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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
|
// SPDX-License-Identifier: BSD-2-Clause
/* Copyright (C) 2014 - 2021 Intel Corporation. */
#include "allocator_perf_tool/HugePageOrganizer.hpp"
#include <memkind/internal/memkind_hbw.h>
#include "check.h"
#include "trial_generator.h"
#include <numa.h>
#include <numaif.h>
#include <unistd.h>
#include <vector>
trial_t TrialGenerator ::create_trial_tuple(alloc_api_t api, size_t size,
size_t alignment, int page_size,
memkind_t memkind, int free_index)
{
trial_t ltrial;
ltrial.api = api;
ltrial.size = size;
ltrial.alignment = alignment;
ltrial.page_size = page_size;
ltrial.memkind = memkind;
ltrial.free_index = free_index;
return ltrial;
}
void TrialGenerator ::generate_gb(alloc_api_t api, int number_of_gb_pages,
memkind_t memkind, alloc_api_t api_free,
bool psize_strict, size_t align)
{
std::vector<size_t> sizes_to_alloc;
// When API = HBW_MEMALIGN_PSIZE: psize is set to HBW_PAGESIZE_1GB_STRICT
// when allocation is a multiple of 1GB. Otherwise it is set to
// HBW_PAGESIZE_1GB.
for (int i = 1; i <= number_of_gb_pages; i++) {
if (psize_strict || api != HBW_MEMALIGN_PSIZE)
sizes_to_alloc.push_back(i * GB);
else
sizes_to_alloc.push_back(i * GB + 1);
}
int k = 0;
trial_vec.clear();
for (int i = 0; i < (int)sizes_to_alloc.size(); i++) {
trial_vec.push_back(create_trial_tuple(api, sizes_to_alloc[i], align,
2 * MB, memkind, -1));
if (i > 0)
k++;
trial_vec.push_back(
create_trial_tuple(api_free, 0, 0, 0, memkind, k++));
}
}
int n_random(int i)
{
return random() % i;
}
void TrialGenerator ::generate_size_2bytes_2KB_2MB(alloc_api_t api)
{
size_t size[] = {2, 2 * KB, 2 * MB};
int k = 0;
trial_vec.clear();
for (unsigned int i = 0; i < (int)(sizeof(size) / sizeof(size[0])); i++) {
trial_vec.push_back(create_trial_tuple(
api, size[i], 32, sysconf(_SC_PAGESIZE), MEMKIND_HBW, -1));
if (i > 0)
k++;
trial_vec.push_back(
create_trial_tuple(HBW_FREE, 0, 0, 0, MEMKIND_HBW, k));
k++;
}
}
void TrialGenerator ::print()
{
std::vector<trial_t>::iterator it;
std::cout << "*********** Size: " << trial_vec.size() << "********\n";
std::cout << "SIZE PSIZE ALIGN FREE KIND" << std::endl;
for (it = trial_vec.begin(); it != trial_vec.end(); it++) {
std::cout << it->size << " " << it->page_size << " " << it->alignment
<< " " << it->free_index << " " << it->memkind << " "
<< std::endl;
}
}
void TrialGenerator ::run(int num_bandwidth, std::vector<int> &bandwidth)
{
int num_trial = trial_vec.size();
int i, ret = 0;
void **ptr_vec = NULL;
ptr_vec = (void **)malloc(num_trial * sizeof(void *));
if (NULL == ptr_vec) {
fprintf(stderr, "Error in allocating ptr array\n");
exit(-1);
}
for (i = 0; i < num_trial; ++i) {
ptr_vec[i] = NULL;
}
for (i = 0; i < num_trial; ++i) {
switch (trial_vec[i].api) {
case HBW_FREE:
if (i == num_trial - 1 || trial_vec[i + 1].api != HBW_REALLOC) {
hbw_free(ptr_vec[trial_vec[i].free_index]);
ptr_vec[trial_vec[i].free_index] = NULL;
ptr_vec[i] = NULL;
} else {
ptr_vec[i + 1] =
hbw_realloc(ptr_vec[trial_vec[i].free_index],
trial_vec[i + 1].size);
ptr_vec[trial_vec[i].free_index] = NULL;
}
break;
case MEMKIND_FREE:
if (i == num_trial - 1 ||
trial_vec[i + 1].api != MEMKIND_REALLOC) {
memkind_free(trial_vec[i].memkind,
ptr_vec[trial_vec[i].free_index]);
ptr_vec[trial_vec[i].free_index] = NULL;
ptr_vec[i] = NULL;
} else {
ptr_vec[i + 1] = memkind_realloc(
trial_vec[i].memkind, ptr_vec[trial_vec[i].free_index],
trial_vec[i + 1].size);
ptr_vec[trial_vec[i].free_index] = NULL;
}
break;
case HBW_MALLOC:
fprintf(stdout, "Allocating %zd bytes using hbw_malloc\n",
trial_vec[i].size);
ptr_vec[i] = hbw_malloc(trial_vec[i].size);
break;
case HBW_CALLOC:
fprintf(stdout, "Allocating %zd bytes using hbw_calloc\n",
trial_vec[i].size);
ptr_vec[i] = hbw_calloc(trial_vec[i].size, 1);
break;
case HBW_REALLOC:
fprintf(stdout, "Allocating %zd bytes using hbw_realloc\n",
trial_vec[i].size);
fflush(stdout);
if (NULL == ptr_vec[i]) {
ptr_vec[i] = hbw_realloc(NULL, trial_vec[i].size);
}
break;
case HBW_MEMALIGN:
fprintf(stdout, "Allocating %zd bytes using hbw_memalign\n",
trial_vec[i].size);
ret = hbw_posix_memalign(&ptr_vec[i], trial_vec[i].alignment,
trial_vec[i].size);
break;
case HBW_MEMALIGN_PSIZE:
fprintf(stdout,
"Allocating %zd bytes using hbw_memalign_psize\n",
trial_vec[i].size);
hbw_pagesize_t psize;
if (trial_vec[i].page_size == (size_t)sysconf(_SC_PAGESIZE))
psize = HBW_PAGESIZE_4KB;
else if (trial_vec[i].page_size == 2097152)
psize = HBW_PAGESIZE_2MB;
else if (trial_vec[i].size % trial_vec[i].page_size > 0)
psize = HBW_PAGESIZE_1GB;
else
psize = HBW_PAGESIZE_1GB_STRICT;
ret = hbw_posix_memalign_psize(&ptr_vec[i],
trial_vec[i].alignment,
trial_vec[i].size, psize);
break;
case MEMKIND_MALLOC:
fprintf(stdout, "Allocating %zd bytes using memkind_malloc\n",
trial_vec[i].size);
ptr_vec[i] =
memkind_malloc(trial_vec[i].memkind, trial_vec[i].size);
break;
case MEMKIND_CALLOC:
fprintf(stdout, "Allocating %zd bytes using memkind_calloc\n",
trial_vec[i].size);
ptr_vec[i] =
memkind_calloc(trial_vec[i].memkind, trial_vec[i].size, 1);
break;
case MEMKIND_REALLOC:
fprintf(stdout, "Allocating %zd bytes using memkind_realloc\n",
trial_vec[i].size);
if (NULL == ptr_vec[i]) {
ptr_vec[i] = memkind_realloc(trial_vec[i].memkind,
ptr_vec[i], trial_vec[i].size);
}
break;
case MEMKIND_POSIX_MEMALIGN:
fprintf(stdout,
"Allocating %zd bytes using memkind_posix_memalign\n",
trial_vec[i].size);
ret = memkind_posix_memalign(trial_vec[i].memkind, &ptr_vec[i],
trial_vec[i].alignment,
trial_vec[i].size);
break;
}
if (trial_vec[i].api != HBW_FREE && trial_vec[i].api != MEMKIND_FREE &&
trial_vec[i].memkind != MEMKIND_DEFAULT) {
ASSERT_TRUE(ptr_vec[i] != NULL);
memset(ptr_vec[i], 0, trial_vec[i].size);
Check check(ptr_vec[i], trial_vec[i]);
if (trial_vec[i].api == HBW_CALLOC) {
EXPECT_EQ(0, check.check_zero());
}
if (trial_vec[i].api == HBW_MEMALIGN ||
trial_vec[i].api == HBW_MEMALIGN_PSIZE ||
trial_vec[i].api == MEMKIND_POSIX_MEMALIGN) {
EXPECT_EQ(0, check.check_align(trial_vec[i].alignment));
EXPECT_EQ(0, ret);
}
if (trial_vec[i].api == HBW_MEMALIGN_PSIZE ||
(trial_vec[i].api == MEMKIND_MALLOC &&
(trial_vec[i].memkind == MEMKIND_HBW_HUGETLB ||
trial_vec[i].memkind == MEMKIND_HBW_PREFERRED_HUGETLB))) {
EXPECT_EQ(0, check.check_page_size(trial_vec[i].page_size));
}
}
}
for (i = 0; i < num_trial; ++i) {
if (ptr_vec[i]) {
hbw_free(ptr_vec[i]);
}
}
}
void TGTest ::SetUp()
{
size_t node;
char *hbw_nodes_env, *endptr;
tgen = std::unique_ptr<TrialGenerator>(new TrialGenerator());
hbw_nodes_env = getenv("MEMKIND_HBW_NODES");
if (hbw_nodes_env) {
num_bandwidth = 128;
for (node = 0; node < num_bandwidth; node++) {
bandwidth.push_back(1);
}
node = strtol(hbw_nodes_env, &endptr, 10);
bandwidth.push_back(2);
while (*endptr == ':') {
hbw_nodes_env = endptr + 1;
node = strtol(hbw_nodes_env, &endptr, 10);
if (endptr != hbw_nodes_env && node < num_bandwidth) {
bandwidth.push_back(2);
}
}
} else {
num_bandwidth = NUMA_NUM_NODES;
nodemask_t nodemask;
struct bitmask nodemask_bm = {NUMA_NUM_NODES, nodemask.n};
numa_bitmask_clearall(&nodemask_bm);
memkind_hbw_all_get_mbind_nodemask(NULL, nodemask.n, NUMA_NUM_NODES);
int i, max_node = numa_max_node();
for (i = 0; i < NUMA_NUM_NODES; i++) {
if (i > max_node) {
bandwidth.push_back(0);
} else if (numa_bitmask_isbitset(&nodemask_bm, i)) {
bandwidth.push_back(2);
} else {
bandwidth.push_back(1);
}
}
}
}
void TGTest ::TearDown()
{}
|
