1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
|
/**********************************************************************
* Copyright (c) 2017 Pieter Wuille *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or http://www.opensource.org/licenses/mit-license.php.*
**********************************************************************/
#include <stdio.h>
#include "include/secp256k1.h"
#include "util.h"
#include "hash_impl.h"
#include "num_impl.h"
#include "field_impl.h"
#include "group_impl.h"
#include "scalar_impl.h"
#include "ecmult_impl.h"
#include "bench.h"
#include "secp256k1.c"
#define POINTS 32768
typedef struct {
/* Setup once in advance */
secp256k1_context* ctx;
secp256k1_scratch_space* scratch;
secp256k1_scalar* scalars;
secp256k1_ge* pubkeys;
secp256k1_scalar* seckeys;
secp256k1_gej* expected_output;
secp256k1_ecmult_multi_func ecmult_multi;
/* Changes per test */
size_t count;
int includes_g;
/* Changes per test iteration */
size_t offset1;
size_t offset2;
/* Test output. */
secp256k1_gej* output;
} bench_data;
static int bench_callback(secp256k1_scalar* sc, secp256k1_ge* ge, size_t idx, void* arg) {
bench_data* data = (bench_data*)arg;
if (data->includes_g) ++idx;
if (idx == 0) {
*sc = data->scalars[data->offset1];
*ge = secp256k1_ge_const_g;
} else {
*sc = data->scalars[(data->offset1 + idx) % POINTS];
*ge = data->pubkeys[(data->offset2 + idx - 1) % POINTS];
}
return 1;
}
static void bench_ecmult(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int includes_g = data->includes_g;
int iter;
int count = data->count;
iters = iters / data->count;
for (iter = 0; iter < iters; ++iter) {
data->ecmult_multi(&data->ctx->error_callback, &data->ctx->ecmult_ctx, data->scratch, &data->output[iter], data->includes_g ? &data->scalars[data->offset1] : NULL, bench_callback, arg, count - includes_g);
data->offset1 = (data->offset1 + count) % POINTS;
data->offset2 = (data->offset2 + count - 1) % POINTS;
}
}
static void bench_ecmult_setup(void* arg) {
bench_data* data = (bench_data*)arg;
data->offset1 = (data->count * 0x537b7f6f + 0x8f66a481) % POINTS;
data->offset2 = (data->count * 0x7f6f537b + 0x6a1a8f49) % POINTS;
}
static void bench_ecmult_teardown(void* arg, int iters) {
bench_data* data = (bench_data*)arg;
int iter;
iters = iters / data->count;
/* Verify the results in teardown, to avoid doing comparisons while benchmarking. */
for (iter = 0; iter < iters; ++iter) {
secp256k1_gej tmp;
secp256k1_gej_add_var(&tmp, &data->output[iter], &data->expected_output[iter], NULL);
CHECK(secp256k1_gej_is_infinity(&tmp));
}
}
static void generate_scalar(uint32_t num, secp256k1_scalar* scalar) {
secp256k1_sha256 sha256;
unsigned char c[11] = {'e', 'c', 'm', 'u', 'l', 't', 0, 0, 0, 0};
unsigned char buf[32];
int overflow = 0;
c[6] = num;
c[7] = num >> 8;
c[8] = num >> 16;
c[9] = num >> 24;
secp256k1_sha256_initialize(&sha256);
secp256k1_sha256_write(&sha256, c, sizeof(c));
secp256k1_sha256_finalize(&sha256, buf);
secp256k1_scalar_set_b32(scalar, buf, &overflow);
CHECK(!overflow);
}
static void run_test(bench_data* data, size_t count, int includes_g, int num_iters) {
char str[32];
static const secp256k1_scalar zero = SECP256K1_SCALAR_CONST(0, 0, 0, 0, 0, 0, 0, 0);
size_t iters = 1 + num_iters / count;
size_t iter;
data->count = count;
data->includes_g = includes_g;
/* Compute (the negation of) the expected results directly. */
data->offset1 = (data->count * 0x537b7f6f + 0x8f66a481) % POINTS;
