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
|
/* SPDX-License-Identifier: Apache-2.0 */
#include "BLI_ressource_strings.h"
#include "testing/testing.h"
#include "atomic_ops.h"
#define GHASH_INTERNAL_API
#include "MEM_guardedalloc.h"
#include "BLI_utildefines.h"
#include "BLI_listbase.h"
#include "BLI_mempool.h"
#include "BLI_task.h"
#include "PIL_time.h"
#define NUM_RUN_AVERAGED 100
static uint gen_pseudo_random_number(uint num)
{
/* NOTE: this is taken from BLI_ghashutil_uinthash(), don't want to depend on external code that
* might change here... */
num += ~(num << 16);
num ^= (num >> 5);
num += (num << 3);
num ^= (num >> 13);
num += ~(num << 9);
num ^= (num >> 17);
/* Make final number in [65 - 16385] range. */
return ((num & 255) << 6) + 1;
}
/* *** Parallel iterations over double-linked list items. *** */
static void task_listbase_light_iter_func(void *UNUSED(userdata),
void *item,
int index,
const TaskParallelTLS *__restrict UNUSED(tls))
{
LinkData *data = (LinkData *)item;
data->data = POINTER_FROM_INT(POINTER_AS_INT(data->data) + index);
}
static void task_listbase_light_membarrier_iter_func(void *userdata,
void *item,
int index,
const TaskParallelTLS *__restrict UNUSED(tls))
{
LinkData *data = (LinkData *)item;
int *count = (int *)userdata;
data->data = POINTER_FROM_INT(POINTER_AS_INT(data->data) + index);
atomic_sub_and_fetch_uint32((uint32_t *)count, 1);
}
static void task_listbase_heavy_iter_func(void *UNUSED(userdata),
void *item,
int index,
const TaskParallelTLS *__restrict UNUSED(tls))
{
LinkData *data = (LinkData *)item;
/* 'Random' number of iterations. */
const uint num = gen_pseudo_random_number((uint)index);
for (uint i = 0; i < num; i++) {
data->data = POINTER_FROM_INT(POINTER_AS_INT(data->data) + ((i % 2) ? -index : index));
}
}
static void task_listbase_heavy_membarrier_iter_func(void *userdata,
void *item,
int index,
const TaskParallelTLS *__restrict UNUSED(tls))
{
LinkData *data = (LinkData *)item;
int *count = (int *)userdata;
/* 'Random' number of iterations. */
const uint num = gen_pseudo_random_number((uint)index);
for (uint i = 0; i < num; i++) {
data->data = POINTER_FROM_INT(POINTER_AS_INT(data->data) + ((i % 2) ? -index : index));
}
atomic_sub_and_fetch_uint32((uint32_t *)count, 1);
}
static void task_listbase_test_do(ListBase *list,
const int num_items,
int *num_items_tmp,
const char *id,
TaskParallelIteratorFunc func,
const bool use_threads,
const bool check_num_items_tmp)
{
TaskParallelSettings settings;
BLI_parallel_range_settings_defaults(&settings);
settings.use_threading = use_threads;
double averaged_timing = 0.0;
for (int i = 0; i < NUM_RUN_AVERAGED; i++) {
const double init_time = PIL_check_seconds_timer();
BLI_task_parallel_listbase(list, num_items_tmp, func, &settings);
averaged_timing += PIL_check_seconds_timer() - init_time;
/* Those checks should ensure us all items of the listbase were processed once, and only once -
* as expected. */
if (check_num_items_tmp) {
EXPECT_EQ(*num_items_tmp, 0);
}
LinkData *item;
int j;
for (j = 0, item = (LinkData *)list->first; j < num_items && item != nullptr;
j++, item = item->next) {
EXPECT_EQ(POINTER_AS_INT(item->data), j);
item->data = POINTER_FROM_INT(0);
}
EXPECT_EQ(num_items, j);
*num_items_tmp = num_items;
}
printf("\t%s: done in %fs on average over %d runs\n",
id,
averaged_timing / NUM_RUN_AVERAGED,
NUM_RUN_AVERAGED);
}
static void task_listbase_test(const char *id, const int nbr, const bool use_threads)
{
printf("\n========== STARTING %s ==========\n", id);
ListBase list = {nullptr, nullptr};
LinkData *items_buffer = (LinkData *)MEM_calloc_arrayN(nbr, sizeof(*items_buffer), __func__);
BLI_threadapi_init();
int num_items = 0;
for (int i = 0; i < nbr; i++) {
BLI_addtail(&list, &items_buffer[i]);
num_items++;
}
int num_items_tmp = num_items;
task_listbase_test_do(&list,
num_items,
&num_items_tmp,
"Light iter",
task_listbase_light_iter_func,
use_threads,
false);
task_listbase_test_do(&list,
num_items,
&num_items_tmp,
"Light iter with mem barrier",
task_listbase_light_membarrier_iter_func,
use_threads,
true);
task_listbase_test_do(&list,
num_items,
&num_items_tmp,
"Heavy iter",
task_listbase_heavy_iter_func,
use_threads,
false);
task_listbase_test_do(&list,
num_items,
&num_items_tmp,
"Heavy iter with mem barrier",
task_listbase_heavy_membarrier_iter_func,
use_threads,
true);
MEM_freeN(items_buffer);
BLI_threadapi_exit();
printf("========== ENDED %s ==========\n\n", id);
}
TEST(task, ListBaseIterNoThread10k)
{
task_listbase_test("ListBase parallel iteration - Single thread - 10000 items", 10000, false);
}
TEST(task, ListBaseIter10k)
{
task_listbase_test("ListBase parallel iteration - Threaded - 10000 items", 10000, true);
}
TEST(task, ListBaseIterNoThread100k)
{
task_listbase_test("ListBase parallel iteration - Single thread - 100000 items", 100000, false);
}
TEST(task, ListBaseIter100k)
{
task_listbase_test("ListBase parallel iteration - Threaded - 100000 items", 100000, true);
}
|
