Skip to content

Commit

Permalink
Effective affine addition in EC multiplication
Browse files Browse the repository at this point in the history 
* Make secp256k1_gej_add_var and secp256k1_gej_double return the
  Z ratio to go from a.z to r.z.
* Use these Z ratios to speed up batch point conversion to affine
  coordinates, and to speed up batch conversion of points to a
  common Z coordinate.
* Add a point addition function that takes a point with a known
  Z inverse.
* Due to secp256k1's endomorphism, all additions in the EC
  multiplication code can work on affine coordinate (with an
  implicit common Z coordinate), correcting the Z coordinate of
  the result afterwards.

Refactoring by Pieter Wuille:
* Move more global-z logic into the group code.
* Separate code for computing the odd multiples from the code to bring it
  to either storage or globalz format.
* Rename functions.
* Make all addition operations return Z ratios, and test them.
* Make the zr table format compatible with future batch chaining
  (the first entry in zr becomes the ratio between the input and the
  first output).

Original idea and code by Peter Dettman.
  • Loading branch information
@peterdettman @sipa
peterdettman authored and sipa committed Apr 23, 2015
1 parent 22f60a6 commit 8e242c9
Show file tree
Hide file tree
Showing 6 changed files with 328 additions and 84 deletions.
4 changes: 2 additions & 2 deletions src/bench_internal.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -193,7 +193,7 @@ void bench_group_double_var(void* arg) {
bench_inv_t *data = (bench_inv_t*)arg;

for (i = 0; i < 200000; i++) {
secp256k1_gej_double_var(&data->gej_x, &data->gej_x);
secp256k1_gej_double_var(&data->gej_x, &data->gej_x, NULL);
}
}

Expand All @@ -202,7 +202,7 @@ void bench_group_add_var(void* arg) {
bench_inv_t *data = (bench_inv_t*)arg;

for (i = 0; i < 200000; i++) {
secp256k1_gej_add_var(&data->gej_x, &data->gej_x, &data->gej_y);
secp256k1_gej_add_var(&data->gej_x, &data->gej_x, &data->gej_y, NULL);
}
}

Expand Down
8 changes: 4 additions & 4 deletions src/ecmult_gen_impl.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -54,18 +54,18 @@ static void secp256k1_ecmult_gen_context_build(secp256k1_ecmult_gen_context_t *c
/* Set precj[j*16 .. j*16+15] to (numsbase, numsbase + gbase, ..., numsbase + 15*gbase). */
precj[j*16] = numsbase;
for (i = 1; i < 16; i++) {
secp256k1_gej_add_var(&precj[j*16 + i], &precj[j*16 + i - 1], &gbase);
secp256k1_gej_add_var(&precj[j*16 + i], &precj[j*16 + i - 1], &gbase, NULL);
}
/* Multiply gbase by 16. */
for (i = 0; i < 4; i++) {
secp256k1_gej_double_var(&gbase, &gbase);
secp256k1_gej_double_var(&gbase, &gbase, NULL);
}
/* Multiply numbase by 2. */
secp256k1_gej_double_var(&numsbase, &numsbase);
secp256k1_gej_double_var(&numsbase, &numsbase, NULL);
if (j == 62) {
/* In the last iteration, numsbase is (1 - 2^j) * nums instead. */
secp256k1_gej_neg(&numsbase, &numsbase);
secp256k1_gej_add_var(&numsbase, &numsbase, &nums_gej);
secp256k1_gej_add_var(&numsbase, &numsbase, &nums_gej, NULL);
}
}
secp256k1_ge_set_all_gej_var(1024, prec, precj);
Expand Down
144 changes: 90 additions & 54 deletions src/ecmult_impl.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -24,62 +24,85 @@
#define WINDOW_G 16
#endif

