forked from taskflow/taskflow
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathfor_each.hpp
More file actions
169 lines (139 loc) · 4.75 KB
/
Copy pathfor_each.hpp
File metadata and controls
169 lines (139 loc) · 4.75 KB
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
#pragma once
#include "launch.hpp"
namespace tf {
// Function: make_for_each_task
template <typename B, typename E, typename C, typename P = GuidedPartitioner>
TF_FORCE_INLINE auto make_for_each_task(B beg, E end, C c, P&& part = P()) {
using B_t = std::decay_t<unwrap_ref_decay_t<B>>;
using E_t = std::decay_t<unwrap_ref_decay_t<E>>;
using namespace std::string_literals;
return [b=beg, e=end, c, part=std::forward<P>(part)] (Runtime& rt) mutable {
// fetch the stateful values
B_t beg = b;
E_t end = e;
size_t W = rt.executor().num_workers();
size_t N = std::distance(beg, end);
// only myself - no need to spawn another graph
if(W <= 1 || N <= part.chunk_size()) {
std::for_each(beg, end, c);
return;
}
if(N < W) {
W = N;
}
// static partitioner
if constexpr(std::is_same_v<std::decay_t<P>, StaticPartitioner>) {
size_t chunk_size;
for(size_t w=0, curr_b=0; w<W && curr_b < N; ++w, curr_b += chunk_size) {
chunk_size = part.adjusted_chunk_size(N, W, w);
launch_loop(W, w, rt, [=, &c, &part] () mutable {
part.loop(N, W, curr_b, chunk_size,
[&, prev_e=size_t{0}](size_t curr_b, size_t curr_e) mutable {
std::advance(beg, curr_b - prev_e);
for(size_t x = curr_b; x<curr_e; x++) {
c(*beg++);
}
prev_e = curr_e;
}
);
});
}
rt.join();
}
// dynamic partitioner
else {
std::atomic<size_t> next(0);
launch_loop(N, W, rt, next, part, [=, &c, &next, &part] () mutable {
part.loop(N, W, next,
[&, prev_e=size_t{0}](size_t curr_b, size_t curr_e) mutable {
std::advance(beg, curr_b - prev_e);
for(size_t x = curr_b; x<curr_e; x++) {
c(*beg++);
}
prev_e = curr_e;
}
);
});
}
};
}
// Function: make_for_each_index_task
template <typename B, typename E, typename S, typename C, typename P = GuidedPartitioner>
TF_FORCE_INLINE auto make_for_each_index_task(B beg, E end, S inc, C c, P&& part = P()){
using namespace std::string_literals;
using B_t = std::decay_t<unwrap_ref_decay_t<B>>;
using E_t = std::decay_t<unwrap_ref_decay_t<E>>;
using S_t = std::decay_t<unwrap_ref_decay_t<S>>;
return [b=beg, e=end, a=inc, c, part=std::forward<P>(part)]
(Runtime& rt) mutable {
// fetch the iterator values
B_t beg = b;
E_t end = e;
S_t inc = a;
size_t W = rt.executor().num_workers();
size_t N = distance(beg, end, inc);
// only myself - no need to spawn another graph
if(W <= 1 || N <= part.chunk_size()) {
for(size_t x=0; x<N; x++, beg+=inc) {
c(beg);
}
return;
}
if(N < W) {
W = N;
}
// static partitioner
if constexpr(std::is_same_v<std::decay_t<P>, StaticPartitioner>) {
size_t chunk_size;
for(size_t w=0, curr_b=0; w<W && curr_b < N; ++w, curr_b += chunk_size) {
chunk_size = part.adjusted_chunk_size(N, W, w);
launch_loop(W, w, rt, [=, &c, &part] () mutable {
part.loop(N, W, curr_b, chunk_size,
[&](size_t curr_b, size_t curr_e) {
auto idx = static_cast<B_t>(curr_b) * inc + beg;
for(size_t x=curr_b; x<curr_e; x++, idx += inc) {
c(idx);
}
}
);
});
}
rt.join();
}
// dynamic partitioner
else {
std::atomic<size_t> next(0);
launch_loop(N, W, rt, next, part, [=, &c, &next, &part] () mutable {
part.loop(N, W, next,
[&](size_t curr_b, size_t curr_e) {
auto idx = static_cast<B_t>(curr_b) * inc + beg;
for(size_t x=curr_b; x<curr_e; x++, idx += inc) {
c(idx);
}
}
);
});
}
};
}
// ----------------------------------------------------------------------------
// for_each
// ----------------------------------------------------------------------------
// Function: for_each
template <typename B, typename E, typename C, typename P>
Task FlowBuilder::for_each(B beg, E end, C c, P&& part) {
return emplace(
make_for_each_task(beg, end, c, std::forward<P>(part))
);
}
// ----------------------------------------------------------------------------
// for_each_index
// ----------------------------------------------------------------------------
// Function: for_each_index
template <typename B, typename E, typename S, typename C, typename P>
Task FlowBuilder::for_each_index(B beg, E end, S inc, C c, P&& part){
return emplace(
make_for_each_index_task(beg, end, inc, c, std::forward<P>(part))
);
}
} // end of namespace tf -----------------------------------------------------