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
|
/*
* Copyright (c) 2016 University of Cordoba and University of Illinois
* All rights reserved.
*
* Developed by: IMPACT Research Group
* University of Cordoba and University of Illinois
* http://impact.crhc.illinois.edu/
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* with the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, 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:
*
* > Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimers.
* > Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimers in the
* documentation and/or other materials provided with the distribution.
* > Neither the names of IMPACT Research Group, University of Cordoba,
* University of Illinois nor the names of its contributors may be used
* to endorse or promote products derived from this Software without
* specific prior written permission.
*
* 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 NONINFRINGEMENT. IN NO EVENT SHALL THE
* CONTRIBUTORS OR COPYRIGHT HOLDERS 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 WITH
* THE SOFTWARE.
*
*/
#ifndef _PARTITIONER_H_
#define _PARTITIONER_H_
#ifndef _CUDA_COMPILER_
#include <iostream>
#endif
#if !defined(_CUDA_COMPILER_) && defined(CUDA_8_0)
#include <atomic>
#endif
// Partitioner definition -----------------------------------------------------
typedef struct Partitioner {
int n_tasks;
int cut;
int current;
#ifndef _CUDA_COMPILER_
int thread_id;
int n_threads;
#endif
#ifdef CUDA_8_0
// CUDA 8.0 support for dynamic partitioning
int strategy;
#ifdef _CUDA_COMPILER_
int *worklist;
int *tmp;
#else
std::atomic_int *worklist;
#endif
#endif
} Partitioner;
// Partitioning strategies
#define STATIC_PARTITIONING 0
#define DYNAMIC_PARTITIONING 1
// Create a partitioner -------------------------------------------------------
#ifdef _CUDA_COMPILER_
__device__
#endif
inline Partitioner partitioner_create(int n_tasks, float alpha
#ifndef _CUDA_COMPILER_
, int thread_id, int n_threads
#endif
#ifdef CUDA_8_0
#ifdef _CUDA_COMPILER_
, int *worklist
, int *tmp
#else
, std::atomic_int *worklist
#endif
#endif
) {
Partitioner p;
p.n_tasks = n_tasks;
#ifndef _CUDA_COMPILER_
p.thread_id = thread_id;
p.n_threads = n_threads;
#endif
if(alpha >= 0.0 && alpha <= 1.0) {
p.cut = p.n_tasks * alpha;
#ifdef CUDA_8_0
p.strategy = STATIC_PARTITIONING;
#endif
} else {
#ifdef CUDA_8_0
p.strategy = DYNAMIC_PARTITIONING;
p.worklist = worklist;
#ifdef _CUDA_COMPILER_
p.tmp = tmp;
#endif
#endif
}
return p;
}
// Partitioner iterators: first() ---------------------------------------------
#ifndef _CUDA_COMPILER_
inline int cpu_first(Partitioner *p) {
#ifdef CUDA_8_0
if(p->strategy == DYNAMIC_PARTITIONING) {
p->current = p->worklist->fetch_add(1);
} else
#endif
{
p->current = p->thread_id;
}
return p->current;
}
#else
__device__ inline int gpu_first(Partitioner *p) {
#ifdef CUDA_8_0
if(p->strategy == DYNAMIC_PARTITIONING) {
if(threadIdx.y == 0 && threadIdx.x == 0) {
p->tmp[0] = atomicAdd_system(p->worklist, 1);
}
__syncthreads();
p->current = p->tmp[0];
} else
#endif
{
p->current = p->cut + blockIdx.x;
}
return p->current;
}
#endif
// Partitioner iterators: more() ----------------------------------------------
#ifndef _CUDA_COMPILER_
inline bool cpu_more(const Partitioner *p) {
#ifdef CUDA_8_0
if(p->strategy == DYNAMIC_PARTITIONING) {
return (p->current < p->n_tasks);
} else
#endif
{
return (p->current < p->cut);
}
}
#else
__device__ inline bool gpu_more(const Partitioner *p) {
return (p->current < p->n_tasks);
}
#endif
// Partitioner iterators: next() ----------------------------------------------
#ifndef _CUDA_COMPILER_
inline int cpu_next(Partitioner *p) {
#ifdef CUDA_8_0
if(p->strategy == DYNAMIC_PARTITIONING) {
p->current = p->worklist->fetch_add(1);
} else
#endif
{
p->current = p->current + p->n_threads;
}
return p->current;
}
#else
__device__ inline int gpu_next(Partitioner *p) {
#ifdef CUDA_8_0
if(p->strategy == DYNAMIC_PARTITIONING) {
if(threadIdx.y == 0 && threadIdx.x == 0) {
p->tmp[0] = atomicAdd_system(p->worklist, 1);
}
__syncthreads();
p->current = p->tmp[0];
} else
#endif
{
p->current = p->current + gridDim.x;
}
return p->current;
}
#endif
#endif
|
