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
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
|
/****************************************************************/
/* 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 _FULLY_DIST_SP_VEC_H_
#define _FULLY_DIST_SP_VEC_H_
#include <iostream>
#include <vector>
#include <utility>
#include "CommGrid.h"
#include "promote.h"
#include "SpParMat.h"
#include "FullyDist.h"
#include "Exception.h"
#include "OptBuf.h"
#include "CombBLAS.h"
namespace combblas {
template <class IT, class NT, class DER>
class SpParMat;
template <class IT>
class DistEdgeList;
template <class IU, class NU>
class FullyDistVec;
template <class IU, class NU>
class SparseVectorLocalIterator;
/**
* A sparse vector of length n (with nnz <= n of them being nonzeros) is distributed to
* "all the processors" in a way that "respects ordering" of the nonzero indices
* Example: x = [5,1,6,2,9] for nnz(x)=5 and length(x)=12
* we use 4 processors P_00, P_01, P_10, P_11
* Then P_00 owns [1,2] (in the range [0,...,2]), P_01 ow`ns [5] (in the range [3,...,5]), and so on.
* In the case of A(v,w) type sparse matrix indexing, this doesn't matter because n = nnz
* After all, A(v,w) will have dimensions length(v) x length (w)
* v and w will be of numerical type (NT) "int" and their indices (IT) will be consecutive integers
* It is possibly that nonzero counts are distributed unevenly
* Example: x=[1,2,3,4,5] and length(x) = 20, then P_00 would own all the nonzeros and the rest will hold empry vectors
* Just like in SpParMat case, indices are local to processors (they belong to range [0,...,length-1] on each processor)
* \warning Always create vectors with the right length, setting elements won't increase its length (similar to operator[] on std::vector)
**/
template <class IT, class NT>
class FullyDistSpVec: public FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>
{
public:
FullyDistSpVec ( );
explicit FullyDistSpVec ( IT glen );
FullyDistSpVec ( std::shared_ptr<CommGrid> grid);
FullyDistSpVec ( std::shared_ptr<CommGrid> grid, IT glen);
template <typename _UnaryOperation>
FullyDistSpVec (const FullyDistVec<IT,NT> & rhs, _UnaryOperation unop);
FullyDistSpVec (const FullyDistVec<IT,NT> & rhs); // Conversion copy-constructor
FullyDistSpVec (IT globalsize, const FullyDistVec<IT,IT> & inds, const FullyDistVec<IT,NT> & vals, bool SumDuplicates = false);
FullyDistSpVec (std::shared_ptr<CommGrid> grid, IT globallen, const std::vector<IT>& indvec, const std::vector<NT> & numvec, bool SumDuplicates = false, bool sorted=false);
IT NnzUntil() const;
FullyDistSpVec<IT,NT> Invert (IT globallen);
template <typename _BinaryOperationIdx, typename _BinaryOperationVal, typename _BinaryOperationDuplicate>
FullyDistSpVec<IT,NT> Invert (IT globallen, _BinaryOperationIdx __binopIdx, _BinaryOperationVal __binopVal, _BinaryOperationDuplicate __binopDuplicate);
template <typename _BinaryOperationIdx, typename _BinaryOperationVal>
FullyDistSpVec<IT,NT> InvertRMA (IT globallen, _BinaryOperationIdx __binopIdx, _BinaryOperationVal __binopVal);
template <typename NT1, typename _UnaryOperation>
void Select (const FullyDistVec<IT,NT1> & denseVec, _UnaryOperation unop);
template <typename _UnaryOperation>
void FilterByVal (FullyDistSpVec<IT,IT> Selector, _UnaryOperation __unop, bool filterByIndex);
template <typename NT1>
void Setminus (const FullyDistSpVec<IT,NT1> & other);
//template <typename NT1, typename _UnaryOperation>
//void Set (FullyDistSpVec<IT,NT1> Selector, _UnaryOperation __unop);
template <typename NT1, typename _UnaryOperation, typename _BinaryOperation>
void SelectApply (const FullyDistVec<IT,NT1> & denseVec, _UnaryOperation __unop, _BinaryOperation __binop);
//! like operator=, but instead of making a deep copy it just steals the contents.
