From 436281afdcb68991395f97338197d208212965e2 Mon Sep 17 00:00:00 2001 From: Pieter Wuille Date: Sun, 11 Oct 2020 15:41:54 -0700 Subject: [PATCH] Move secp256k1_fe_inverse{_var} to per-impl files This temporarily duplicates the inversion code across the 5x52 and 10x26 implementations. Those implementations will be replaced in a next commit. --- src/field_10x26_impl.h | 127 +++++++++++++++++++++++++++++++++++++++++ src/field_5x52_impl.h | 127 +++++++++++++++++++++++++++++++++++++++++ src/field_impl.h | 127 ----------------------------------------- 3 files changed, 254 insertions(+), 127 deletions(-) diff --git a/src/field_10x26_impl.h b/src/field_10x26_impl.h index 62bffdc21bd44..3539d5b89fd0d 100644 --- a/src/field_10x26_impl.h +++ b/src/field_10x26_impl.h @@ -1164,4 +1164,131 @@ static SECP256K1_INLINE void secp256k1_fe_from_storage(secp256k1_fe *r, const se #endif } +static void secp256k1_fe_inv(secp256k1_fe *r, const secp256k1_fe *a) { + secp256k1_fe x2, x3, x6, x9, x11, x22, x44, x88, x176, x220, x223, t1; + int j; + + /** The binary representation of (p - 2) has 5 blocks of 1s, with lengths in + * { 1, 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block: + * [1], [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223] + */ + + secp256k1_fe_sqr(&x2, a); + secp256k1_fe_mul(&x2, &x2, a); + + secp256k1_fe_sqr(&x3, &x2); + secp256k1_fe_mul(&x3, &x3, a); + + x6 = x3; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x6, &x6); + } + secp256k1_fe_mul(&x6, &x6, &x3); + + x9 = x6; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x9, &x9); + } + secp256k1_fe_mul(&x9, &x9, &x3); + + x11 = x9; + for (j=0; j<2; j++) { + secp256k1_fe_sqr(&x11, &x11); + } + secp256k1_fe_mul(&x11, &x11, &x2); + + x22 = x11; + for (j=0; j<11; j++) { + secp256k1_fe_sqr(&x22, &x22); + } + secp256k1_fe_mul(&x22, &x22, &x11); + + x44 = x22; + for (j=0; j<22; j++) { + secp256k1_fe_sqr(&x44, &x44); + } + secp256k1_fe_mul(&x44, &x44, &x22); + + x88 = x44; + for (j=0; j<44; j++) { + secp256k1_fe_sqr(&x88, &x88); + } + secp256k1_fe_mul(&x88, &x88, &x44); + + x176 = x88; + for (j=0; j<88; j++) { + secp256k1_fe_sqr(&x176, &x176); + } + secp256k1_fe_mul(&x176, &x176, &x88); + + x220 = x176; + for (j=0; j<44; j++) { + secp256k1_fe_sqr(&x220, &x220); + } + secp256k1_fe_mul(&x220, &x220, &x44); + + x223 = x220; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x223, &x223); + } + secp256k1_fe_mul(&x223, &x223, &x3); + + /* The final result is then assembled using a sliding window over the blocks. */ + + t1 = x223; + for (j=0; j<23; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, &x22); + for (j=0; j<5; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, a); + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, &x2); + for (j=0; j<2; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(r, a, &t1); +} + +static void secp256k1_fe_inv_var(secp256k1_fe *r, const secp256k1_fe *a) { +#if defined(USE_FIELD_INV_BUILTIN) + secp256k1_fe_inv(r, a); +#elif defined(USE_FIELD_INV_NUM) + secp256k1_num n, m; + static const secp256k1_fe negone = SECP256K1_FE_CONST( + 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, + 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFEUL, 0xFFFFFC2EUL + ); + /* secp256k1 field prime, value p defined in "Standards for Efficient Cryptography" (SEC2) 2.7.1. */ + static const unsigned char prime[32] = { + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F + }; + unsigned char b[32]; + int res; + secp256k1_fe c = *a; + secp256k1_fe_normalize_var(&c); + secp256k1_fe_get_b32(b, &c); + secp256k1_num_set_bin(&n, b, 32); + secp256k1_num_set_bin(&m, prime, 32); + secp256k1_num_mod_inverse(&n, &n, &m); + secp256k1_num_get_bin(b, 32, &n); + res = secp256k1_fe_set_b32(r, b); + (void)res; + VERIFY_CHECK(res); + /* Verify