This repository was archived by the owner on Jul 23, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathmain.cpp
More file actions
236 lines (194 loc) · 8.86 KB
/
main.cpp
File metadata and controls
236 lines (194 loc) · 8.86 KB
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
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
#include "bench_merkle_tree.hpp"
#include "bench_ntt.hpp"
#include "bench_rescue_prime.hpp"
#include <iomanip>
using namespace sycl;
const uint32_t N = 1 << 10;
const uint32_t B = 1 << 7;
const size_t BENCH_ROUND = 4;
int
main(int argc, char** argv)
{
// device selection is based on flag provided to compiler
// such as `clang++ main.cpp {...}.cpp -D{CPU,GPU,HOST,DEFAULT} -fsycl
// -std=c++20 -Wall`
//
// but someone using make utility, should be invoking it as
// `DEVICE=cpu|gpu|host make <target>`
#if defined CPU
device d{ cpu_selector{} };
#elif defined GPU
device d{ gpu_selector{} };
#elif defined HOST
device d{ host_selector{} };
#else
device d{ default_selector{} };
#endif
queue q{ d, { sycl::property::queue::enable_profiling() } };
std::cout << "running on " << d.get_info<info::device::name>() << "\n"
<< std::endl;
std::cout << "\nRescue prime hash on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "dimension"
<< "\t\t" << std::setw(10) << "iterations"
<< "\t\t" << std::setw(15) << "total"
<< "\t\t" << std::setw(20) << "avg"
<< "\t\t" << std::setw(20) << "op/s" << std::endl;
for (uint dim = B; dim <= (1ul << 12); dim <<= 1) {
// note iteration count is set to 1, so each work-item
// only hashes input one time
//
// time in nanoseconds --- be careful !
uint64_t tm = benchmark_hash_elements(q, dim, 1ul << 6, 1);
std::cout << std::setw(5) << std::left << dim << "x" << std::setw(5)
<< std::right << dim << "\t\t" << std::setw(8) << std::right << 1
<< "\t\t" << std::setw(15) << std::right << tm << " ns"
<< "\t\t" << std::setw(15) << std::right
<< (double)tm / (double)(dim * dim * 1) << " ns"
<< "\t\t" << std::setw(15) << std::right
<< 1e9 / ((double)tm / (double)(dim * dim * 1)) << std::endl;
}
std::cout << "\nRescue prime merge hashes ( using global memory ) on F(2**64 "
"- 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "dimension"
<< "\t\t" << std::setw(10) << "iterations"
<< "\t\t" << std::setw(15) << "total"
<< "\t\t" << std::setw(20) << "avg"
<< "\t\t" << std::setw(20) << "op/s" << std::endl;
for (uint dim = B; dim <= (1ul << 12); dim <<= 1) {
// time in nanoseconds --- beware !
uint64_t tm = benchmark_merge(q, dim, 1ul << 6, 1);
std::cout << std::setw(5) << std::left << dim << "x" << std::setw(5)
<< std::right << dim << "\t\t" << std::setw(8) << std::right << 1
<< "\t\t" << std::setw(15) << std::right << tm << " ns"
<< "\t\t" << std::setw(15) << std::right
<< (double)tm / (double)(dim * dim * 1) << " ns"
<< "\t\t" << std::setw(15) << std::right
<< 1e9 / ((double)tm / (double)(dim * dim * 1)) << std::endl;
}
std::cout << "\nRescue prime merge hashes ( using scratch pad ) on F(2**64 - "
"2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "dimension"
<< "\t\t" << std::setw(10) << "iterations"
<< "\t\t" << std::setw(15) << "total"
<< "\t\t" << std::setw(20) << "avg"
<< "\t\t" << std::setw(20) << "op/s" << std::endl;
for (uint dim = B; dim <= (1ul << 12); dim <<= 1) {
// time in nanoseconds --- beware !
uint64_t tm = benchmark_merge_using_scratch_pad(q, dim, 1ul << 6, 1);
std::cout << std::setw(5) << std::left << dim << "x" << std::setw(5)
<< std::right << dim << "\t\t" << std::setw(8) << std::right << 1
<< "\t\t" << std::setw(15) << std::right << tm << " ns"
<< "\t\t" << std::setw(15) << std::right
<< (double)tm / (double)(dim * dim * 1) << " ns"
<< "\t\t" << std::setw(15) << std::right
<< 1e9 / ((double)tm / (double)(dim * dim * 1)) << std::endl;
}
std::cout << "\nMerklize ( approach 1 ) using Rescue Prime on F(2**64 - "
"2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "leaves"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = (1ul << 20); dim <= (1ul << 24); dim <<= 1) {
// time in nanoseconds --- beware !
