Squashed 'src/secp256k1/' changes from 8746600ee..44c2452fd

44c2452fd Merge bitcoin-core/secp256k1#1105: Don't export symbols in static libraries
6f6cab998 abi: Don't export symbols in static Windows libraries
485f608fa Merge bitcoin-core/secp256k1#1104: Fix the false positive of `SECP_64BIT_ASM_CHECK`
8b013fce5 Merge bitcoin-core/secp256k1#1056: Save negations in var-time group addition
7efc9835a Fix the false positive of `SECP_64BIT_ASM_CHECK`
2f984ffc4 Save negations in var-time group addition

git-subtree-dir: src/secp256k1
git-subtree-split: 44c2452fd387f7ca604ab42d73746e7d3a44d8a2

2 files changed, 71 insertions, 38 deletions

diff --git a/src/bench_internal.c b/src/bench_internal.c
index 3c145f306c..7eb3af28d7 100644
--- a/src/bench_internal.c
+++ b/src/bench_internal.c
@@ -254,6 +254,15 @@ void bench_group_add_affine_var(void* arg, int iters) {
     }
 }
 
+void bench_group_add_zinv_var(void* arg, int iters) {
+    int i;
+    bench_inv *data = (bench_inv*)arg;
+
+    for (i = 0; i < iters; i++) {
+        secp256k1_gej_add_zinv_var(&data->gej[0], &data->gej[0], &data->ge[1], &data->gej[0].y);
+    }
+}
+
 void bench_group_to_affine_var(void* arg, int iters) {
     int i;
     bench_inv *data = (bench_inv*)arg;
@@ -376,6 +385,7 @@ int main(int argc, char **argv) {
     if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_var", bench_group_add_var, bench_setup, NULL, &data, 10, iters*10);
     if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_affine", bench_group_add_affine, bench_setup, NULL, &data, 10, iters*10);
     if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_affine_var", bench_group_add_affine_var, bench_setup, NULL, &data, 10, iters*10);
+    if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "add")) run_benchmark("group_add_zinv_var", bench_group_add_zinv_var, bench_setup, NULL, &data, 10, iters*10);
     if (d || have_flag(argc, argv, "group") || have_flag(argc, argv, "to_affine")) run_benchmark("group_to_affine_var", bench_group_to_affine_var, bench_setup, NULL, &data, 10, iters);
 
     if (d || have_flag(argc, argv, "ecmult") || have_flag(argc, argv, "wnaf")) run_benchmark("wnaf_const", bench_wnaf_const, bench_setup, NULL, &data, 10, iters);
diff --git a/src/group_impl.h b/src/group_impl.h
index b19b02a01f..63735ab682 100644
--- a/src/group_impl.h
+++ b/src/group_impl.h
@@ -330,15 +330,14 @@ static void secp256k1_gej_double_var(secp256k1_gej *r, const secp256k1_gej *a, s
 }
 
 static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_gej *b, secp256k1_fe *rzr) {
-    /* Operations: 12 mul, 4 sqr, 2 normalize, 12 mul_int/add/negate */
-    secp256k1_fe z22, z12, u1, u2, s1, s2, h, i, i2, h2, h3, t;
+    /* 12 mul, 4 sqr, 11 add/negate/normalizes_to_zero (ignoring special cases) */
+    secp256k1_fe z22, z12, u1, u2, s1, s2, h, i, h2, h3, t;
 
     if (a->infinity) {
         VERIFY_CHECK(rzr == NULL);
         *r = *b;
         return;
     }
-
     if (b->infinity) {
         if (rzr != NULL) {
             secp256k1_fe_set_int(rzr, 1);
@@ -347,7 +346,6 @@ static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, cons
         return;
     }
 