the result is the (unique) valid inverse using non-GMP code. */ + secp256k1_fe_mul(&c, &c, r); + secp256k1_fe_add(&c, &negone); + CHECK(secp256k1_fe_normalizes_to_zero_var(&c)); +#else +#error "Please select field inverse implementation" +#endif +} + #endif /* SECP256K1_FIELD_REPR_IMPL_H */ diff --git a/src/field_5x52_impl.h b/src/field_5x52_impl.h index 3465ea3247b42..b5645674934a3 100644 --- a/src/field_5x52_impl.h +++ b/src/field_5x52_impl.h @@ -498,4 +498,131 @@ static SECP256K1_INLINE void secp256k1_fe_from_storage(secp256k1_fe *r, const se #endif } +static void secp256k1_fe_inv(secp256k1_fe *r, const secp256k1_fe *a) { + secp256k1_fe x2, x3, x6, x9, x11, x22, x44, x88, x176, x220, x223, t1; + int j; + + /** The binary representation of (p - 2) has 5 blocks of 1s, with lengths in + * { 1, 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block: + * [1], [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223] + */ + + secp256k1_fe_sqr(&x2, a); + secp256k1_fe_mul(&x2, &x2, a); + + secp256k1_fe_sqr(&x3, &x2); + secp256k1_fe_mul(&x3, &x3, a); + + x6 = x3; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x6, &x6); + } + secp256k1_fe_mul(&x6, &x6, &x3); + + x9 = x6; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x9, &x9); + } + secp256k1_fe_mul(&x9, &x9, &x3); + + x11 = x9; + for (j=0; j<2; j++) { + secp256k1_fe_sqr(&x11, &x11); + } + secp256k1_fe_mul(&x11, &x11, &x2); + + x22 = x11; + for (j=0; j<11; j++) { + secp256k1_fe_sqr(&x22, &x22); + } + secp256k1_fe_mul(&x22, &x22, &x11); + + x44 = x22; + for (j=0; j<22; j++) { + secp256k1_fe_sqr(&x44, &x44); + } + secp256k1_fe_mul(&x44, &x44, &x22); + + x88 = x44; + for (j=0; j<44; j++) { + secp256k1_fe_sqr(&x88, &x88); + } + secp256k1_fe_mul(&x88, &x88, &x44); + + x176 = x88; + for (j=0; j<88; j++) { + secp256k1_fe_sqr(&x176, &x176); + } + secp256k1_fe_mul(&x176, &x176, &x88); + + x220 = x176; + for (j=0; j<44; j++) { + secp256k1_fe_sqr(&x220, &x220); + } + secp256k1_fe_mul(&x220, &x220, &x44); + + x223 = x220; + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&x223, &x223); + } + secp256k1_fe_mul(&x223, &x223, &x3); + + /* The final result is then assembled using a sliding window over the blocks. */ + + t1 = x223; + for (j=0; j<23; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, &x22); + for (j=0; j<5; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, a); + for (j=0; j<3; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(&t1, &t1, &x2); + for (j=0; j<2; j++) { + secp256k1_fe_sqr(&t1, &t1); + } + secp256k1_fe_mul(r, a, &t1); +} + +static void secp256k1_fe_inv_var(secp256k1_fe *r, const secp256k1_fe *a) { +#if defined(USE_FIELD_INV_BUILTIN) + secp256k1_fe_inv(r, a); +#elif defined(USE_FIELD_INV_NUM) + secp256k1_num n, m; + static const secp256k1_fe negone = SECP256K1_FE_CONST( + 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, + 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFEUL, 0xFFFFFC2EUL + ); + /* secp256k1 field prime, value p defined in "Standards for Efficient Cryptography" (SEC2) 2.7.1. */ + static const unsigned char prime[32] = { + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F + }; + unsigned char b[32]; + int res; + secp256k1_fe c = *a; + secp256k1_fe_normalize_var(&c); + secp256k1_fe_get_b32(b, &c); + secp256k1_num_set_bin(&n, b, 32); + secp256k1_num_set_bin(&m, prime, 32); + secp256k1_num_mod_inverse(&n, &n, &m); + secp256k1_num_get_bin(b, 32, &n); + res = secp256k1_fe_set_b32(r, b); + (void)res; + VERIFY_CHECK(res); + /* Verify the result is the (unique) valid inverse using non-GMP code. */ + secp256k1_fe_mul(&c, &c, r); + secp256k1_fe_add(&c, &negone); + CHECK(secp256k1_fe_normalizes_to_zero_var(&c)); +#else +#error "Please select field inverse implementation" +#endif +} + #endif /* SECP256K1_FIELD_REPR_IMPL_H */ diff --git a/src/field_impl.h b/src/field_impl.h index f0096f63122fc..7b75e98601f0c 100644 --- a/src/field_impl.h +++ b/src/field_impl.h @@ -136,133 +136,6 @@ static int secp256k1_fe_sqrt(secp256k1_fe *r, const secp256k1_fe *a) { return secp256k1_fe_equal(&t1, a); } -static void secp256k1_fe_inv(secp256k1_fe *r, const secp256k1_fe *a) { - secp256k1_fe x2, x3, x6, x9, x11, x22, x44, x88, x176, x220, x223, t1; - int j; - - /** The binary representation of (p - 2) has 5 blocks of 1s, with lengths in - * { 1, 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block: - * [1], [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223] - */ - - secp256k1_fe_sqr(&x2, a); - secp256k1_fe_mul(&x2, &x2, a); - - secp256k1_fe_sqr(&x3, &x2); - secp256k1_fe_mul(&x3, &x3, a); - - x6 = x3; - for (j=0; j<3; j++) { - secp256k1_fe_sqr(&x6, &x6); - } - secp256k1_fe_mul(&x6, &x6, &x3); - - x9 = x6; - for (j=0; j<3; j++) { - secp256k1_fe_sqr(&x9, &x9); - } - secp256k1_fe_mul(&x9, &x9, &x3); - - x11 = x9; - for (j=0; j<2; j++) { - secp256k1_fe_sqr(&x11, &x11); - } - secp256k1_fe_mul(&x11, &x11, &x2); - - x22 = x11; - for (j=0; j<11; j++) { - secp256k1_fe_sqr(&x22, &x22); - } - secp256k1_fe_mul(&x22, &x22, &x11); - - x44 = x22; - for (j=0; j<22; j++) { - secp256k1_fe_sqr(&x44, &x44); - } - secp256k1_fe_mul(&x44, &x44, &x22); - - x88 = x44; - for (j=0; j<44; j++) { - secp256k1_fe_sqr(&x88, &x88); - } - secp256k1_fe_mul(&x88, &x88, &x44); - - x176 = x88; - for (j=0; j<88; j++) { - secp256k1_fe_sqr(&x176, &x176); - } - secp256k1_fe_mul(&x176, &x176, &x88); - - x220 = x176; - for (j=0; j<44; j++) { - secp256k1_fe_sqr(&x220, &x220); - } - secp256k1_fe_mul(&x220, &x220, &x44); - - x223 = x220; - for (j=0; j<3; j++) { - secp256k1_fe_sqr(&x223, &x223); - } - secp256k1_fe_mul(&x223, &x223, &x3); - - /* The final result is then assembled using a sliding window over the blocks. */ - - t1 = x223; - for (j=0; j<23; j++) { - secp256k1_fe_sqr(&t1, &t1); - } - secp256k1_fe_mul(&t1, &t1, &x22); - for (j=0; j<5; j++) { - secp256k1_fe_sqr(&t1, &t1); - } - secp256k1_fe_mul(&t1, &t1, a); - for (j=0; j<3; j++) { - secp256k1_fe_sqr(&t1, &t1); - } - secp256k1_fe_mul(&t1, &t1, &x2); - for (j=0; j<2; j++) { - secp256k1_fe_sqr(&t1, &t1); - } - secp256k1_fe_mul(r, a, &t1); -} - -static void secp256k1_fe_inv_var(secp256k1_fe *r, const secp256k1_fe *a) { -#if defined(USE_FIELD_INV_BUILTIN) - secp256k1_fe_inv(r, a); -#elif defined(USE_FIELD_INV_NUM) - secp256k1_num n, m; - static const secp256k1_fe negone = SECP256K1_FE_CONST( - 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFFUL, - 0xFFFFFFFFUL, 0xFFFFFFFFUL, 0xFFFFFFFEUL, 0xFFFFFC2EUL - ); - /* secp256k1 field prime, value p defined in "Standards for Efficient Cryptography" (SEC2) 2.7.1. */ - static const unsigned char prime[32] = { - 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, - 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, - 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, - 0xFF,0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F - }; - unsigned char b[32]; - int res; - secp256k1_fe c = *a; - secp256k1_fe_normalize_var(&c); - secp256k1_fe_get_b32(b, &c); - secp256k1_num_set_bin(&n, b, 32); - secp256k1_num_set_bin(&m, prime, 32); - secp256k1_num_mod_inverse(&n, &n, &m); - secp256k1_num_get_bin(b, 32, &n); - res = secp256k1_fe_set_b32(r, b); - (void)res; - VERIFY_CHECK(res); - /* Verify the result is the (unique) valid inverse using non-GMP code. */ - secp256k1_fe_mul(&c, &c, r); - secp256k1_fe_add(&c, &negone); - CHECK(secp256k1_fe_normalizes_to_zero_var(&c)); -#else -#error "Please select field inverse implementation" -#endif -} - static int secp256k1_fe_is_quad_var(const secp256k1_fe *a) { #ifndef USE_NUM_NONE unsigned char b[32];