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
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
|
/*
* Copyright © 2016 Advanced Micro Devices, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sub license, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
* AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial portions
* of the Software.
*/
#include "u_queue.h"
#include "util/u_string.h"
static void util_queue_killall_and_wait(struct util_queue *queue);
/****************************************************************************
* Wait for all queues to assert idle when exit() is called.
*
* Otherwise, C++ static variable destructors can be called while threads
* are using the static variables.
*/
static once_flag atexit_once_flag = ONCE_FLAG_INIT;
static struct list_head queue_list;
static mtx_t exit_mutex = _MTX_INITIALIZER_NP;
static void
atexit_handler(void)
{
struct util_queue *iter;
mtx_lock(&exit_mutex);
/* Wait for all queues to assert idle. */
LIST_FOR_EACH_ENTRY(iter, &queue_list, head) {
util_queue_killall_and_wait(iter);
}
mtx_unlock(&exit_mutex);
}
static void
global_init(void)
{
LIST_INITHEAD(&queue_list);
atexit(atexit_handler);
}
static void
add_to_atexit_list(struct util_queue *queue)
{
call_once(&atexit_once_flag, global_init);
mtx_lock(&exit_mutex);
LIST_ADD(&queue->head, &queue_list);
mtx_unlock(&exit_mutex);
}
static void
remove_from_atexit_list(struct util_queue *queue)
{
struct util_queue *iter, *tmp;
mtx_lock(&exit_mutex);
LIST_FOR_EACH_ENTRY_SAFE(iter, tmp, &queue_list, head) {
if (iter == queue) {
LIST_DEL(&iter->head);
break;
}
}
mtx_unlock(&exit_mutex);
}
/****************************************************************************
* util_queue_fence
*/
static void
util_queue_fence_signal(struct util_queue_fence *fence)
{
mtx_lock(&fence->mutex);
fence->signalled = true;
cnd_broadcast(&fence->cond);
mtx_unlock(&fence->mutex);
}
void
util_queue_fence_wait(struct util_queue_fence *fence)
{
mtx_lock(&fence->mutex);
while (!fence->signalled)
cnd_wait(&fence->cond, &fence->mutex);
mtx_unlock(&fence->mutex);
}
void
util_queue_fence_init(struct util_queue_fence *fence)
{
memset(fence, 0, sizeof(*fence));
(void) mtx_init(&fence->mutex, mtx_plain);
cnd_init(&fence->cond);
fence->signalled = true;
}
void
util_queue_fence_destroy(struct util_queue_fence *fence)
{
assert(fence->signalled);
cnd_destroy(&fence->cond);
mtx_destroy(&fence->mutex);
}
/****************************************************************************
* util_queue implementation
*/
struct thread_input {
struct util_queue *queue;
int thread_index;
};
static int
util_queue_thread_func(void *input)
{
struct util_queue *queue = ((struct thread_input*)input)->queue;
int thread_index = ((struct thread_input*)input)->thread_index;
free(input);
if (queue->name) {
char name[16];
util_snprintf(name, sizeof(name), "%s:%i", queue->name, thread_index);
u_thread_setname(name);
}
if (queue->flags & UTIL_QUEUE_INIT_USE_MINIMUM_PRIORITY) {
#if defined(__linux__)
struct sched_param sched_param = {0};
/* The nice() function can only set a maximum of 19.
* SCHED_IDLE is the same as nice = 20.
*
* Note that Linux only allows decreasing the priority. The original
* priority can't be restored.
*/
pthread_setschedparam(queue->threads[thread_index], SCHED_IDLE,
&sched_param);
#endif
}
while (1) {
struct util_queue_job job;
mtx_lock(&queue->lock);
assert(queue->num_queued >= 0 && queue->num_queued <= queue->max_jobs);
/* wait if the queue is empty */
while (!queue->kill_threads && queue->num_queued == 0)
cnd_wait(&queue->has_queued_cond, &queue->lock);
if (queue->kill_threads) {
mtx_unlock(&queue->lock);
break;
}
job = queue->jobs[queue->read_idx];
memset(&queue->jobs[queue->read_idx], 0, sizeof(struct util_queue_job));
queue->read_idx = (queue->read_idx + 1) % queue->max_jobs;
queue->num_queued--;
cnd_signal(&queue->has_space_cond);
mtx_unlock(&queue->lock);
if (job.job) {
job.execute(job.job, thread_index);
util_queue_fence_signal(job.fence);
if (job.cleanup)
job.cleanup(job.job, thread_index);
}
}
/* signal remaining jobs before terminating */
mtx_lock(&queue->lock);
for (unsigned i = queue->read_idx; i != queue->write_idx;
i = (i + 1) % queue->max_jobs) {
if (queue->jobs[i].job) {
util_queue_fence_signal(queue->jobs[i].fence);
queue->jobs[i].job = NULL;
