abc/GmSSL
1
0
Fork 0
mirror of https://github.com/guanzhi/GmSSL.git synced 2026-08-07 21:53:41 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update sm9_z256_alg.c

Browse Source
This commit is contained in:
Zhi Guan
2024-03-31 16:54:36 +08:00
parent 6b36c51cdf
commit d8bd8993d2
1 changed files with 123 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

137
src/sm9_z256_alg.c
View File

@@ -612,6 +612,7 @@ void sm9_z256_fp_mont_mul_2way(sm9_z256_t r, const sm9_z256_t a, const sm9_z256_
} }
#endif #endif
#ifndef ENABLE_SM9_Z256_ARMV8 #ifndef ENABLE_SM9_Z256_ARMV8
void sm9_z256_fp_to_mont(sm9_z256_t r, const sm9_z256_t a) void sm9_z256_fp_to_mont(sm9_z256_t r, const sm9_z256_t a)
{ {
@@ -622,15 +623,17 @@ void sm9_z256_fp_from_mont(sm9_z256_t r, const sm9_z256_t a)
{ {
sm9_z256_fp_mont_mul(r, a, SM9_Z256_ONE); sm9_z256_fp_mont_mul(r, a, SM9_Z256_ONE);
} }
#endif
// ASM
void sm9_z256_fp_mont_sqr(sm9_z256_t r, const sm9_z256_t a) void sm9_z256_fp_mont_sqr(sm9_z256_t r, const sm9_z256_t a)
{ {
sm9_z256_fp_mont_mul(r, a, a); sm9_z256_fp_mont_mul(r, a, a);
} }
#endif
// change args name to a_mont, r_mont // change args name to a_mont, r_mont
// 这个函数反复多次调用mont_mul，如果展开为asm，可以节约一些开销
// 也可以不展开这个函数，只是展开inv那个函数
// 所有256次调用可以完全展开，展开之后不需要判断w，直接就做计算就可以了，好处就是所有的初始化都不需要了
void sm9_z256_fp_pow(sm9_z256_t r, const sm9_z256_t a, const sm9_z256_t e) void sm9_z256_fp_pow(sm9_z256_t r, const sm9_z256_t a, const sm9_z256_t e)
{ {
sm9_z256_t t; sm9_z256_t t;
@@ -823,49 +826,65 @@ void sm9_z256_fp2_a_mul_u(sm9_z256_fp2 r, sm9_z256_fp2 a)
sm9_z256_fp_copy(r[0], r0); sm9_z256_fp_copy(r[0], r0);
} }
void sm9_z256_fp2_mul(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_fp2 b) void sm9_z256_fp2_mul(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_fp2 b)
{ {
sm9_z256_t r0, r1, t; sm9_z256_t t0;
sm9_z256_t t1;
sm9_z256_t t2;
sm9_z256_fp_add(r0, a[0], a[1]); // t2 = (a0 + a1) * (b0 + b1)
sm9_z256_fp_add(t, b[0], b[1]); sm9_z256_fp_add(t0, a[0], a[1]);
sm9_z256_fp_mul(r1, t, r0); sm9_z256_fp_add(t1, b[0], b[1]);
sm9_z256_fp_mul(t2, t0, t1);
// r0 = a0 * b0 - 2 * a1 * b1 // t0 = a0 * b0
sm9_z256_fp_mul(r0, a[0], b[0]); sm9_z256_fp_mul(t0, a[0], b[0]);
sm9_z256_fp_mul(t, a[1], b[1]);
// r1 = (a0 + a1) * (b0 + b1) - a0 * b0 - a1 * b1 // t1 = a1 * b1
sm9_z256_fp_sub(r1, r1, r0); sm9_z256_fp_mul(t1, a[1], b[1]);
sm9_z256_fp_sub(r1, r1, t);
// r0 // r1 = t2 - t0 - t1 = a0 * b1 + a1 * b0
sm9_z256_fp_dbl(t, t); sm9_z256_fp_sub(t2, t2, t0);
sm9_z256_fp_sub(r0, r0, t); sm9_z256_fp_sub(t2, t2, t1);
sm9_z256_fp_copy(r[0], r0); // r0 = t0 - 2*t1 = a0 * b0 - 2(a1 * b1)
sm9_z256_fp_copy(r[1], r1); sm9_z256_fp_dbl(t1, t1);
sm9_z256_fp_sub(t0, t0, t1);
sm9_z256_fp_copy(r[0], t0);
sm9_z256_fp_copy(r[1], t2);
} }
