d[3] < SECP256K1_N_3); /* No need for a > check. */ @@ -66,7 +66,7 @@ SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scal return yes; } -SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar_t *r, unsigned int overflow) { +SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar *r, unsigned int overflow) { uint128_t t; VERIFY_CHECK(overflow <= 1); t = (uint128_t)r->d[0] + overflow * SECP256K1_N_C_0; @@ -80,7 +80,7 @@ SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar_t *r, unsig return overflow; } -static int secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { +static int secp256k1_scalar_add(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) { int overflow; uint128_t t = (uint128_t)a->d[0] + b->d[0]; r->d[0] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; @@ -96,9 +96,10 @@ static int secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t return overflow; } -static void secp256k1_scalar_add_bit(secp256k1_scalar_t *r, unsigned int bit) { +static void secp256k1_scalar_cadd_bit(secp256k1_scalar *r, unsigned int bit, int flag) { uint128_t t; VERIFY_CHECK(bit < 256); + bit += ((uint32_t) flag - 1) & 0x100; /* forcing (bit >> 6) > 3 makes this a noop */ t = (uint128_t)r->d[0] + (((uint64_t)((bit >> 6) == 0)) << (bit & 0x3F)); r->d[0] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; t += (uint128_t)r->d[1] + (((uint64_t)((bit >> 6) == 1)) << (bit & 0x3F)); @@ -113,7 +114,7 @@ static void secp256k1_scalar_add_bit(secp256k1_scalar_t *r, unsigned int bit) { #endif } -static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char *b32, int *overflow) { +static void secp256k1_scalar_set_b32(secp256k1_scalar *r, const unsigned char *b32, int *overflow) { int over; r->d[0] = (uint64_t)b32[31] | (uint64_t)b32[30] << 8 | (uint64_t)b32[29] << 16 | (uint64_t)b32[28] << 24 | (uint64_t)b32[27] << 32 | (uint64_t)b32[26] << 40 | (uint64_t)b32[25] << 48 | (uint64_t)b32[24] << 56; r->d[1] = (uint64_t)b32[23] | (uint64_t)b32[22] << 8 | (uint64_t)b32[21] << 16 | (uint64_t)b32[20] << 24 | (uint64_t)b32[19] << 32 | (uint64_t)b32[18] << 40 | (uint64_t)b32[17] << 48 | (uint64_t)b32[16] << 56; @@ -125,18 +126,18 @@ static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char } } -static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar_t* a) { +static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar* a) { bin[0] = a->d[3] >> 56; bin[1] = a->d[3] >> 48; bin[2] = a->d[3] >> 40; bin[3] = a->d[3] >> 32; bin[4] = a->d[3] >> 24; bin[5] = a->d[3] >> 16; bin[6] = a->d[3] >> 8; bin[7] = a->d[3]; bin[8] = a->d[2] >> 56; bin[9] = a->d[2] >> 48; bin[10] = a->d[2] >> 40; bin[11] = a->d[2] >> 32; bin[12] = a->d[2] >> 24; bin[13] = a->d[2] >> 16; bin[14] = a->d[2] >> 8; bin[15] = a->d[2]; bin[16] = a->d[1] >> 56; bin[17] = a->d[1] >> 48; bin[18] = a->d[1] >> 40; bin[19] = a->d[1] >> 32; bin[20] = a->d[1] >> 24; bin[21] = a->d[1] >> 16; bin[22] = a->d[1] >> 8; bin[23] = a->d[1]; bin[24] = a->d[0] >> 56; bin[25] = a->d[0] >> 48; bin[26] = a->d[0] >> 40; bin[27] = a->d[0] >> 32; bin[28] = a->d[0] >> 24; bin[29] = a->d[0] >> 16; bin[30] = a->d[0] >> 8; bin[31] = a->d[0]; } -SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar_t *a) { +SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar *a) { return (a->d[0] | a->d[1] | a->d[2] | a->d[3]) == 0; } -static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { +static void secp256k1_scalar_negate(secp256k1_scalar *r, const secp256k1_scalar *a) { uint64_t nonzero = 0xFFFFFFFFFFFFFFFFULL * (secp256k1_scalar_is_zero(a) == 0); uint128_t t = (uint128_t)(~a->d[0]) + SECP256K1_N_0 + 1; r->d[0] = t & nonzero; t >>= 64; @@ -148,11 +149,11 @@ static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scala r->d[3] = t & nonzero; } -SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar_t *a) { +SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar *a) { return ((a->d[0] ^ 1) | a->d[1] | a->d[2] | a->d[3]) == 0; } -static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) { +static int secp256k1_scalar_is_high(const secp256k1_scalar *a) { int yes = 0; int no = 0; no |= (a->d[3] < SECP256K1_N_H_3); @@ -164,6 +165,22 @@ static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) { return yes; } +static int secp256k1_scalar_cond_negate(secp256k1_scalar *r, int flag) { + /* If we are flag = 0, mask = 00...00 and this is a no-op; + * if we are flag = 1, mask = 11...11 and this is identical to secp256k1_scalar_negate */ + uint64_t mask = !flag - 1; + uint64_t nonzero = (secp256k1_scalar_is_zero(r) != 0) - 1; + uint128_t t = (uint128_t)(r->d[0] ^ mask) + ((SECP256K1_N_0 + 1) & mask); + r->d[0] = t & nonzero; t >>= 64; + t += (uint128_t)(r->d[1] ^ mask) + (SECP256K1_N_1 & mask); + r->d[1] = t & nonzero; t >>= 64; + t += (uint128_t)(r->d[2] ^ mask) + (SECP256K1_N_2 & mask); + r->d[2] = t & nonzero; t >>= 64; + t += (uint128_t)(r->d[3] ^ mask) + (SECP256K1_N_3 & mask); + r->d[3] = t & nonzero; + return 2 * (mask == 0) - 1; +} + /* Inspired by the macros in OpenSSL's crypto/bn/asm/x86_64-gcc.c. */ /** Add a*b to the number defined by (c0,c1,c2). c2 must never overflow. */ @@ -250,7 +267,7 @@ static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) { VERIFY_CHECK(c2 == 0); \ } -static void secp256k1_scalar_reduce_512(secp256k1_scalar_t *r, const uint64_t *l) { +static void secp256k1_scalar_reduce_512(secp256k1_scalar *r, const uint64_t *l) { #ifdef USE_ASM_X86_64 /* Reduce 512 bits into 385. */ uint64_t m0, m1, m2, m3, m4, m5, m6; @@ -559,7 +576,7 @@ static void secp256k1_scalar_reduce_512(secp256k1_scalar_t *r, const uint64_t *l secp256k1_scalar_reduce(r, c + secp256k1_scalar_check_overflow(r)); } -static void secp256k1_scalar_mul_512(uint64_t l[8], const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { +static void secp256k1_scalar_mul_512(uint64_t l[8], const secp256k1_scalar *a, const secp256k1_scalar *b) { #ifdef USE_ASM_X86_64 const uint64_t *pb = b->d; __asm__ __volatile__( @@ -721,12 +738,12 @@ static void secp256k1_scalar_mul_512(uint64_t l[8], const secp256k1_scalar_t *a, extract(l[5]); muladd_fast(a->d[3], b->d[3]); extract_fast(l[6]); - VERIFY_CHECK(c1 <= 0); + VERIFY_CHECK(c1 == 0); l[7] = c0; #endif } -static void secp256k1_scalar_sqr_512(uint64_t l[8], const secp256k1_scalar_t *a) { +static void secp256k1_scalar_sqr_512(uint64_t l[8], const secp256k1_scalar *a) { #ifdef USE_ASM_X86_64 __asm__ __volatile__( /* Preload */ @@ -871,19 +888,32 @@ static void secp256k1_scalar_sqr_512(uint64_t l[8], const secp256k1_scalar_t *a) #undef extract #undef extract_fast -static void secp256k1_scalar_mul(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { +static void secp256k1_scalar_mul(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b) { uint64_t l[8]; secp256k1_scalar_mul_512(l, a, b); secp256k1_scalar_reduce_512(r, l); } -static void secp256k1_scalar_sqr(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { +static int secp256k1_scalar_shr_int(secp256k1_scalar *r, int n) { + int ret; + VERIFY_CHECK(n > 0); + VERIFY_CHECK(n < 16); + ret = r->d[0] & ((1 << n) - 1); + r->d[0] = (r->d[0] >> n) + (r->d[1] << (64 - n)); + r->d[1] = (r->d[1] >> n) + (r->d[2] << (64 - n)); + r->d[2] = (r->d[2] >> n) + (r->d[3] << (64 - n)); + r->d[3] = (r->d[3] >> n); + return ret; +} + +static void secp256k1_scalar_sqr(secp256k1_scalar *r, const secp256k1_scalar *a) { uint64_t l[8]; secp256k1_scalar_sqr_512(l, a); secp256k1_scalar_reduce_512(r, l); } -static void secp256k1_scalar_split_128(secp256k1_scalar_t *r1, secp256k1_scalar_t *r2, const secp256k1_scalar_t *a) { +#ifdef USE_ENDOMORPHISM +static void secp256k1_scalar_split_128(secp256k1_scalar *r1, secp256k1_scalar *r2, const secp256k1_scalar *a) { r1->d[0] = a->d[0]; r1->d[1] = a->d[1]; r1->d[2] = 0; @@ -893,12 +923,13 @@ static void secp256k1_scalar_split_128(secp256k1_scalar_t *r1, secp256k1_scalar_ r2->d[2] = 0; r2->d[3] = 0; } +#endif -SECP256K1_INLINE static int secp256k1_scalar_eq(const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { +SECP256K1_INLINE static int secp256k1_scalar_eq(const secp256k1_scalar *a, const secp256k1_scalar *b) { return ((a->d[0] ^ b->d[0]) | (a->d[1] ^ b->d[1]) | (a->d[2] ^ b->d[2]) | (a->d[3] ^ b->d[3])) == 0; } -SECP256K1_INLINE static void secp256k1_scalar_mul_shift_var(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b, unsigned int shift) { +SECP256K1_INLINE static void secp256k1_scalar_mul_shift_var(secp256k1_scalar *r, const secp256k1_scalar *a, const secp256k1_scalar *b, unsigned int shift) { uint64_t l[8]; unsigned int shiftlimbs; unsigned int shiftlow; @@ -912,9 +943,7 @@ SECP256K1_INLINE static void secp256k1_scalar_mul_shift_var(secp256k1_scalar_t * r->d[1] = shift < 448 ? (l[1 + shiftlimbs] >> shiftlow | (shift < 384 && shiftlow ? (l[2 + shiftlimbs] << shifthigh) : 0)) : 0; r->d[2] = shift < 384 ? (l[2 + shiftlimbs] >> shiftlow | (shift < 320 && shiftlow ? (l[3 + shiftlimbs] << shifthigh) : 0)) : 0; r->d[3] = shift < 320 ? (l[3 + shiftlimbs] >> shiftlow) : 0; - if ((l[(shift - 1) >> 6] >> ((shift - 1) & 0x3f)) & 1) { - secp256k1_scalar_add_bit(r, 0); - } + secp256k1_scalar_cadd_bit(r, 0, (l[(shift - 1) >> 6] >> ((shift - 1) & 0x3f)) & 1); } #endif |