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
|
/****************************************************************/
/* Parallel Combinatorial BLAS Library (for Graph Computations) */
/* version 1.6 -------------------------------------------------*/
/* date: 6/15/2017 ---------------------------------------------*/
/* authors: Ariful Azad, Aydin Buluc --------------------------*/
/****************************************************************/
/*
Copyright (c) 2010-2017, The Regents of the University of California
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in 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:
The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.
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
AUTHORS 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 IN
THE SOFTWARE.
*/
#ifndef _PRE_ALLOCATED_SPA_H
#define _PRE_ALLOCATED_SPA_H
#include "BitMap.h"
namespace combblas {
/**
* This special data structure is used for optimizing BFS iterations
* by providing a pre-allocated SPA data structure
*/
template <class OVT > // output value type
class PreAllocatedSPA
{
public:
PreAllocatedSPA():initialized(false) {}; // hide default constructor
template <class LMAT>
PreAllocatedSPA(LMAT & A):initialized(true) // the one and only constructor
{
int64_t mA = A.getnrow();
if( A.getnsplit() > 0) // multithreaded
{
int64_t perpiece = mA / A.getnsplit();
for(int i=0; i<A.getnsplit(); ++i)
{
if(i != A.getnsplit()-1)
{
V_isthere.push_back(BitMap(perpiece));
V_localy.push_back(std::vector<OVT>(perpiece));
std::vector<bool> isthere(perpiece, false);
for(auto colit = A.begcol(i); colit != A.endcol(i); ++colit)
{
for(auto nzit = A.begnz(colit,i); nzit != A.endnz(colit,i); ++nzit)
{
size_t rowid = nzit.rowid();
if(!isthere[rowid]) isthere[rowid] = true;
}
}
size_t maxvector = std::count(isthere.begin(), isthere.end(), true);
V_inds.push_back(std::vector<uint32_t>(maxvector));
}
else
{
V_isthere.push_back(BitMap(mA - i*perpiece));
V_localy.push_back(std::vector<OVT>(mA - i*perpiece));
std::vector<bool> isthere(mA - i*perpiece, false);
for(auto colit = A.begcol(i); colit != A.endcol(i); ++colit)
{
for(auto nzit = A.begnz(colit,i); nzit != A.endnz(colit,i); ++nzit)
{
size_t rowid = nzit.rowid();
if(!isthere[rowid]) isthere[rowid] = true;
}
}
size_t maxvector = std::count(isthere.begin(), isthere.end(), true);
V_inds.push_back(std::vector<uint32_t>(maxvector));
}
}
}
else // single threaded
{
V_isthere.push_back(BitMap(mA));
V_localy.push_back(std::vector<OVT>(mA));
std::vector<bool> isthere(mA, false);
for(auto colit = A.begcol(); colit != A.endcol(); ++colit)
{
for(auto nzit = A.begnz(colit); nzit != A.endnz(colit); ++nzit)
{
size_t rowid = nzit.rowid();
if(!isthere[rowid]) isthere[rowid] = true;
}
}
size_t maxvector = std::count(isthere.begin(), isthere.end(), true);
V_inds.push_back(std::vector<uint32_t>(maxvector));
}
};
// for manual splitting. just a hack. need to be fixed
template <class LMAT>
PreAllocatedSPA(LMAT & A, int splits):initialized(true)
{
buckets = splits;
int64_t mA = A.getnrow();
V_isthere.push_back(BitMap(mA));
V_localy.push_back(std::vector<OVT>(mA));
V_inds.push_back(std::vector<uint32_t>(mA)); // for better indexing among threads
std::vector<int32_t> nnzSplitA(buckets,0);
int32_t rowPerSplit = mA / splits;
//per thread because writing vector<bool> is not thread safe
for(int i=0; i<splits-1; i++)
V_isthereBool.push_back(std::vector<bool>(rowPerSplit));
V_isthereBool.push_back(std::vector<bool>(mA - (splits-1)*rowPerSplit));
//vector<bool> isthere(mA, false);
for(auto colit = A.begcol(); colit != A.endcol(); ++colit)
{
for(auto nzit = A.begnz(colit); nzit != A.endnz(colit); ++nzit)
{
size_t rowid = nzit.rowid();
//if(!isthere[rowid]) isthere[rowid] = true;
size_t splitId = (rowid/rowPerSplit > splits-1) ? splits-1 : rowid/rowPerSplit;
nnzSplitA[splitId]++;
}
}
// prefix sum
disp.resize(splits+1);
disp[0] = 0;
for(int i=0; i<splits; i++)
{
disp[i+1] = disp[i] + nnzSplitA[i];
}
indSplitA.resize(disp[splits]);
numSplitA.resize(disp[splits]);
};
// initialize an uninitialized SPA
template <class LMAT>
void Init(LMAT & A, int splits) // not done for DCSC matrices with A.getnsplit()
{
if(!initialized)
{
initialized = true;
buckets = splits;
int64_t mA = A.getnrow();
V_isthere.push_back(BitMap(mA));
V_localy.push_back(std::vector<OVT>(mA));
V_inds.push_back(std::vector<uint32_t>(mA)); // for better indexing among threads
std::vector<int32_t> nnzSplitA(buckets,0);
int32_t rowPerSplit = mA / splits;
for(int i=0; i<splits-1; i++)
V_isthereBool.push_back(std::vector<bool>(rowPerSplit));
V_isthereBool.push_back(std::vector<bool>(mA - (splits-1)*rowPerSplit));
//vector<bool> isthere(mA, false);
for(auto colit = A.begcol(); colit != A.endcol(); ++colit)
{
for(auto nzit = A.begnz(colit); nzit != A.endnz(colit); ++nzit)
{
size_t rowid = nzit.rowid();
//if(!isthere[rowid]) isthere[rowid] = true;
size_t splitId = (rowid/rowPerSplit > splits-1) ? splits-1 : rowid/rowPerSplit;
nnzSplitA[splitId]++;
}
}
// prefix sum
disp.resize(splits+1);
disp[0] = 0;
for(int i=0; i<splits; i++)
{
disp[i+1] = disp[i] + nnzSplitA[i];
}
indSplitA.resize(disp[splits]);
numSplitA.resize(disp[splits]);
}
};
int buckets; // number of buckets
std::vector< std::vector<uint32_t> > V_inds; // ABAB: is this big enough?
std::vector< BitMap > V_isthere;
std::vector< std::vector<bool> > V_isthereBool; // for thread safe access
std::vector< std::vector<OVT> > V_localy;
bool initialized;
std::vector<int32_t> indSplitA;
std::vector<OVT> numSplitA;
std::vector<uint32_t> disp;
};
}
#endif
|