void sm9_z256_fp2_mul_u(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_fp2 b) void sm9_z256_fp2_mul_u(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_fp2 b)
{ {
sm9_z256_t r0, r1, t; sm9_z256_t t0;
sm9_z256_t t1;
sm9_z256_t t2;
// r0 = -2 * (a0 * b1 + a1 * b0) // t2 = (a0 + a1) * (b0 + b1)
sm9_z256_fp_mul(r0, a[0], b[1]); sm9_z256_fp_add(t0, a[0], a[1]);
sm9_z256_fp_mul(t, a[1], b[0]); sm9_z256_fp_add(t1, b[0], b[1]);
sm9_z256_fp_add(r0, r0, t); sm9_z256_fp_mul(t2, t0, t1);
sm9_z256_fp_dbl(r0, r0);
sm9_z256_fp_neg(r0, r0);
// r1 = a0 * b0 - 2 * a1 * b1 // t0 = a0 * b0
sm9_z256_fp_mul(r1, a[0], b[0]); sm9_z256_fp_mul(t0, a[0], b[0]);
sm9_z256_fp_mul(t, a[1], b[1]);
sm9_z256_fp_dbl(t, t);
sm9_z256_fp_sub(r1, r1, t);
sm9_z256_fp_copy(r[0], r0); // t1 = a1 * b1
sm9_z256_fp_copy(r[1], r1); sm9_z256_fp_mul(t1, a[1], b[1]);
// r0 = -2 *(t2 - t0 - t1) = -2 * (a0 * b1 + a1 * b0)
sm9_z256_fp_sub(t2, t2, t0);
sm9_z256_fp_sub(t2, t2, t1);
sm9_z256_fp_dbl(t2, t2);
sm9_z256_fp_neg(t2, t2);
// r1 = t0 - 2*t1 = a0 * b0 - 2(a1 * b1)
sm9_z256_fp_dbl(t1, t1);
sm9_z256_fp_sub(t0, t0, t1);
sm9_z256_fp_copy(r[0], t2);
sm9_z256_fp_copy(r[1], t0);
} }
void sm9_z256_fp2_mul_fp(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_t k) void sm9_z256_fp2_mul_fp(sm9_z256_fp2 r, const sm9_z256_fp2 a, const sm9_z256_t k)
@@ -895,22 +914,31 @@ void sm9_z256_fp2_sqr(sm9_z256_fp2 r, const sm9_z256_fp2 a)
void sm9_z256_fp2_sqr_u(sm9_z256_fp2 r, const sm9_z256_fp2 a) void sm9_z256_fp2_sqr_u(sm9_z256_fp2 r, const sm9_z256_fp2 a)
{ {
sm9_z256_t r0, r1, t; sm9_z256_t t0;
sm9_z256_t t1;
sm9_z256_t t2;
// r0 = -4 * a0 * a1 // t0 = a0 * a1
sm9_z256_fp_mul(r0, a[0], a[1]); sm9_z256_fp_mont_mul(t0, a[0], a[1]);
sm9_z256_fp_dbl(r0, r0);
sm9_z256_fp_dbl(r0, r0);
sm9_z256_fp_neg(r0, r0);
// r1 = a0^2 - 2 * a1^2 // t1 = a0 + a1
sm9_z256_fp_sqr(r1, a[0]); sm9_z256_fp_add(t1, a[0], a[1]);
sm9_z256_fp_sqr(t, a[1]);
sm9_z256_fp_dbl(t, t);
sm9_z256_fp_sub(r1, r1, t);
sm9_z256_fp_copy(r[0], r0); // t2 = a0 - 2*a
sm9_z256_fp_copy(r[1], r1); sm9_z256_fp_sub(t2, a[0], a[1]);
sm9_z256_fp_sub(t2, t2, a[1]);
// r1 = t1 * t2 + t0
sm9_z256_fp_mont_mul(t2, t2, t1);
sm9_z256_fp_add(t2, t2, t0);
// r0 = -4 * t0
sm9_z256_fp_dbl(t0, t0);
sm9_z256_fp_dbl(t0, t0);
sm9_z256_fp_neg(t0, t0);
sm9_z256_copy(r[0], t0);
sm9_z256_copy(r[1], t2);
} }
void sm9_z256_fp2_inv(sm9_z256_fp2 r, const sm9_z256_fp2 a) void sm9_z256_fp2_inv(sm9_z256_fp2 r, const sm9_z256_fp2 a)
@@ -1064,17 +1092,29 @@ void sm9_z256_fp4_mul(sm9_z256_fp4 r, const sm9_z256_fp4 a, const sm9_z256_fp4 b
{ {
sm9_z256_fp2 r0, r1, t; sm9_z256_fp2 r0, r1, t;
// r0 = a0 + a1
sm9_z256_fp2_add(r0, a[0], a[1]); sm9_z256_fp2_add(r0, a[0], a[1]);
// t = b0 + b1