}
}
queue->read_idx = queue->write_idx;
queue->num_queued = 0;
mtx_unlock(&queue->lock);
return 0;
}
bool
util_queue_init(struct util_queue *queue,
const char *name,
unsigned max_jobs,
unsigned num_threads,
unsigned flags)
{
unsigned i;
memset(queue, 0, sizeof(*queue));
queue->name = name;
queue->num_threads = num_threads;
queue->flags = flags;
queue->max_jobs = max_jobs;
queue->jobs = (struct util_queue_job*)
calloc(max_jobs, sizeof(struct util_queue_job));
if (!queue->jobs)
goto fail;
(void) mtx_init(&queue->lock, mtx_plain);
queue->num_queued = 0;
cnd_init(&queue->has_queued_cond);
cnd_init(&queue->has_space_cond);
queue->threads = (thrd_t*) calloc(num_threads, sizeof(thrd_t));
if (!queue->threads)
goto fail;
/* start threads */
for (i = 0; i < num_threads; i++) {
struct thread_input *input =
(struct thread_input *) malloc(sizeof(struct thread_input));
input->queue = queue;
input->thread_index = i;
queue->threads[i] = u_thread_create(util_queue_thread_func, input);
if (!queue->threads[i]) {
free(input);
if (i == 0) {
/* no threads created, fail */
goto fail;
} else {
/* at least one thread created, so use it */
queue->num_threads = i;
break;
}
}
}
add_to_atexit_list(queue);
return true;
fail:
free(queue->threads);
if (queue->jobs) {
cnd_destroy(&queue->has_space_cond);
cnd_destroy(&queue->has_queued_cond);
mtx_destroy(&queue->lock);
free(queue->jobs);
}
/* also util_queue_is_initialized can be used to check for success */
memset(queue, 0, sizeof(*queue));
return false;
}
static void
util_queue_killall_and_wait(struct util_queue *queue)
{
unsigned i;
/* Signal all threads to terminate. */
mtx_lock(&queue->lock);
queue->kill_threads = 1;
cnd_broadcast(&queue->has_queued_cond);
mtx_unlock(&queue->lock);
for (i = 0; i < queue->num_threads; i++)
thrd_join(queue->threads[i], NULL);
queue->num_threads = 0;
}
void
util_queue_destroy(struct util_queue *queue)
{
util_queue_killall_and_wait(queue);
remove_from_atexit_list(queue);
cnd_destroy(&queue->has_space_cond);
cnd_destroy(&queue->has_queued_cond);
mtx_destroy(&queue->lock);
free(queue->jobs);
free(queue->threads);
}
void
util_queue_add_job(struct util_queue *queue,
void *job,
struct util_queue_fence *fence,
util_queue_execute_func execute,
util_queue_execute_func cleanup)
{
struct util_queue_job *ptr;
assert(fence->signalled);
mtx_lock(&queue->lock);
if (queue->kill_threads) {
mtx_unlock(&queue->lock);
/* well no good option here, but any leaks will be
* short-lived as things are shutting down..
*/
return;
}
fence->signalled = false;
assert(queue->num_queued >= 0 && queue->num_queued <= queue->max_jobs);
/* if the queue is full, wait until there is space */
while (queue->num_queued == queue->max_jobs)
cnd_wait(&queue->has_space_cond, &queue->lock);
ptr = &queue->jobs[queue->write_idx];
assert(ptr->job == NULL);
ptr->job = job;
ptr->fence = fence;
ptr->execute = execute;
ptr->cleanup = cleanup;
queue->write_idx = (queue->write_idx + 1) % queue->max_jobs;
queue->num_queued++;
cnd_signal(&queue->has_queued_cond);
mtx_unlock(&queue->lock);
}
/**
* Remove a queued job. If the job hasn't started execution, it's removed from
* the queue. If the job has started execution, the function waits for it to
* complete.
*
* In all cases, the fence is signalled when the function returns.
*
* The function can be used when destroying an object associated with the job
* when you don't care about the job completion state.
*/
void
util_queue_drop_job(struct util_queue *queue, struct util_queue_fence *fence)
{
bool removed = false;
if (util_queue_fence_is_signalled(fence))
return;
mtx_lock(&queue->lock);
for (unsigned i = queue->read_idx; i != queue->write_idx;
i = (i + 1) % queue->max_jobs) {
if (queue->jobs[i].fence == fence) {
if (queue->jobs[i].cleanup)
queue->jobs[i].cleanup(queue->jobs[i].job, -1);
/* Just clear it. The threads will treat as a no-op job. */
memset(&queue->jobs[i], 0, sizeof(queue->jobs[i]));
removed = true;
break;
}
}
mtx_unlock(&queue->lock);
if (removed)
util_queue_fence_signal(fence);
else
util_queue_fence_wait(fence);
}
int64_t
util_queue_get_thread_time_nano(struct util_queue *queue, unsigned thread_index)
{
/* Allow some flexibility by not raising an error. */
if (thread_index >= queue->num_threads)
return 0;
return u_thread_get_time_nano(queue->threads[thread_index]);
}
|