double tm = 0;
for (size_t i = 0; i < BENCH_ROUND; i++) {
tm +=
static_cast<double>(benchmark_merklize_approach_1(q, dim, 1ul << 5));
}
tm /= static_cast<double>(BENCH_ROUND);
std::cout << std::setw(11) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << tm * 1e-6 << " ms" << std::endl;
}
std::cout << "\nMerklize ( approach 2 ) using Rescue Prime on F(2**64 - "
"2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "leaves"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = (1ul << 20); dim <= (1ul << 24); dim <<= 1) {
// time in nanoseconds --- beware !
double tm = 0;
for (size_t i = 0; i < BENCH_ROUND; i++) {
tm +=
static_cast<double>(benchmark_merklize_approach_2(q, dim, 1ul << 5));
}
tm /= static_cast<double>(BENCH_ROUND);
std::cout << std::setw(11) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << tm * 1e-6 << " ms" << std::endl;
}
std::cout << "\nMerklize ( approach 3 ) using Rescue Prime on F(2**64 - "
"2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "leaves"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = (1ul << 20); dim <= (1ul << 24); dim <<= 1) {
// time in nanoseconds --- beware !
double tm = 0;
for (size_t i = 0; i < BENCH_ROUND; i++) {
tm +=
static_cast<double>(benchmark_merklize_approach_3(q, dim, 1ul << 5));
}
tm /= static_cast<double>(BENCH_ROUND);
std::cout << std::setw(11) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << tm * 1e-6 << " ms" << std::endl;
}
std::cout << "\nForward NTT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(5) << "dimension"
<< "\t\t" << std::setw(10) << "total" << std::endl;
for (uint dim = B; dim <= N; dim <<= 1) {
int64_t tm = benchmark_forward_transform(q, dim, B);
std::cout << std::setw(5) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << tm << " us" << std::endl;
}
std::cout << "\nInverse NTT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(5) << "dimension"
<< "\t\t" << std::setw(10) << "total" << std::endl;
for (uint dim = B; dim <= N; dim <<= 1) {
int64_t tm = benchmark_inverse_transform(q, dim, B);
std::cout << std::setw(5) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << tm << " us" << std::endl;
}
std::cout << "\nCooley-Tukey FFT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(10) << "dimension"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = 1ul << 16; dim <= (1ul << 23); dim <<= 1) {
int64_t tm = benchmark_cooley_tukey_fft(q, dim, B);
std::cout << std::setw(8) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << (float)tm / 1000.f << " ms" << std::endl;
}
std::cout << "\nCooley-Tukey IFFT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(10) << "dimension"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = 1ul << 16; dim <= (1ul << 23); dim <<= 1) {
int64_t tm = benchmark_cooley_tukey_ifft(q, dim, B);
std::cout << std::setw(8) << std::right << dim << "\t\t" << std::setw(15)
<< std::right << (float)tm / 1000.f << " ms" << std::endl;
}
std::cout << "\nSix-Step FFT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "dimension"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = 16; dim <= 24; dim++) {
int64_t tm = benchmark_six_step_fft(q, 1ul << dim, 1 << 6);
std::cout << std::setw(9) << std::right << (1ul << dim) << "\t\t"
<< std::setw(15) << std::right << (float)tm / 1000.f << " ms"
<< std::endl;
}
std::cout << "\nSix-Step IFFT on F(2**64 - 2**32 + 1) elements 👇\n"
<< std::endl;
std::cout << std::setw(11) << "dimension"
<< "\t\t" << std::setw(15) << "total" << std::endl;
for (uint dim = 16; dim <= 24; dim++) {
int64_t tm = benchmark_six_step_ifft(q, 1ul << dim, 1 << 6);
std::cout << std::setw(9) << std::right << (1ul << dim) << "\t\t"
<< std::setw(15) << std::right << (float)tm / 1000.f << " ms"
<< std::endl;
}
return 0;
}