Add _fe_half and use in _gej_add_ge

- Trades 1 _half for 3 _mul_int and 2 _normalize_weak
- Updated formula and comments in _gej_add_ge
- Added internal benchmark for _fe_half
This commit is contained in:
Peter Dettman
2021-12-06 01:24:30 +07:00
parent d8a2463246
commit 925f78d55e
5 changed files with 172 additions and 20 deletions

View File

@@ -492,7 +492,7 @@ static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a,
static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b) {
/* Operations: 7 mul, 5 sqr, 4 normalize, 21 mul_int/add/negate/cmov */
/* Operations: 7 mul, 5 sqr, 24 add/cmov/half/mul_int/negate/normalize_weak/normalizes_to_zero */
secp256k1_fe zz, u1, u2, s1, s2, t, tt, m, n, q, rr;
secp256k1_fe m_alt, rr_alt;
int infinity, degenerate;
@@ -513,11 +513,11 @@ static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const
* Z = Z1*Z2
* T = U1+U2
* M = S1+S2
* Q = T*M^2
* Q = -T*M^2
* R = T^2-U1*U2
* X3 = 4*(R^2-Q)
* Y3 = 4*(R*(3*Q-2*R^2)-M^4)
* Z3 = 2*M*Z
* X3 = R^2+Q
* Y3 = -(R*(2*X3+Q)+M^4)/2
* Z3 = M*Z
* (Note that the paper uses xi = Xi / Zi and yi = Yi / Zi instead.)
*
* This formula has the benefit of being the same for both addition
@@ -581,7 +581,8 @@ static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const
* and denominator of lambda; R and M represent the explicit
* expressions x1^2 + x2^2 + x1x2 and y1 + y2. */
secp256k1_fe_sqr(&n, &m_alt); /* n = Malt^2 (1) */
secp256k1_fe_mul(&q, &n, &t); /* q = Q = T*Malt^2 (1) */
secp256k1_fe_negate(&q, &t, 2); /* q = -T (3) */
secp256k1_fe_mul(&q, &q, &n); /* q = Q = -T*Malt^2 (1) */
/* These two lines use the observation that either M == Malt or M == 0,
* so M^3 * Malt is either Malt^4 (which is computed by squaring), or
* zero (which is "computed" by cmov). So the cost is one squaring
@@ -589,21 +590,16 @@ static void secp256k1_gej_add_ge(secp256k1_gej *r, const secp256k1_gej *a, const
secp256k1_fe_sqr(&n, &n);
secp256k1_fe_cmov(&n, &m, degenerate); /* n = M^3 * Malt (2) */
secp256k1_fe_sqr(&t, &rr_alt); /* t = Ralt^2 (1) */
secp256k1_fe_mul(&r->z, &a->z, &m_alt); /* r->z = Malt*Z (1) */
secp256k1_fe_mul(&r->z, &a->z, &m_alt); /* r->z = Z3 = Malt*Z (1) */
infinity = secp256k1_fe_normalizes_to_zero(&r->z) & ~a->infinity;
secp256k1_fe_mul_int(&r->z, 2); /* r->z = Z3 = 2*Malt*Z (2) */
secp256k1_fe_negate(&q, &q, 1); /* q = -Q (2) */
secp256k1_fe_add(&t, &q); /* t = Ralt^2-Q (3) */
secp256k1_fe_normalize_weak(&t);
r->x = t; /* r->x = Ralt^2-Q (1) */
secp256k1_fe_mul_int(&t, 2); /* t = 2*x3 (2) */
secp256k1_fe_add(&t, &q); /* t = 2*x3 - Q: (4) */
secp256k1_fe_mul(&t, &t, &rr_alt); /* t = Ralt*(2*x3 - Q) (1) */
secp256k1_fe_add(&t, &n); /* t = Ralt*(2*x3 - Q) + M^3*Malt (3) */
secp256k1_fe_negate(&r->y, &t, 3); /* r->y = Ralt*(Q - 2x3) - M^3*Malt (4) */
secp256k1_fe_normalize_weak(&r->y);
secp256k1_fe_mul_int(&r->x, 4); /* r->x = X3 = 4*(Ralt^2-Q) */
secp256k1_fe_mul_int(&r->y, 4); /* r->y = Y3 = 4*Ralt*(Q - 2x3) - 4*M^3*Malt (4) */
secp256k1_fe_add(&t, &q); /* t = Ralt^2 + Q (2) */
r->x = t; /* r->x = X3 = Ralt^2 + Q (2) */
secp256k1_fe_mul_int(&t, 2); /* t = 2*X3 (4) */
secp256k1_fe_add(&t, &q); /* t = 2*X3 + Q (5) */
secp256k1_fe_mul(&t, &t, &rr_alt); /* t = Ralt*(2*X3 + Q) (1) */
secp256k1_fe_add(&t, &n); /* t = Ralt*(2*X3 + Q) + M^3*Malt (3) */
secp256k1_fe_negate(&r->y, &t, 3); /* r->y = -(Ralt*(2*X3 + Q) + M^3*Malt) (4) */
secp256k1_fe_half(&r->y); /* r->y = Y3 = -(Ralt*(2*X3 + Q) + M^3*Malt)/2 (3) */
/** In case a->infinity == 1, replace r with (b->x, b->y, 1). */
secp256k1_fe_cmov(&r->x, &b->x, a->infinity);