-    r->infinity = 0;
     secp256k1_fe_sqr(&z22, &b->z);
     secp256k1_fe_sqr(&z12, &a->z);
     secp256k1_fe_mul(&u1, &a->x, &z22);
@@ -355,7 +353,7 @@ static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, cons
     secp256k1_fe_mul(&s1, &a->y, &z22); secp256k1_fe_mul(&s1, &s1, &b->z);
     secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z);
     secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2);
-    secp256k1_fe_negate(&i, &s1, 1); secp256k1_fe_add(&i, &s2);
+    secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
     if (secp256k1_fe_normalizes_to_zero_var(&h)) {
         if (secp256k1_fe_normalizes_to_zero_var(&i)) {
             secp256k1_gej_double_var(r, a, rzr);
@@ -367,24 +365,33 @@ static void secp256k1_gej_add_var(secp256k1_gej *r, const secp256k1_gej *a, cons
         }
         return;
     }
-    secp256k1_fe_sqr(&i2, &i);
-    secp256k1_fe_sqr(&h2, &h);
-    secp256k1_fe_mul(&h3, &h, &h2);
-    secp256k1_fe_mul(&h, &h, &b->z);
+
+    r->infinity = 0;
+    secp256k1_fe_mul(&t, &h, &b->z);
     if (rzr != NULL) {
-        *rzr = h;
+        *rzr = t;
     }
-    secp256k1_fe_mul(&r->z, &a->z, &h);
+    secp256k1_fe_mul(&r->z, &a->z, &t);
+
+    secp256k1_fe_sqr(&h2, &h);
+    secp256k1_fe_negate(&h2, &h2, 1);
+    secp256k1_fe_mul(&h3, &h2, &h);
     secp256k1_fe_mul(&t, &u1, &h2);
-    r->x = t; secp256k1_fe_mul_int(&r->x, 2); secp256k1_fe_add(&r->x, &h3); secp256k1_fe_negate(&r->x, &r->x, 3); secp256k1_fe_add(&r->x, &i2);
-    secp256k1_fe_negate(&r->y, &r->x, 5); secp256k1_fe_add(&r->y, &t); secp256k1_fe_mul(&r->y, &r->y, &i);
-    secp256k1_fe_mul(&h3, &h3, &s1); secp256k1_fe_negate(&h3, &h3, 1);
+
+    secp256k1_fe_sqr(&r->x, &i);
+    secp256k1_fe_add(&r->x, &h3);
+    secp256k1_fe_add(&r->x, &t);
+    secp256k1_fe_add(&r->x, &t);
+
+    secp256k1_fe_add(&t, &r->x);
+    secp256k1_fe_mul(&r->y, &t, &i);
+    secp256k1_fe_mul(&h3, &h3, &s1);
     secp256k1_fe_add(&r->y, &h3);
 }
 
 static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, secp256k1_fe *rzr) {
-    /* 8 mul, 3 sqr, 4 normalize, 12 mul_int/add/negate */
-    secp256k1_fe z12, u1, u2, s1, s2, h, i, i2, h2, h3, t;
+    /* 8 mul, 3 sqr, 13 add/negate/normalize_weak/normalizes_to_zero (ignoring special cases) */
+    secp256k1_fe z12, u1, u2, s1, s2, h, i, h2, h3, t;
     if (a->infinity) {
         VERIFY_CHECK(rzr == NULL);
         secp256k1_gej_set_ge(r, b);
@@ -397,7 +404,6 @@ static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, c
         *r = *a;
         return;
     }
-    r->infinity = 0;
 
     secp256k1_fe_sqr(&z12, &a->z);
     u1 = a->x; secp256k1_fe_normalize_weak(&u1);
@@ -405,7 +411,7 @@ static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, c
     s1 = a->y; secp256k1_fe_normalize_weak(&s1);
     secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z);
     secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2);
-    secp256k1_fe_negate(&i, &s1, 1); secp256k1_fe_add(&i, &s2);
+    secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
     if (secp256k1_fe_normalizes_to_zero_var(&h)) {
         if (secp256k1_fe_normalizes_to_zero_var(&i)) {
             secp256k1_gej_double_var(r, a, rzr);
@@ -417,28 +423,33 @@ static void secp256k1_gej_add_ge_var(secp256k1_gej *r, const secp256k1_gej *a, c
         }
         return;
     }
-    secp256k1_fe_sqr(&i2, &i);
-    secp256k1_fe_sqr(&h2, &h);
-    secp256k1_fe_mul(&h3, &h, &h2);
+
+    r->infinity = 0;
     if (rzr != NULL) {
         *rzr = h;
     }
     secp256k1_fe_mul(&r->z, &a->z, &h);
+
+    secp256k1_fe_sqr(&h2, &h);
+    secp256k1_fe_negate(&h2, &h2, 1);
+    secp256k1_fe_mul(&h3, &h2, &h);
     secp256k1_fe_mul(&t, &u1, &h2);
-    r->x = t; secp256k1_fe_mul_int(&r->x, 2); secp256k1_fe_add(&r->x, &h3); secp256k1_fe_negate(&r->x, &r->x, 3); secp256k1_fe_add(&r->x, &i2);
-    secp256k1_fe_negate(&r->y, &r->x, 5); secp256k1_fe_add(&r->y, &t); secp256k1_fe_mul(&r->y, &r->y, &i);
-    secp256k1_fe_mul(&h3, &h3, &s1); secp256k1_fe_negate(&h3, &h3, 1);
+
+    secp256k1_fe_sqr(&r->x, &i);
+    secp256k1_fe_add(&r->x, &h3);
+    secp256k1_fe_add(&r->x, &t);
+    secp256k1_fe_add(&r->x, &t);
+
+    secp256k1_fe_add(&t, &r->x);
+    secp256k1_fe_mul(&r->y, &t, &i);
+    secp256k1_fe_mul(&h3, &h3, &s1);
     secp256k1_fe_add(&r->y, &h3);
 }
 