//! Useful for places where the "victim" will be distroyed immediately after the call.
void stealFrom(FullyDistSpVec<IT,NT> & victim);
FullyDistSpVec<IT,NT> & operator=(const FullyDistSpVec< IT,NT > & rhs);
FullyDistSpVec<IT,NT> & operator=(const FullyDistVec< IT,NT > & rhs); // convert from dense
FullyDistSpVec<IT,NT> & operator=(NT fixedval) // assign fixed value
{
#ifdef _OPENMP
#pragma omp parallel for
#endif
for(size_t i=0; i < ind.size(); ++i)
num[i] = fixedval;
return *this;
}
FullyDistSpVec<IT,NT> & operator+=(const FullyDistSpVec<IT,NT> & rhs);
FullyDistSpVec<IT,NT> & operator-=(const FullyDistSpVec<IT,NT> & rhs);
class ScalarReadSaveHandler
{
public:
NT getNoNum(IT index) { return static_cast<NT>(1); }
template <typename c, typename t>
NT read(std::basic_istream<c,t>& is, IT index)
{
NT v;
is >> v;
return v;
}
template <typename c, typename t>
void save(std::basic_ostream<c,t>& os, const NT& v, IT index)
{
os << v;
}
};
template <class HANDLER>
void ParallelWrite(const std::string & filename, bool onebased, HANDLER handler, bool includeindices = true, bool includeheader = false);
void ParallelWrite(const std::string & filename, bool onebased, bool includeindices = true) { ParallelWrite(filename, onebased, ScalarReadSaveHandler(), includeindices); };
template <typename _BinaryOperation>
void ParallelRead (const std::string & filename, bool onebased, _BinaryOperation BinOp);
//! Totally obsolete version that only accepts an ifstream object and ascii files
template <class HANDLER>
std::ifstream& ReadDistribute (std::ifstream& infile, int master, HANDLER handler);
std::ifstream& ReadDistribute (std::ifstream& infile, int master) { return ReadDistribute(infile, master, ScalarReadSaveHandler()); }
template <class HANDLER>
void SaveGathered(std::ofstream& outfile, int master, HANDLER handler, bool printProcSplits = false);
void SaveGathered(std::ofstream& outfile, int master) { SaveGathered(outfile, master, ScalarReadSaveHandler()); }
template <typename NNT> operator FullyDistSpVec< IT,NNT > () const //!< Type conversion operator
{
FullyDistSpVec<IT,NNT> CVT(commGrid);
CVT.ind = std::vector<IT>(ind.begin(), ind.end());
CVT.num = std::vector<NNT>(num.begin(), num.end());
CVT.glen = glen;
return CVT;
}
bool operator==(const FullyDistSpVec<IT,NT> & rhs) const
{
FullyDistVec<IT,NT> v = *this;
FullyDistVec<IT,NT> w = rhs;
return (v == w);
}
void PrintInfo(std::string vecname) const;
void iota(IT globalsize, NT first);
void nziota(NT first);
FullyDistVec<IT,NT> operator() (const FullyDistVec<IT,IT> & ri) const; //!< SpRef (expects ri to be 0-based)
void SetElement (IT indx, NT numx); // element-wise assignment
void DelElement (IT indx); // element-wise deletion
NT operator[](IT indx);
bool WasFound() const { return wasFound; }
//! sort the vector itself, return the permutation vector (0-based)
FullyDistSpVec<IT, IT> sort();
#if __cplusplus > 199711L
template <typename _BinaryOperation = minimum<NT> >
FullyDistSpVec<IT, NT> Uniq(_BinaryOperation __binary_op = _BinaryOperation(), MPI_Op mympiop = MPI_MIN);
#else
template <typename _BinaryOperation >
FullyDistSpVec<IT, NT> Uniq(_BinaryOperation __binary_op, MPI_Op mympiop);
#endif
// Aydin TODO: parallelize with OpenMP
template <typename _UnaryOperation>
FullyDistSpVec<IT,NT> Prune(_UnaryOperation __unary_op, bool inPlace = true) //<! Prune any nonzero entries for which the __unary_op evaluates to true (solely based on value)
{
FullyDistSpVec<IT,NT> temp(commGrid);
IT spsize = ind.size();
for(IT i=0; i< spsize; ++i)
{
if(!(__unary_op(num[i]))) // keep this nonzero
