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
|
/*
* Copyright © 2012 Intel Corporation
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library. If not, see <http://www.gnu.org/licenses/>.
*
* Author: Benjamin Segovia <benjamin.segovia@intel.com>
*/
#ifndef __CL_COMMAND_QUEUE_H__
#define __CL_COMMAND_QUEUE_H__
#include "cl_internals.h"
#include "cl_driver.h"
#include "cl_thread.h"
#include "cl_base_object.h"
#include "CL/cl.h"
#include <stdint.h>
struct intel_gpgpu;
typedef struct _cl_command_queue_enqueue_worker {
cl_command_queue queue;
pthread_t tid;
cl_uint cookie;
cl_bool quit;
list_head enqueued_events;
cl_uint in_exec_status; // Same value as CL_COMPLETE, CL_SUBMITTED ...
} _cl_command_queue_enqueue_worker;
typedef _cl_command_queue_enqueue_worker *cl_command_queue_enqueue_worker;
/* Basically, this is a (kind-of) batch buffer */
struct _cl_command_queue {
_cl_base_object base;
_cl_command_queue_enqueue_worker worker;
cl_context ctx; /* Its parent context */
cl_event* barrier_events; /* Point to array of non-complete user events that block this command queue */
cl_int barrier_events_num; /* Number of Non-complete user events */
cl_int barrier_events_size; /* The size of array that wait_events point to */
cl_event* wait_events; /* Point to array of non-complete user events that block this command queue */
cl_int wait_events_num; /* Number of Non-complete user events */
cl_int wait_events_size; /* The size of array that wait_events point to */
cl_command_queue_properties props; /* Queue properties */
void *thread_data; /* Used to store thread context data */
cl_mem perf; /* Where to put the perf counters */
void* cmrt_event; /* the latest CmEvent* of the command queue */
};
#define CL_OBJECT_COMMAND_QUEUE_MAGIC 0x83650a12b79ce4efLL
#define CL_OBJECT_IS_COMMAND_QUEUE(obj) (((cl_base_object)obj)->magic == CL_OBJECT_COMMAND_QUEUE_MAGIC)
/* The macro to get the thread specified gpgpu struct. */
#define GET_QUEUE_THREAD_GPGPU(queue) \
cl_gpgpu gpgpu = queue ? cl_get_thread_gpgpu(queue) : NULL; \
if (queue) \
assert(gpgpu);
/* Allocate and initialize a new command queue. Also insert it in the list of
* command queue in the associated context
*/
extern cl_command_queue cl_command_queue_new(cl_context);
/* Destroy and deallocate the command queue */
extern void cl_command_queue_delete(cl_command_queue);
/* Keep one more reference on the queue */
extern void cl_command_queue_add_ref(cl_command_queue);
/* Map ND range kernel from OCL API */
extern cl_int cl_command_queue_ND_range(cl_command_queue queue,
cl_kernel ker,
const uint32_t work_dim,
const size_t *global_work_offset,
const size_t *global_work_size,
const size_t *local_work_size);
/* The memory object where to report the performance */
extern cl_int cl_command_queue_set_report_buffer(cl_command_queue, cl_mem);
/* Flush for the command queue */
extern cl_int cl_command_queue_flush(cl_command_queue);
/* Flush for the specified gpgpu */
extern int cl_command_queue_flush_gpgpu(cl_command_queue, cl_gpgpu);
/* Wait for the completion of the command queue */
extern cl_int cl_command_queue_finish(cl_command_queue);
/* Bind all the surfaces in the GPGPU state */
extern cl_int cl_command_queue_bind_surface(cl_command_queue, cl_kernel);
/* Bind all the image surfaces in the GPGPU state */
extern cl_int cl_command_queue_bind_image(cl_command_queue, cl_kernel);
/* Insert a user event to command's wait_events */
extern void cl_command_queue_insert_event(cl_command_queue, cl_event);
/* Remove a user event from command's wait_events */
extern void cl_command_queue_remove_event(cl_command_queue, cl_event);
extern void cl_command_queue_insert_barrier_event(cl_command_queue queue, cl_event event);
extern void cl_command_queue_remove_barrier_event(cl_command_queue queue, cl_event event);
extern void cl_command_queue_notify(cl_command_queue queue);
extern void cl_command_queue_enqueue_event(cl_command_queue queue, cl_event event);
extern cl_int cl_command_queue_init_enqueue(cl_command_queue queue);
extern void cl_command_queue_destroy_enqueue(cl_command_queue queue);
extern cl_int cl_command_queue_wait_finish(cl_command_queue queue);
extern cl_int cl_command_queue_wait_flush(cl_command_queue queue);
/* Note: Must call this function with queue's lock. */
extern cl_event *cl_command_queue_record_in_queue_events(cl_command_queue queue, cl_uint *list_num);
#endif /* __CL_COMMAND_QUEUE_H__ */
|