/** Fill a table 'pre' with precomputed odd multiples of a. W determines the size of the table.
* pre will contains the values [1*a,3*a,5*a,...,(2^(w-1)-1)*a], so it needs place for
* 2^(w-2) entries.
*
* There are two versions of this function:
* - secp256k1_ecmult_precomp_wnaf_gej, which operates on group elements in jacobian notation,
* fast to precompute, but slower to use in later additions.
* - secp256k1_ecmult_precomp_wnaf_ge, which operates on group elements in affine notations,
* (much) slower to precompute, but a bit faster to use in later additions.
* To compute a*P + b*G, we use the jacobian version for P, and the affine version for G, as
* G is constant, so it only needs to be done once in advance.
/** The number of entries a table with precomputed multiples needs to have. */
#define ECMULT_TABLE_SIZE(w) (1 << ((w)-2))

/** Fill a table 'prej' with precomputed odd multiples of a. Prej will contain
* the values [1*a,3*a,...,(2*n-1)*a], so it space for n values. zr[0] will
* contain prej[0].z / a.z. The other zr[i] values = prej[i].z / prej[i-1].z.
*/
static void secp256k1_ecmult_table_precomp_gej_var(secp256k1_gej_t *pre, const secp256k1_gej_t *a, int w) {
static void secp256k1_ecmult_odd_multiples_table(int n, secp256k1_gej_t *prej, secp256k1_fe_t *zr, const secp256k1_gej_t *a) {
secp256k1_gej_t d;
int i;
pre[0] = *a;
secp256k1_gej_double_var(&d, &pre[0]);
for (i = 1; i < (1 << (w-2)); i++) {
secp256k1_gej_add_var(&pre[i], &d, &pre[i-1]);

VERIFY_CHECK(!a->infinity);

prej[0] = *a;
secp256k1_gej_double_var(&d, &prej[0], NULL);
secp256k1_fe_set_int(zr, 1);
for (i = 1; i < n; i++) {
secp256k1_gej_add_var(&prej[i], &prej[i-1], &d, &zr[i]);
}
}

static void secp256k1_ecmult_table_precomp_ge_storage_var(secp256k1_ge_storage_t *pre, const secp256k1_gej_t *a, int w) {
secp256k1_gej_t d;
/** Fill a table 'pre' with precomputed odd multiples of a.
*
* There are two versions of this function:
* - secp256k1_ecmult_odd_multiples_table_globalz_windowa which brings its
* resulting point set to a single constant Z denominator, stores the X and Y
* coordinates as ge_storage points in pre, and stores the global Z in rz.
* It only operates on tables sized for WINDOW_A wnaf multiples.
* - secp256k1_ecmult_odd_multiples_table_storage_var, which converts its
* resulting point set to actually affine points, and stores those in pre.
* It operates on tables of any size, but uses heap-allocated temporaries.
*
* To compute a*P + b*G, we compute a table for P using the first function,
* and for G using the second (which requires an inverse, but it only needs to
* happen once).
*/
static void secp256k1_ecmult_odd_multiples_table_globalz_windowa(secp256k1_ge_t *pre, secp256k1_fe_t *globalz, const secp256k1_gej_t *a) {
secp256k1_gej_t prej[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_fe_t zr[ECMULT_TABLE_SIZE(WINDOW_A)];

/* Compute the odd multiples in Jacobian form. */
secp256k1_ecmult_odd_multiples_table(ECMULT_TABLE_SIZE(WINDOW_A), prej, zr, a);
/* Bring them to the same Z denominator. */
secp256k1_ge_globalz_set_table_gej(ECMULT_TABLE_SIZE(WINDOW_A), pre, globalz, prej, zr);
}