{
temp.ind.push_back(ind[i]);
temp.num.push_back(num[i]);
}
}
if (inPlace)
{
ind.swap(temp.ind);
ind.swap(temp.num);
return FullyDistSpVec<IT,NT>(commGrid); // return blank to match signature
}
else
{
return temp;
}
}
IT getlocnnz() const
{
return ind.size();
}
IT getnnz() const
{
IT totnnz = 0;
IT locnnz = ind.size();
MPI_Allreduce( &locnnz, &totnnz, 1, MPIType<IT>(), MPI_SUM, commGrid->GetWorld());
return totnnz;
}
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::LengthUntil;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::MyLocLength;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::MyRowLength;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::TotalLength;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::Owner;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::RowLenUntil;
void setNumToInd()
{
IT offset = LengthUntil();
IT spsize = ind.size();
#ifdef _OPENMP
#pragma omp parallel for
#endif
for(IT i=0; i< spsize; ++i)
num[i] = ind[i] + offset;
}
template <typename _Predicate>
IT Count(_Predicate pred) const; //!< Return the number of elements for which pred is true
template <typename _UnaryOperation>
void Apply(_UnaryOperation __unary_op)
{
//transform(num.begin(), num.end(), num.begin(), __unary_op);
IT spsize = num.size();
#ifdef _OPENMP
#pragma omp parallel for
#endif
for(IT i=0; i < spsize; ++i)
num[i] = __unary_op(num[i]);
}
template <typename _BinaryOperation>
void ApplyInd(_BinaryOperation __binary_op)
{
IT offset = LengthUntil();
IT spsize = ind.size();
#ifdef _OPENMP
#pragma omp parallel for
#endif
for(IT i=0; i < spsize; ++i)
num[i] = __binary_op(num[i], ind[i] + offset);
}
template <typename _BinaryOperation>
NT Reduce(_BinaryOperation __binary_op, NT init) const;
template <typename OUT, typename _BinaryOperation, typename _UnaryOperation>
OUT Reduce(_BinaryOperation __binary_op, OUT default_val, _UnaryOperation __unary_op) const;
void DebugPrint();
std::shared_ptr<CommGrid> getcommgrid() const { return commGrid; }
void Reset();
NT GetLocalElement(IT indx);
void BulkSet(IT inds[], int count);
std::vector<IT> GetLocalInd (){std::vector<IT> rind = ind; return rind;};
std::vector<NT> GetLocalNum (){std::vector<NT> rnum = num; return rnum;};
template <typename _Predicate>
FullyDistVec<IT,IT> FindInds(_Predicate pred) const;
template <typename _Predicate>
FullyDistVec<IT,NT> FindVals(_Predicate pred) const;
protected:
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::glen;
using FullyDist<IT,NT,typename combblas::disable_if< combblas::is_boolean<NT>::value, NT >::type>::commGrid;
private:
std::vector< IT > ind; // ind.size() give the number of nonzeros
std::vector< NT > num;
bool wasFound; // true if the last GetElement operation returned an actual value
template <typename _BinaryOperation>
void SparseCommon(std::vector< std::vector < std::pair<IT,NT> > > & data, _BinaryOperation BinOp);
#if __cplusplus > 199711L
template <typename _BinaryOperation = minimum<NT> >
FullyDistSpVec<IT, NT> UniqAll2All(_BinaryOperation __binary_op = _BinaryOperation(), MPI_Op mympiop = MPI_MIN);
#else
template <typename _BinaryOperation >
FullyDistSpVec<IT, NT> UniqAll2All(_BinaryOperation __binary_op, MPI_Op mympiop);
#endif
template <class IU, class NU>
friend class FullyDistSpVec;
template <class IU, class NU>
friend class FullyDistVec;
template <class IU, class NU, class UDER>
friend class SpParMat;
template <class IU, class NU>
friend class SparseVectorLocalIterator;
template <typename SR, typename IU, typename NUM, typename NUV, typename UDER>
