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
|
// Copyright 2009-2020 Intel Corporation
// SPDX-License-Identifier: Apache-2.0
#pragma once
#include "../tasking/taskscheduler.h"
#include "../sys/array.h"
#include "../math/math.h"
#include "../math/range.h"
#if defined(TASKING_GCD) && defined(BUILD_IOS)
#include <dispatch/dispatch.h>
#include <algorithm>
#include <type_traits>
#endif
namespace embree
{
/* parallel_for without range */
template<typename Index, typename Func>
__forceinline void parallel_for( const Index N, const Func& func)
{
#if defined(TASKING_INTERNAL)
if (N) {
TaskScheduler::spawn(Index(0),N,Index(1),[&] (const range<Index>& r) {
assert(r.size() == 1);
func(r.begin());
});
if (!TaskScheduler::wait())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
}
#elif defined(TASKING_GCD) && defined(BUILD_IOS)
const size_t baselineNumBlocks = (TaskScheduler::threadCount() > 1)? TaskScheduler::threadCount() : 1;
const size_t length = N;
const size_t blockSize = (length + baselineNumBlocks-1) / baselineNumBlocks;
const size_t numBlocks = (length + blockSize-1) / blockSize;
dispatch_apply(numBlocks, DISPATCH_APPLY_AUTO, ^(size_t currentBlock) {
const size_t start = (currentBlock * blockSize);
const size_t blockLength = std::min(length - start, blockSize);
const size_t end = start + blockLength;
for(size_t i=start; i < end; i++)
{
func(i);
}
});
#elif defined(TASKING_TBB)
#if TBB_INTERFACE_VERSION >= 12002
tbb::task_group_context context;
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
},context);
if (context.is_group_execution_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#else
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
});
if (tbb::task::self().is_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#endif
#elif defined(TASKING_PPL)
concurrency::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
});
#else
# error "no tasking system enabled"
#endif
}
/* parallel for with range and granulatity */
template<typename Index, typename Func>
__forceinline void parallel_for( const Index first, const Index last, const Index minStepSize, const Func& func)
{
assert(first <= last);
#if defined(TASKING_INTERNAL)
TaskScheduler::spawn(first,last,minStepSize,func);
if (!TaskScheduler::wait())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#elif defined(TASKING_GCD) && defined(BUILD_IOS)
const size_t baselineNumBlocks = (TaskScheduler::threadCount() > 1)? 4*TaskScheduler::threadCount() : 1;
const size_t length = last - first;
const size_t blockSizeByThreads = (length + baselineNumBlocks-1) / baselineNumBlocks;
size_t blockSize = std::max<size_t>(minStepSize,blockSizeByThreads);
blockSize += blockSize % 4;
const size_t numBlocks = (length + blockSize-1) / blockSize;
dispatch_apply(numBlocks, DISPATCH_APPLY_AUTO, ^(size_t currentBlock) {
const size_t start = first + (currentBlock * blockSize);
const size_t end = std::min<size_t>(last, start + blockSize);
func( embree::range<Index>(start,end) );
});
#elif defined(TASKING_TBB)
#if TBB_INTERFACE_VERSION >= 12002
tbb::task_group_context context;
tbb::parallel_for(tbb::blocked_range<Index>(first,last,minStepSize),[&](const tbb::blocked_range<Index>& r) {
func(range<Index>(r.begin(),r.end()));
},context);
if (context.is_group_execution_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#else
tbb::parallel_for(tbb::blocked_range<Index>(first,last,minStepSize),[&](const tbb::blocked_range<Index>& r) {
func(range<Index>(r.begin(),r.end()));
});
if (tbb::task::self().is_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#endif
#elif defined(TASKING_PPL)
concurrency::parallel_for(first, last, Index(1) /*minStepSize*/, [&](Index i) {
func(range<Index>(i,i+1));
});
#else
# error "no tasking system enabled"
#endif
}
/* parallel for with range */
template<typename Index, typename Func>
__forceinline void parallel_for( const Index first, const Index last, const Func& func)
{
assert(first <= last);
parallel_for(first,last,(Index)1,func);
}
#if defined(TASKING_TBB) && (TBB_INTERFACE_VERSION > 4001)
template<typename Index, typename Func>
__forceinline void parallel_for_static( const Index N, const Func& func)
{
#if TBB_INTERFACE_VERSION >= 12002
tbb::task_group_context context;
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
},tbb::simple_partitioner(),context);
if (context.is_group_execution_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#else
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
},tbb::simple_partitioner());
if (tbb::task::self().is_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#endif
}
typedef tbb::affinity_partitioner affinity_partitioner;
template<typename Index, typename Func>
__forceinline void parallel_for_affinity( const Index N, const Func& func, tbb::affinity_partitioner& ap)
{
#if TBB_INTERFACE_VERSION >= 12002
tbb::task_group_context context;
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
},ap,context);
if (context.is_group_execution_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#else
tbb::parallel_for(Index(0),N,Index(1),[&](Index i) {
func(i);
},ap);
if (tbb::task::self().is_cancelled())
// -- GODOT start --
// throw std::runtime_error("task cancelled");
abort();
// -- GODOT end --
#endif
}
#else
template<typename Index, typename Func>
__forceinline void parallel_for_static( const Index N, const Func& func)
{
parallel_for(N,func);
}
struct affinity_partitioner {
};
template<typename Index, typename Func>
__forceinline void parallel_for_affinity( const Index N, const Func& func, affinity_partitioner& ap)
{
parallel_for(N,func);
}
#endif
}
|