 static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_ge *b, const secp256k1_fe *bzinv) {
-    /* 9 mul, 3 sqr, 4 normalize, 12 mul_int/add/negate */
-    secp256k1_fe az, z12, u1, u2, s1, s2, h, i, i2, h2, h3, t;
+    /* 9 mul, 3 sqr, 13 add/negate/normalize_weak/normalizes_to_zero (ignoring special cases) */
+    secp256k1_fe az, z12, u1, u2, s1, s2, h, i, h2, h3, t;
 
-    if (b->infinity) {
-        *r = *a;
-        return;
-    }
     if (a->infinity) {
         secp256k1_fe bzinv2, bzinv3;
         r->infinity = b->infinity;
@@ -449,7 +460,10 @@ static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a,
         secp256k1_fe_set_int(&r->z, 1);
         return;
     }
-    r->infinity = 0;
+    if (b->infinity) {
+        *r = *a;
+        return;
+    }
 
     /** We need to calculate (rx,ry,rz) = (ax,ay,az) + (bx,by,1/bzinv). Due to
      *  secp256k1's isomorphism we can multiply the Z coordinates on both sides
@@ -467,7 +481,7 @@ static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a,
     s1 = a->y; secp256k1_fe_normalize_weak(&s1);
     secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &az);
     secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2);
-    secp256k1_fe_negate(&i, &s1, 1); secp256k1_fe_add(&i, &s2);
+    secp256k1_fe_negate(&i, &s2, 1); secp256k1_fe_add(&i, &s1);
     if (secp256k1_fe_normalizes_to_zero_var(&h)) {
         if (secp256k1_fe_normalizes_to_zero_var(&i)) {
             secp256k1_gej_double_var(r, a, NULL);
@@ -476,14 +490,23 @@ static void secp256k1_gej_add_zinv_var(secp256k1_gej *r, const secp256k1_gej *a,
         }
         return;
     }
-    secp256k1_fe_sqr(&i2, &i);
+
+    r->infinity = 0;
+    secp256k1_fe_mul(&r->z, &a->z, &h);
+
     secp256k1_fe_sqr(&h2, &h);
-    secp256k1_fe_mul(&h3, &h, &h2);
-    r->z = a->z; secp256k1_fe_mul(&r->z, &r->z, &h);
+    secp256k1_fe_negate(&h2, &h2, 1);
+    secp256k1_fe_mul(&h3, &h2, &h);
     secp256k1_fe_mul(&t, &u1, &h2);
-    r->x = t; secp256k1_fe_mul_int(&r->x, 2); secp256k1_fe_add(&r->x, &h3); secp256k1_fe_negate(&r->x, &r->x, 3); secp256k1_fe_add(&r->x, &i2);
-    secp256k1_fe_negate(&r->y, &r->x, 5); secp256k1_fe_add(&r->y, &t); secp256k1_fe_mul(&r->y, &r->y, &i);
-    secp256k1_fe_mul(&h3, &h3, &s1); secp256k1_fe_negate(&h3, &h3, 1);
+
+    secp256k1_fe_sqr(&r->x, &i);
+    secp256k1_fe_add(&r->x, &h3);
+    secp256k1_fe_add(&r->x, &t);
+    secp256k1_fe_add(&r->x, &t);
+
+    secp256k1_fe_add(&t, &r->x);
+    secp256k1_fe_mul(&r->y, &t, &i);
+    secp256k1_fe_mul(&h3, &h3, &s1);
     secp256k1_fe_add(&r->y, &h3);
 }