sm9_z256_fp2_add(t, b[0], b[1]); sm9_z256_fp2_add(t, b[0], b[1]);
// r1 = (a0 + a1) * (b0 + b1)
sm9_z256_fp2_mul(r1, t, r0); sm9_z256_fp2_mul(r1, t, r0);
// r0 = a0 * b0
sm9_z256_fp2_mul(r0, a[0], b[0]); sm9_z256_fp2_mul(r0, a[0], b[0]);
// t = a1 * b1
sm9_z256_fp2_mul(t, a[1], b[1]); sm9_z256_fp2_mul(t, a[1], b[1]);
// r1 = a0 * b1 + a1 * b0
sm9_z256_fp2_sub(r1, r1, r0); sm9_z256_fp2_sub(r1, r1, r0);
sm9_z256_fp2_sub(r1, r1, t); sm9_z256_fp2_sub(r1, r1, t);
// t = a1 * b1 * u
sm9_z256_fp2_a_mul_u(t, t); sm9_z256_fp2_a_mul_u(t, t);
// r0 = a0 * b0 + a1 * b1 * u
sm9_z256_fp2_add(r0, r0, t); sm9_z256_fp2_add(r0, r0, t);
sm9_z256_fp2_copy(r[0], r0); sm9_z256_fp2_copy(r[0], r0);
@@ -1691,6 +1731,7 @@ int sm9_z256_point_is_on_curve(const SM9_Z256_POINT *P)
return 1; return 1;
} }
// E(F_p): y^2 = x^3 + b， 计算公式和SM2不同
void sm9_z256_point_dbl(SM9_Z256_POINT *R, const SM9_Z256_POINT *P) void sm9_z256_point_dbl(SM9_Z256_POINT *R, const SM9_Z256_POINT *P)
{ {
const uint64_t *X1 = P->X; const uint64_t *X1 = P->X;
@@ -2018,6 +2059,8 @@ void sm9_z256_twist_point_neg(SM9_Z256_TWIST_POINT *R, const SM9_Z256_TWIST_POIN
sm9_z256_fp2_copy(R->Z, P->Z); sm9_z256_fp2_copy(R->Z, P->Z);
} }
// E(Fp^2)的计算也比较重要，但是fp2上的计算并不容易做2路并发
void sm9_z256_twist_point_dbl(SM9_Z256_TWIST_POINT *R, const SM9_Z256_TWIST_POINT *P) void sm9_z256_twist_point_dbl(SM9_Z256_TWIST_POINT *R, const SM9_Z256_TWIST_POINT *P)
{ {
const sm9_z256_t *X1 = P->X; const sm9_z256_t *X1 = P->X;
@@ -2369,6 +2412,8 @@ void sm9_z256_final_exponent(sm9_z256_fp12 r, const sm9_z256_fp12 f)
sm9_z256_fp12_copy(r, t0); sm9_z256_fp12_copy(r, t0);
} }
// 这个计算是否有更快速的算法
// 特别是主循环中的计算时否需要再Fp12上面
void sm9_z256_pairing(sm9_z256_fp12 r, const SM9_Z256_TWIST_POINT *Q, const SM9_Z256_POINT *P) void sm9_z256_pairing(sm9_z256_fp12 r, const SM9_Z256_TWIST_POINT *Q, const SM9_Z256_POINT *P)
{ {
const char *abits = "00100000000000000000000000000000000000010000101100020200101000020"; const char *abits = "00100000000000000000000000000000000000010000101100020200101000020";
@@ -2457,7 +2502,7 @@ void sm9_z256_fn_sub(sm9_z256_t r, const sm9_z256_t a, const sm9_z256_t b)
} }
} }
void sm9_z256_mul_5(uint64_t r[10], const uint64_t a[5], const uint64_t b[5]) void sm9_z320_mul(uint64_t r[10], const uint64_t a[5], const uint64_t b[5])
{ {
uint64_t a_[10]; uint64_t a_[10];
uint64_t b_[10]; uint64_t b_[10];
@@ -2499,7 +2544,7 @@ void sm9_z256_fn_mul(sm9_z256_t r, const sm9_z256_t a, const sm9_z256_t b)
sm9_z256_mul(z, a, b); sm9_z256_mul(z, a, b);
// (z // 2^192) = z[3-7] // (z // 2^192) = z[3-7]
sm9_z256_mul_5(h, z + 3, SM9_Z256_N_BARRETT_MU); sm9_z320_mul(h, z + 3, SM9_Z256_N_BARRETT_MU);
// (h // 2^320) = h[5-9] // (h // 2^320) = h[5-9]
sm9_z256_mul(s, h + 5, SM9_Z256_N); sm9_z256_mul(s, h + 5, SM9_Z256_N);