-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathpi_benchmarks.cpp
More file actions
224 lines (195 loc) · 10.4 KB
/
Copy pathpi_benchmarks.cpp
File metadata and controls
224 lines (195 loc) · 10.4 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
//
// Created by khondar on 08.02.20.
//
#include <mpi.h>
#include <chrono>
#include "pi_benchmarks.h"
#include "matrix_formats/csr.hpp"
#include "power_iteration/poweriteration.h"
#include "communication.h"
void
iteration_counter(const std::vector<int> &sizes, const std::vector<double> &densities, const std::vector<double> &etas,
const unsigned max_iterations, const int num_tests) {
const auto seed = std::random_device{}();
std::cout << "seed: " << seed << " max_iterations: " << max_iterations << " num_tests: " << num_tests << "\n";
std::mt19937 rng(seed);
for (const auto size : sizes) {
for (const auto d : densities) {
for (const auto eta : etas) {
std::cout << "parameters: " << size << " " << d << " " << eta << "\n";
const std::vector<double> x(size, 1.);
const std::vector<int> rowcnt{size};
for (int i{0}; i < num_tests; ++i) {
const auto matrix = CSR::fixed_eta(size, d, eta, rng);
// test 2 4 6 8
{
unsigned avail_iterations = max_iterations;
std::vector<double> result_2;
const auto meta_2 = distributed::seg_uint::power_iteration_2(matrix, x, result_2,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_2.second;
std::vector<double> result_4;
const auto meta_4 = distributed::seg_uint::power_iteration_4(matrix, result_2, result_4,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_4.second;
std::vector<double> result_6;
const auto meta_6 = distributed::seg_uint::power_iteration_6(matrix, result_4, result_6,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_6.second;
std::vector<double> result;
const auto meta_8 = local::power_iteration(matrix, result_6, result, avail_iterations);
std::cout << "2_4_6_8 ";
if (meta_2.first && meta_4.first && meta_6.first && meta_8.first) {
std::cout << "all_converged ";
} else {
std::cout << "not_all_converged ";
}
std::cout << meta_2.second << " " << meta_4.second << " " << meta_6.second << " "
<< meta_8.second << "\n";
}
// test 4 6 8
{
unsigned avail_iterations = max_iterations;
std::vector<double> result_4;
const auto meta_4 = distributed::seg_uint::power_iteration_4(matrix, x, result_4,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_4.second;
std::vector<double> result_6;
const auto meta_6 = distributed::seg_uint::power_iteration_6(matrix, result_4, result_6,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_6.second;
std::vector<double> result;
const auto meta_8 = local::power_iteration(matrix, result_6, result, avail_iterations);
std::cout << "4_6_8 ";
if (meta_4.first && meta_6.first && meta_8.first) {
std::cout << "all_converged ";
} else {
std::cout << "not_all_converged ";
}
std::cout << meta_4.second << " " << meta_6.second << " " << meta_8.second << "\n";
}
// test 4 8
{
unsigned avail_iterations = max_iterations;
std::vector<double> result_4;
const auto meta_4 = distributed::seg_uint::power_iteration_4(matrix, x, result_4,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_4.second;
std::vector<double> result;
const auto meta_8 = local::power_iteration(matrix, result_4, result, avail_iterations);
std::cout << "4_8 ";
if (meta_4.first && meta_8.first) {
std::cout << "all_converged ";
} else {
std::cout << "not_all_converged ";
}
std::cout << meta_4.second << " " << meta_8.second << "\n";
}
// test 6 8
{
unsigned avail_iterations = max_iterations;
std::vector<double> result_6;
const auto meta_6 = distributed::seg_uint::power_iteration_6(matrix, x, result_6,
MPI_COMM_WORLD, rowcnt,
avail_iterations);
avail_iterations -= meta_6.second;
std::vector<double> result;
const auto meta_8 = local::power_iteration(matrix, result_6, result, avail_iterations);
std::cout << "6_8 ";
if (meta_6.first && meta_8.first) {
std::cout << "all_converged ";
} else {
std::cout << "not_all_converged";
}
std::cout << meta_6.second << " " << meta_8.second << "\n";
}
// test 8
{
std::vector<double> result;
const auto meta_8 = local::power_iteration(matrix, x, result, max_iterations);
std::cout << "8 ";
if (meta_8.first) {
std::cout << "all_converged ";
} else {
std::cout << "not_all_converged ";
}
std::cout << meta_8.second << "\n";
}
}
}
}
}
}
void
speedup_test(const int size, const double density, const double eta, const int max_iterations, const int num_tests,
const std::vector<int> &rowcnt, std::vector<unsigned> &seg_timings, std::vector<unsigned> &fixed_timings,
MPI_Comm comm) {
int rank, comm_size;
MPI_Comm_rank(comm, &rank);
MPI_Comm_size(comm, &comm_size);
std::mt19937 rng{std::random_device{}()};
const int root = 0;
const auto matrix = CSR::fixed_eta(size, density, eta, rng);
const auto matrix_slice = distribute_matrix(matrix, comm, root);
std::vector<double> x(matrix_slice.num_cols(), 1);
double warmup_sum{0};
for (int i{0}; i < 10; ++i) {
x.at(0) = i + 1;
MPI_Barrier(comm);
int avail_iterations = max_iterations;
std::vector<double> result_4;
const auto meta_4 = distributed::seg_uint::power_iteration_4(matrix_slice, x, result_4, comm, rowcnt,
avail_iterations);
avail_iterations -= meta_4.second;
std::vector<double> result;
distributed::fixed::power_iteration(matrix_slice, result_4, result, comm, avail_iterations);
warmup_sum += result.at(0);
}
for (int i{0}; i < num_tests; ++i) {
x.at(0) = i + 1;
MPI_Barrier(comm);
using namespace std::chrono;
const auto start = high_resolution_clock::now();
int avail_iterations = max_iterations;
std::vector<double> result_4;
const auto meta_4 = distributed::seg_uint::power_iteration_4(matrix_slice, x, result_4, comm, rowcnt,
avail_iterations);
avail_iterations -= meta_4.second;
std::vector<double> result;
const auto meta_8 = distributed::fixed::power_iteration(matrix_slice, result_4, result, comm, avail_iterations);
const auto stop = high_resolution_clock::now();
if (meta_8.first) {
seg_timings.push_back(duration_cast<milliseconds>(stop - start).count());
}
}
for (int i{0}; i < 10; ++i) {
x.at(0) = i + 1;
MPI_Barrier(comm);
int avail_iterations = max_iterations;
std::vector<double> result;
distributed::fixed::power_iteration(matrix_slice, x, result, comm, avail_iterations);
warmup_sum += result.at(0);
}
for (int i{0}; i < num_tests; ++i) {
x.at(0) = i + 1;
MPI_Barrier(comm);
using namespace std::chrono;
const auto start = high_resolution_clock::now();
int avail_iterations = max_iterations;
std::vector<double> result;
const auto meta = distributed::fixed::power_iteration(matrix_slice, x, result, comm, avail_iterations);
const auto stop = high_resolution_clock::now();
if (meta.first) {
fixed_timings.push_back(duration_cast<milliseconds>(stop - start).count());
}
}
if (rank > comm_size) {
std::cout << "warmup_sum " << warmup_sum << "\n";
}
}