static void secp256k1_ecmult_odd_multiples_table_storage_var(int n, secp256k1_ge_storage_t *pre, const secp256k1_gej_t *a) {
secp256k1_gej_t *prej = checked_malloc(sizeof(secp256k1_gej_t) * n);
secp256k1_ge_t *prea = checked_malloc(sizeof(secp256k1_ge_t) * n);
secp256k1_fe_t *zr = checked_malloc(sizeof(secp256k1_fe_t) * n);
int i;
const int table_size = 1 << (w-2);
secp256k1_gej_t *prej = (secp256k1_gej_t *)checked_malloc(sizeof(secp256k1_gej_t) * table_size);
secp256k1_ge_t *prea = (secp256k1_ge_t *)checked_malloc(sizeof(secp256k1_ge_t) * table_size);
prej[0] = *a;
secp256k1_gej_double_var(&d, a);
for (i = 1; i < table_size; i++) {
secp256k1_gej_add_var(&prej[i], &d, &prej[i-1]);
}
secp256k1_ge_set_all_gej_var(table_size, prea, prej);
for (i = 0; i < table_size; i++) {

/* Compute the odd multiples in Jacobian form. */
secp256k1_ecmult_odd_multiples_table(n, prej, zr, a);
/* Convert them in batch to affine coordinates. */
secp256k1_ge_set_table_gej_var(n, prea, prej, zr);
/* Convert them to compact storage form. */
for (i = 0; i < n; i++) {
secp256k1_ge_to_storage(&pre[i], &prea[i]);
}
free(prej);

free(prea);
free(prej);
free(zr);
}

/** The number of entries a table with precomputed multiples needs to have. */
#define ECMULT_TABLE_SIZE(w) (1 << ((w)-2))

/** The following two macro retrieves a particular odd multiple from a table
* of precomputed multiples. */
#define ECMULT_TABLE_GET_GEJ(r,pre,n,w) do { \
#define ECMULT_TABLE_GET_GE(r,pre,n,w) do { \
VERIFY_CHECK(((n) & 1) == 1); \
VERIFY_CHECK((n) >= -((1 << ((w)-1)) - 1)); \
VERIFY_CHECK((n) <= ((1 << ((w)-1)) - 1)); \
if ((n) > 0) { \
*(r) = (pre)[((n)-1)/2]; \
} else { \
secp256k1_gej_neg((r), &(pre)[(-(n)-1)/2]); \
secp256k1_ge_neg((r), &(pre)[(-(n)-1)/2]); \
} \
} while(0)

#define ECMULT_TABLE_GET_GE_STORAGE(r,pre,n,w) do { \
VERIFY_CHECK(((n) & 1) == 1); \
VERIFY_CHECK((n) >= -((1 << ((w)-1)) - 1)); \
Expand Down Expand Up @@ -112,7 +135,7 @@ static void secp256k1_ecmult_context_build(secp256k1_ecmult_context_t *ctx) {
ctx->pre_g = (secp256k1_ge_storage_t (*)[])checked_malloc(sizeof((*ctx->pre_g)[0]) * ECMULT_TABLE_SIZE(WINDOW_G));

/* precompute the tables with odd multiples */
secp256k1_ecmult_table_precomp_ge_storage_var(*ctx->pre_g, &gj, WINDOW_G);
secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g, &gj);

#ifdef USE_ENDOMORPHISM
{
Expand All @@ -124,9 +147,9 @@ static void secp256k1_ecmult_context_build(secp256k1_ecmult_context_t *ctx) {
/* calculate 2^128*generator */
g_128j = gj;
for (i = 0; i < 128; i++) {
secp256k1_gej_double_var(&g_128j, &g_128j);
secp256k1_gej_double_var(&g_128j, &g_128j, NULL);
}
secp256k1_ecmult_table_precomp_ge_storage_var(*ctx->pre_g_128, &g_128j, WINDOW_G);
secp256k1_ecmult_odd_multiples_table_storage_var(ECMULT_TABLE_SIZE(WINDOW_G), *ctx->pre_g_128, &g_128j);
}
#endif
}
Expand Down Expand Up @@ -208,11 +231,11 @@ static int secp256k1_ecmult_wnaf(int *wnaf, const secp256k1_scalar_t *a, int w)
}