friend FullyDistSpVec<IU,typename promote_trait<NUM,NUV>::T_promote>
SpMV (const SpParMat<IU,NUM,UDER> & A, const FullyDistSpVec<IU,NUV> & x );
template <typename SR, typename IU, typename NUM, typename UDER>
friend FullyDistSpVec<IU,typename promote_trait<NUM,IU>::T_promote>
SpMV (const SpParMat<IU,NUM,UDER> & A, const FullyDistSpVec<IU,IU> & x, bool indexisvalue);
template <typename VT, typename IU, typename UDER> // NoSR version (in BFSFriends.h)
friend FullyDistSpVec<IU,VT> SpMV (const SpParMat<IU,bool,UDER> & A, const FullyDistSpVec<IU,VT> & x, OptBuf<int32_t, VT > & optbuf);
template <typename SR, typename IVT, typename OVT, typename IU, typename NUM, typename UDER>
friend void SpMV (const SpParMat<IU,NUM,UDER> & A, const FullyDistSpVec<IU,IVT> & x, FullyDistSpVec<IU,OVT> & y,bool indexisvalue, OptBuf<int32_t, OVT > & optbuf);
template <typename SR, typename IVT, typename OVT, typename IU, typename NUM, typename UDER>
friend void SpMV (const SpParMat<IU,NUM,UDER> & A, const FullyDistSpVec<IU,IVT> & x, FullyDistSpVec<IU,OVT> & y,bool indexisvalue, OptBuf<int32_t, OVT > & optbuf, PreAllocatedSPA<OVT> & SPA);
template <typename IU, typename NU1, typename NU2>
friend FullyDistSpVec<IU,typename promote_trait<NU1,NU2>::T_promote>
EWiseMult (const FullyDistSpVec<IU,NU1> & V, const FullyDistVec<IU,NU2> & W , bool exclude, NU2 zero);
template <typename RET, typename IU, typename NU1, typename NU2, typename _BinaryOperation, typename _BinaryPredicate>
friend FullyDistSpVec<IU,RET>
EWiseApply (const FullyDistSpVec<IU,NU1> & V, const FullyDistVec<IU,NU2> & W , _BinaryOperation _binary_op, _BinaryPredicate _doOp, bool allowVNulls, NU1 Vzero, const bool useExtendedBinOp);
template <typename RET, typename IU, typename NU1, typename NU2, typename _BinaryOperation, typename _BinaryPredicate>
friend FullyDistSpVec<IU,RET>
EWiseApply_threaded (const FullyDistSpVec<IU,NU1> & V, const FullyDistVec<IU,NU2> & W , _BinaryOperation _binary_op, _BinaryPredicate _doOp, bool allowVNulls, NU1 Vzero, const bool useExtendedBinOp);
template <typename RET, typename IU, typename NU1, typename NU2, typename _BinaryOperation, typename _BinaryPredicate>
friend FullyDistSpVec<IU,RET>
EWiseApply (const FullyDistSpVec<IU,NU1> & V, const FullyDistSpVec<IU,NU2> & W , _BinaryOperation _binary_op, _BinaryPredicate _doOp, bool allowVNulls, bool allowWNulls, NU1 Vzero, NU2 Wzero, const bool allowIntersect, const bool useExtendedBinOp);
template <typename IU>
friend void RandPerm(FullyDistSpVec<IU,IU> & V); // called on an existing object, randomly permutes it
template <typename IU>
friend void RenameVertices(DistEdgeList<IU> & DEL);
//! Helper functions for sparse matrix X sparse vector
// Ariful: I made this an internal function in ParFriends.h
//template <typename SR, typename IU, typename OVT>
//friend void MergeContributions(FullyDistSpVec<IU,OVT> & y, int * & recvcnt, int * & rdispls, int32_t * & recvindbuf, OVT * & recvnumbuf, int rowneighs);
template <typename IU, typename VT>
friend void MergeContributions(FullyDistSpVec<IU,VT> & y, int * & recvcnt, int * & rdispls, int32_t * & recvindbuf, VT * & recvnumbuf, int rowneighs);
template<typename IU, typename NV>
friend void TransposeVector(MPI_Comm & World, const FullyDistSpVec<IU,NV> & x, int32_t & trxlocnz, IU & lenuntil, int32_t * & trxinds, NV * & trxnums, bool indexisvalue);
template <class IU, class NU, class DER, typename _UnaryOperation>
friend SpParMat<IU, bool, DER> PermMat1 (const FullyDistSpVec<IU,NU> & ri, const IU ncol, _UnaryOperation __unop);
};
}
#include "FullyDistSpVec.cpp"
#endif
|
