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
|
/*
* Copyright 2011-2015 Blender Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __KERNEL_QUEUE_H__
#define __KERNEL_QUEUE_H__
CCL_NAMESPACE_BEGIN
/*
* Queue utility functions for split kernel
*/
#ifdef __KERNEL_OPENCL__
#pragma OPENCL EXTENSION cl_khr_global_int32_base_atomics : enable
#pragma OPENCL EXTENSION cl_khr_local_int32_base_atomics : enable
#endif
/*
* Enqueue ray index into the queue
*/
ccl_device void enqueue_ray_index(
int ray_index, /* Ray index to be enqueued. */
int queue_number, /* Queue in which the ray index should be enqueued. */
ccl_global int *queues, /* Buffer of all queues. */
int queue_size, /* Size of each queue. */
ccl_global int *queue_index) /* Array of size num_queues; Used for atomic increment. */
{
/* This thread's queue index. */
int my_queue_index = atomic_fetch_and_inc_uint32((ccl_global uint*)&queue_index[queue_number])
+ (queue_number * queue_size);
queues[my_queue_index] = ray_index;
}
/*
* Get the ray index for this thread
* Returns a positive ray_index for threads that have to do some work;
* Returns 'QUEUE_EMPTY_SLOT' for threads that don't have any work
* i.e All ray's in the queue has been successfully allocated and there
* is no more ray to allocate to other threads.
*/
ccl_device int get_ray_index(
KernelGlobals *kg,
int thread_index, /* Global thread index. */
int queue_number, /* Queue to operate on. */
ccl_global int *queues, /* Buffer of all queues. */
int queuesize, /* Size of a queue. */
int empty_queue) /* Empty the queue slot as soon as we fetch the ray index. */
{
int ray_index = queues[queue_number * queuesize + thread_index];
if(empty_queue && ray_index != QUEUE_EMPTY_SLOT) {
queues[queue_number * queuesize + thread_index] = QUEUE_EMPTY_SLOT;
}
return ray_index;
}
/* The following functions are to realize Local memory variant of enqueue ray index function. */
/* All threads should call this function. */
ccl_device void enqueue_ray_index_local(
int ray_index, /* Ray index to enqueue. */
int queue_number, /* Queue in which to enqueue ray index. */
char enqueue_flag, /* True for threads whose ray index has to be enqueued. */
int queuesize, /* queue size. */
ccl_local_param unsigned int *local_queue_atomics, /* To to local queue atomics. */
ccl_global int *Queue_data, /* Queues. */
ccl_global int *Queue_index) /* To do global queue atomics. */
{
int lidx = ccl_local_id(1) * ccl_local_size(0) + ccl_local_id(0);
/* Get local queue id .*/
unsigned int lqidx;
if(enqueue_flag) {
lqidx = atomic_fetch_and_inc_uint32(local_queue_atomics);
}
ccl_barrier(CCL_LOCAL_MEM_FENCE);
/* Get global queue offset. */
if(lidx == 0) {
*local_queue_atomics = atomic_fetch_and_add_uint32((ccl_global uint*)&Queue_index[queue_number],
*local_queue_atomics);
}
ccl_barrier(CCL_LOCAL_MEM_FENCE);
/* Get global queue index and enqueue ray. */
if(enqueue_flag) {
unsigned int my_gqidx = queue_number * queuesize + (*local_queue_atomics) + lqidx;
Queue_data[my_gqidx] = ray_index;
}
}
ccl_device unsigned int get_local_queue_index(
int queue_number, /* Queue in which to enqueue the ray; -1 if no queue */
ccl_local_param unsigned int *local_queue_atomics)
{
int my_lqidx = atomic_fetch_and_inc_uint32(&local_queue_atomics[queue_number]);
return my_lqidx;
}
ccl_device unsigned int get_global_per_queue_offset(
int queue_number,
ccl_local_param unsigned int *local_queue_atomics,
ccl_global int* global_queue_atomics)
{
unsigned int queue_offset = atomic_fetch_and_add_uint32((ccl_global uint*)&global_queue_atomics[queue_number],
local_queue_atomics[queue_number]);
return queue_offset;
}
ccl_device unsigned int get_global_queue_index(
int queue_number,
int queuesize,
unsigned int lqidx,
ccl_local_param unsigned int * global_per_queue_offset)
{
int my_gqidx = queuesize * queue_number + lqidx + global_per_queue_offset[queue_number];
return my_gqidx;
}
ccl_device int dequeue_ray_index(
int queue_number,
ccl_global int *queues,
int queue_size,
ccl_global int *queue_index)
{
int index = atomic_fetch_and_dec_uint32((ccl_global uint*)&queue_index[queue_number])-1;
if(index < 0) {
return QUEUE_EMPTY_SLOT;
}
return queues[index + queue_number * queue_size];
}
CCL_NAMESPACE_END
#endif // __KERNEL_QUEUE_H__
|