static void secp256k1_ecmult(const secp256k1_ecmult_context_t *ctx, secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_scalar_t *na, const secp256k1_scalar_t *ng) {
secp256k1_gej_t tmpj;
secp256k1_gej_t pre_a[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_ge_t pre_a[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_ge_t tmpa;
secp256k1_fe_t Z;
#ifdef USE_ENDOMORPHISM
secp256k1_gej_t pre_a_lam[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_ge_t pre_a_lam[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_scalar_t na_1, na_lam;
/* Splitted G factors. */
secp256k1_scalar_t ng_1, ng_128;
Expand Down Expand Up @@ -252,12 +275,21 @@ static void secp256k1_ecmult(const secp256k1_ecmult_context_t *ctx, secp256k1_ge
bits = bits_na;
#endif

/* calculate odd multiples of a */
secp256k1_ecmult_table_precomp_gej_var(pre_a, a, WINDOW_A);
/* Calculate odd multiples of a.
* All multiples are brought to the same Z 'denominator', which is stored
* in Z. Due to secp256k1' isomorphism we can do all operations pretending
* that the Z coordinate was 1, use affine addition formulae, and correct
* the Z coordinate of the result once at the end.
* The exception is the precomputed G table points, which are actually
* affine. Compared to the base used for other points, they have a Z ratio
* of 1/Z, so we can use secp256k1_gej_add_zinv_var, which uses the same
* isomorphism to efficiently add with a known Z inverse.
*/
secp256k1_ecmult_odd_multiples_table_globalz_windowa(pre_a, &Z, a);

#ifdef USE_ENDOMORPHISM
for (i = 0; i < ECMULT_TABLE_SIZE(WINDOW_A); i++) {
secp256k1_gej_mul_lambda(&pre_a_lam[i], &pre_a[i]);
secp256k1_ge_mul_lambda(&pre_a_lam[i], &pre_a[i]);
}

/* split ng into ng_1 and ng_128 (where gn = gn_1 + gn_128*2^128, and gn_1 and gn_128 are ~128 bit) */
Expand All @@ -281,37 +313,41 @@ static void secp256k1_ecmult(const secp256k1_ecmult_context_t *ctx, secp256k1_ge

secp256k1_gej_set_infinity(r);

for (i = bits-1; i >= 0; i--) {
for (i = bits - 1; i >= 0; i--) {
int n;
secp256k1_gej_double_var(r, r);
secp256k1_gej_double_var(r, r, NULL);
#ifdef USE_ENDOMORPHISM
if (i < bits_na_1 && (n = wnaf_na_1[i])) {
ECMULT_TABLE_GET_GEJ(&tmpj, pre_a, n, WINDOW_A);
secp256k1_gej_add_var(r, r, &tmpj);
ECMULT_TABLE_GET_GE(&tmpa, pre_a, n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa);
}
if (i < bits_na_lam && (n = wnaf_na_lam[i])) {
ECMULT_TABLE_GET_GEJ(&tmpj, pre_a_lam, n, WINDOW_A);
secp256k1_gej_add_var(r, r, &tmpj);
ECMULT_TABLE_GET_GE(&tmpa, pre_a_lam, n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa);
}
if (i < bits_ng_1 && (n = wnaf_ng_1[i])) {
ECMULT_TABLE_GET_GE_STORAGE(&tmpa, *ctx->pre_g, n, WINDOW_G);
secp256k1_gej_add_ge_var(r, r, &tmpa);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
if (i < bits_ng_128 && (n = wnaf_ng_128[i])) {
ECMULT_TABLE_GET_GE_STORAGE(&tmpa, *ctx->pre_g_128, n, WINDOW_G);
secp256k1_gej_add_ge_var(r, r, &tmpa);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
#else
if (i < bits_na && (n = wnaf_na[i])) {
ECMULT_TABLE_GET_GEJ(&tmpj, pre_a, n, WINDOW_A);
secp256k1_gej_add_var(r, r, &tmpj);
ECMULT_TABLE_GET_GE(&tmpa, pre_a, n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa);
}
if (i < bits_ng && (n = wnaf_ng[i])) {
ECMULT_TABLE_GET_GE_STORAGE(&tmpa, *ctx->pre_g, n, WINDOW_G);
secp256k1_gej_add_ge_var(r, r, &tmpa);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
#endif
}