data->offset2 = (data->count * 0x7f6f537b + 0x6a1a8f49) % POINTS;
for (iter = 0; iter < iters; ++iter) {
secp256k1_scalar tmp;
secp256k1_scalar total = data->scalars[(data->offset1++) % POINTS];
size_t i = 0;
for (i = 0; i + 1 < count; ++i) {
secp256k1_scalar_mul(&tmp, &data->seckeys[(data->offset2++) % POINTS], &data->scalars[(data->offset1++) % POINTS]);
secp256k1_scalar_add(&total, &total, &tmp);
}
secp256k1_scalar_negate(&total, &total);
secp256k1_ecmult(&data->ctx->ecmult_ctx, &data->expected_output[iter], NULL, &zero, &total);
}
/* Run the benchmark. */
sprintf(str, includes_g ? "ecmult_%ig" : "ecmult_%i", (int)count);
run_benchmark(str, bench_ecmult, bench_ecmult_setup, bench_ecmult_teardown, data, 10, count * iters);
}
int main(int argc, char **argv) {
bench_data data;
int i, p;
secp256k1_gej* pubkeys_gej;
size_t scratch_size;
int iters = get_iters(10000);
data.ctx = secp256k1_context_create(SECP256K1_CONTEXT_SIGN | SECP256K1_CONTEXT_VERIFY);
scratch_size = secp256k1_strauss_scratch_size(POINTS) + STRAUSS_SCRATCH_OBJECTS*16;
data.scratch = secp256k1_scratch_space_create(data.ctx, scratch_size);
data.ecmult_multi = secp256k1_ecmult_multi_var;
if (argc > 1) {
if(have_flag(argc, argv, "pippenger_wnaf")) {
printf("Using pippenger_wnaf:\n");
data.ecmult_multi = secp256k1_ecmult_pippenger_batch_single;
} else if(have_flag(argc, argv, "strauss_wnaf")) {
printf("Using strauss_wnaf:\n");
data.ecmult_multi = secp256k1_ecmult_strauss_batch_single;
} else if(have_flag(argc, argv, "simple")) {
printf("Using simple algorithm:\n");
data.ecmult_multi = secp256k1_ecmult_multi_var;
secp256k1_scratch_space_destroy(data.ctx, data.scratch);
data.scratch = NULL;
} else {
fprintf(stderr, "%s: unrecognized argument '%s'.\n", argv[0], argv[1]);
fprintf(stderr, "Use 'pippenger_wnaf', 'strauss_wnaf', 'simple' or no argument to benchmark a combined algorithm.\n");
return 1;
}
}
/* Allocate stuff */
data.scalars = malloc(sizeof(secp256k1_scalar) * POINTS);
data.seckeys = malloc(sizeof(secp256k1_scalar) * POINTS);
data.pubkeys = malloc(sizeof(secp256k1_ge) * POINTS);
data.expected_output = malloc(sizeof(secp256k1_gej) * (iters + 1));
data.output = malloc(sizeof(secp256k1_gej) * (iters + 1));
/* Generate a set of scalars, and private/public keypairs. */
pubkeys_gej = malloc(sizeof(secp256k1_gej) * POINTS);
secp256k1_gej_set_ge(&pubkeys_gej[0], &secp256k1_ge_const_g);
secp256k1_scalar_set_int(&data.seckeys[0], 1);
for (i = 0; i < POINTS; ++i) {
generate_scalar(i, &data.scalars[i]);
if (i) {
secp256k1_gej_double_var(&pubkeys_gej[i], &pubkeys_gej[i - 1], NULL);
secp256k1_scalar_add(&data.seckeys[i], &data.seckeys[i - 1], &data.seckeys[i - 1]);
}
}
secp256k1_ge_set_all_gej_var(data.pubkeys, pubkeys_gej, POINTS);
free(pubkeys_gej);
for (i = 1; i <= 8; ++i) {
run_test(&data, i, 1, iters);
}
/* This is disabled with low count of iterations because the loop runs 77 times even with iters=1
* and the higher it goes the longer the computation takes(more points)
* So we don't run this benchmark with low iterations to prevent slow down */
if (iters > 2) {
for (p = 0; p <= 11; ++p) {
for (i = 9; i <= 16; ++i) {
run_test(&data, i << p, 1, iters);
}
}
}
if (data.scratch != NULL) {
secp256k1_scratch_space_destroy(data.ctx, data.scratch);
}
secp256k1_context_destroy(data.ctx);
free(data.scalars);
free(data.pubkeys);
free(data.seckeys);
free(data.output);
free(data.expected_output);
return(0);
}
|