if (!r->infinity) {
secp256k1_fe_mul(&r->z, &r->z, &Z);
}
}

#endif
20 changes: 17 additions & 3 deletions src/group.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,17 @@ static void secp256k1_ge_set_gej(secp256k1_ge_t *r, secp256k1_gej_t *a);
/** Set a batch of group elements equal to the inputs given in jacobian coordinates */
static void secp256k1_ge_set_all_gej_var(size_t len, secp256k1_ge_t *r, const secp256k1_gej_t *a);

/** Set a batch of group elements equal to the inputs given in jacobian
* coordinates (with known z-ratios). zr must contain the known z-ratios such
* that mul(a[i].z, zr[i+1]) == a[i+1].z. zr[0] is ignored. */
static void secp256k1_ge_set_table_gej_var(size_t len, secp256k1_ge_t *r, const secp256k1_gej_t *a, const secp256k1_fe_t *zr);

/** Bring a batch inputs given in jacobian coordinates (with known z-ratios) to
* the same global z "denominator". zr must contain the known z-ratios such
* that mul(a[i].z, zr[i+1]) == a[i+1].z. zr[0] is ignored. The x and y
* coordinates of the result are stored in r, the common z coordinate is
* stored in globalz. */
static void secp256k1_ge_globalz_set_table_gej(size_t len, secp256k1_ge_t *r, secp256k1_fe_t *globalz, const secp256k1_gej_t *a, const secp256k1_fe_t *zr);

/** Set a group element (jacobian) equal to the point at infinity. */
static void secp256k1_gej_set_infinity(secp256k1_gej_t *r);
Expand All @@ -82,10 +93,10 @@ static void secp256k1_gej_neg(secp256k1_gej_t *r, const secp256k1_gej_t *a);
static int secp256k1_gej_is_infinity(const secp256k1_gej_t *a);

/** Set r equal to the double of a. */
static void secp256k1_gej_double_var(secp256k1_gej_t *r, const secp256k1_gej_t *a);
static void secp256k1_gej_double_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, secp256k1_fe_t *rzr);

/** Set r equal to the sum of a and b. */
static void secp256k1_gej_add_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_gej_t *b);
static void secp256k1_gej_add_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_gej_t *b, secp256k1_fe_t *rzr);

/** Set r equal to the sum of a and b (with b given in affine coordinates, and not infinity). */
static void secp256k1_gej_add_ge(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b);
Expand All @@ -95,9 +106,12 @@ static void secp256k1_gej_add_ge(secp256k1_gej_t *r, const secp256k1_gej_t *a, c
guarantee, and b is allowed to be infinity. */
static void secp256k1_gej_add_ge_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b);

/** Set r equal to the sum of a and b (with the inverse of b's Z coordinate passed as bzinv). */
static void secp256k1_gej_add_zinv_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b, const secp256k1_fe_t *bzinv);

#ifdef USE_ENDOMORPHISM
/** Set r to be equal to lambda times a, where lambda is chosen in a way such that this is very fast. */
static void secp256k1_gej_mul_lambda(secp256k1_gej_t *r, const secp256k1_gej_t *a);
static void secp256k1_ge_mul_lambda(secp256k1_ge_t *r, const secp256k1_ge_t *a);
#endif

/** Clear a secp256k1_gej_t to prevent leaking sensitive information. */
Expand Down
Loading

0 comments on commit 8e242c9

Please sign in to comment.